PHD Thesis

Reliability Assessment of Ageing Distribution
Cable for Replacement in Smart Distribution

Systems
A Thesis submitted to The University of Manchester for the Degree of
Doctor of Philosophy
in the Faculty of Science and Engineering
School of Electrical and Electronic Engineering
2016
Muhammad Buhari
Electrical Energy and Power Division
0
1
0
Table of Contents
List of Figures............................................................................................................... 5
List of Tables ................................................................................................................ 7
List of Abbreviations ................................................................................................... 9
List of Symbols ........................................................................................................... 12
Abstract....................................................................................................................... 14
Chapter one ................................................................................................................ 22
Introduction................................................................................................................ 22
1 Power system reliability studies ...................................................................... 22
1.1 Distribution network reliability ................................................................. 24
1.2 Reliability evaluation methods .................................................................. 26
1.3 Monte Carlo simulation (MCS) ................................................................. 28
1.4 Research motivation .................................................................................. 33
1.5 Research question ...................................................................................... 35
1.6 Research aim and objectives ..................................................................... 35
1.7 Contributions ............................................................................................. 37
1.8 Thesis structure .......................................................................................... 38
Chapter two ................................................................................................................ 42
Thermal modelling of ageing cables ......................................................................... 42
2 Modelling of ageing cables .............................................................................. 42
1
2.1 Common cables for medium voltage distribution systems ........................ 43
2.2 Brief historical overview of studies on conductor thermal ratings ............ 43
2.3 Review of cable thermal rating standards .................................................. 45
2.4 Arrhenius model of ageing cables ............................................................. 55
2.5 Proposed method for the determination of Arrhenius parameters ............. 56
2.6 Cable thermal loss of life model ................................................................ 59
2.7 Chapter summary ....................................................................................... 61
Chapter three.............................................................................................................. 62
Reliability assessment of ageing cables .................................................................... 62
3 Reliability assessment considering end-of-life failure ..................................... 62
3.1 Literature review ........................................................................................ 62
3.2 Basic features of reliability distribution functions ..................................... 68
3.3 Exponential probability distribution function ............................................ 70
3.4 Weibull probability distribution function .................................................. 71
3.5 Weibull renewal process ............................................................................ 73
3.6 Developed IEC-Arrhenius-Weibull reliability model for ageing cable ..... 74
3.7 Chapter summary ....................................................................................... 78
Chapter four ............................................................................................................... 80
Cable ageing study ..................................................................................................... 80
4 Details of SMCS model ................................................................................... 80
2
4.1 Sampling time to next transition................................................................ 81
4.2 Coefficient of variation (CoV) .................................................................. 83
4.3 Simulation methodology ........................................................................... 84
4.4 Test systems............................................................................................... 91
4.5 Study cases and presentation of results ................................................... 100
4.6 Chapter summary..................................................................................... 105
Chapter five .............................................................................................................. 108
Investment planning ................................................................................................ 108
5 Distribution system investment planning ...................................................... 108
5.1 Reinforcement planning (load related planning) ..................................... 109
5.2 Replacement planning ............................................................................. 110
5.3 Regulatory investment constraints .......................................................... 111
5.4 Ranking for cable replacement ................................................................ 112
5.5 Proposed cable ranking index .................................................................. 116
5.6 Cable ranking for replacement planning (based on the first proposed
ranking scheme) ................................................................................................. 118
5.7 Chapter summary..................................................................................... 121
Chapter six................................................................................................................ 122
Smart distribution systems ..................................................................................... 122
6 Smart power networks ................................................................................... 122
6.1 Smart measures to reduce replacement cost ............................................ 124
3
6.2 Optimal wind-DG integration .................................................................. 125
6.3 Smart operations (network reconfiguration) ............................................ 142
6.4 Smart grid planning study ........................................................................ 150
6.5 Details of SMCS approach for smart grid planning studies .................... 151
6.6 Proposed SMCS model with optimal wind-DG integration and network
reconfiguration ................................................................................................... 152
6.7 Results of optimal integration of wind-DG and network reconfiguration to

reduce cable ageing and replacement ................................................................. 155
6.8 Chapter summary ..................................................................................... 162
Chapter seven ........................................................................................................... 164
Conclusions and future work .................................................................................. 164
7 Conclusions .................................................................................................... 164
7.1 Future work .............................................................................................. 166
Appendix ................................................................................................................... 168
References ................................................................................................................. 174
List of Publications................................................................................................... 188
4
LIST OF FIGURES
Figure 1-1: Hierarchical levels of reliability studies in power systems ....................... 23
Figure 1-2: Reliability assessment approaches (thesis approaches in bold) ................ 27
Figure 1-3: Network Assets management planning map [23] ..................................... 31
Figure 2-1: Thermoelectric equivalent (TEE) model of a typical cable. ..................... 49
Figure 2-2: Underground cable geometry.................................................................... 53
Figure 2-3: Age distribution and life estimates of underground cables [6] ................. 57
Figure 2-4: Probabilities associated with normal distribution ..................................... 57
Figure 3-1: Bathtub Curve ........................................................................................... 63
Figure 3-2: Effects of different failure rates on exponential pdf ................................. 71
Figure 3-3: Effects of different failure rates on Weibull pdf ....................................... 73
Figure 3-4: A graph of stochastic point process .......................................................... 75
Figure 3-5: Minimal and general repair ....................................................................... 76
Figure 4-1: SMCS model with non-ageing elements................................................... 85
Figure 4-2: Flow chart of the SMCS for cable ageing analysis ................................... 90
Figure 4-3: IEEE 33 bus single feeder test system ...................................................... 93
Figure 4-4: IEEE 69 bus single feeder test system ...................................................... 95
Figure 4-5: Typical load profile (week 1) for bus 2 of the 33 bus test system ............ 97
Figure 4-6: Time varying soil temperatures used in the SMCS over one year ............ 99
Figure 4-7: Normalized Arrhenius thermal graph ..................................................... 101
5
Figure 4-8: Cable ranking based on thermal loss of life in the historic period: ......... 105
Figure 5-1: Ranking of critical cables for replacement (a) ∆𝑆𝐴𝐼𝐷𝐼 33 Bus system (b)
∆𝑆𝐴𝐼𝐹𝐼 33 Bus system (c) ∆𝑆𝐴𝐼𝐷𝐼 69 Bus system (d) ∆𝑆𝐴𝐼𝐹𝐼 69 Bus system ....... 118
Figure 5-2: First proposed cable ranking scheme for replacement: ........................... 120
Figure 6-1: Typical wind generator output characteristic .......................................... 133
Figure 6-2: Distribution networks optimization - literature review ........................... 145
Figure 6-3: Flowchart of the global methodology for cable replacement in the
presence of wind-DG and network reconfiguration ................................................... 153
Figure 6-4: Flow chart of the SMCS for optimal wind-DG integration and dynamic
network reconfiguration ............................................................................................. 154
Figure 6-5: Measured and 8-hour ahead forecast wind speeds using the ARMA (4,3)
model .......................................................................................................................... 157
6
LIST OF TABLES
Table 1-1: Comparison of analytical and simulation approaches ................................ 28
Table 4-1: Reliability data for main network components [89]................................... 94
Table 4-2: Cable data ................................................................................................... 97
Table 4-3: Time varying soil thermal resistivity ......................................................... 99
Table 4-4: Estimated Arrhenius parameters for different cable types ....................... 100
Table 4-5: Test systems reliability indices ................................................................ 102
Table 4-6: 33 bus network cables ranking list (italicised values indicate critical
ageing) ....................................................................................................................... 103
ageing) ....................................................................................................................... 104
Table 5-1: Losses vs replacement costs for MV cable .............................................. 116
Table 5-2: Cost assumptions ...................................................................................... 119
Table 6-1: Common wind speed forecasting methods............................................... 131
Table 6-2: CPU time and simulation duration for 33 and 69 buses test systems ...... 158
Table 6-3: DG connection, reinforcements, and investment costs for three strategies
................................................................................................................................... 158
Table 6-4: Reliability indices for two strategies ........................................................ 159
Table 6-5: Total ‘operation’ costs for two strategies ................................................. 160
Table 6-6: Cable ranking based on thermal loss-of-life ............................................ 160
Table 6-7: Cable ranking based on: a) ‘Direct’ Cost, b) ‘Total’ Cost ....................... 161
7
8
LIST OF ABBREVIATIONS
AC Alternating Current
ARMA Auto Regressive and Moving Averages
CI Customer Interruptions
CIGRE International Council on Large Electric Systems
CML Customer Minutes Lost
DG Distributed Generation
DNO Distribution Network Operators
DNR Dynamic Network Reconfiguration
ENS Energy Not Served
EPR Ethylene Propylene Rubber
IEC International Electrotechnical Commission
IEEE Institute of Electrical and Electronics Engineering
kM kilometres
LCT Low Carbon Technology
MCS Monte Carlo Simulation
MTTF Mean Time to failure
MTTR Mean Time to Repair
MV Medium Voltage
MVA Mega Volt Ampere
MW Mega Watts
OFGEM Office of Gas and Electricity Markets
OPF Optimal Power Flow
PILC Paper Insulation fluid filled Lead Covered
PLP Power Law Process
RTTR Real Time Thermal Rating
SAIDI System Average Interruption Duration Index
SAIFI System Average Interruption Frequency Index
SG Smart Grid
SMCS Sequential Monte Carlo Simulation
TVTR ` Time Varying Thermal Rating
UGC Underground Cable
WTG Wind Turbine Generator
XLPE Cross-linked Polyethylene insulators
9
10
LIST OF SYMBOLS
Sets
Ω𝐷𝐺 Set of DG units
Ω𝑛𝑜𝑑 , Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 Sets of nodes and potential nodes for DG unit k
Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘 Set of potential ‘DG branches’ for location nk
Ω𝑏𝑟 Set of existing branches
Ω𝑠𝑢𝑏 Set of primary substations
Parameters
𝐴, 𝐵 Parameters of the Arrhenius model
𝑡𝑝 , 𝑡𝑞 Duration of historic period p planning period q
𝑐𝑜𝑠𝑡𝑛𝑛𝑒𝑤 𝑘𝑖
Cost to construct new ‘DG branch’ nki
𝑟𝑒𝑖𝑛
𝑐𝑜𝑠𝑡𝑖𝑗 Cost to reinforce existing branch ij
𝑆𝑖𝑗𝑟𝑎𝑡 , 𝐼𝑖𝑗𝑟𝑎𝑡 Nominal rating of branch ij in MVA and A
𝑃𝐷𝐺𝑖 , 𝑄𝐷𝐺𝑖 Active and reactive power supplied by DG, node i
𝑃𝐷𝑖 , 𝑄𝐷𝑖 Active and reactive demand at node i
𝑒𝑥 𝑟𝑒𝑖 𝑛𝑒𝑤
𝐼̅𝑖𝑗 , 𝐼̅𝑖𝑗 , ̅̅̅̅̅
𝑘
𝐼𝑛 𝑖
Rating of existing, reinforced and new branch
𝑚𝑖𝑛 𝑚𝑎𝑥
𝑉𝑖 , 𝑉𝑖 Minimum/ maximum voltage limits at node i
𝑁𝐷𝐺 , 𝑁𝑏 Number of DG units and existing buses
𝑐𝑜𝑠𝑡𝐸𝐸 , 𝑐𝑜𝑠𝑡𝐸𝑁𝑆 Cost of delivered and non-delivered kWh
𝑅𝑖𝑗 Resistance of branch ij
𝐾𝑖 Fictitious load of 1.0 at DG nodes; 0 otherwise
𝐶𝐼𝐼𝑛𝑐 , 𝐶𝑀𝐿𝐼𝑛𝑐 Incentive rates for CIs and CMLs
𝑁𝑂𝑓𝑎𝑖 , 𝐶𝑓𝑎𝑖𝑙 Number of failures and cost per failure
is the thermal resistance per unit length between a conductor and
𝑇1
the sheath (K m/W);
is the thermal resistance per unit length of the bedding between
𝑇2 cable sheath and armour (K m/W). it takes a value of zero if the
cable does not have an armour;
is the thermal resistance per unit length of the external serving (or
𝑇3
jacket) of the cable (K m/W);
is the thermal resistance between the cable surface and the
𝑇4 surrounding medium which can include the duct and the
soil/backfill (K m/W);
is the ratio of losses in the metal sheath to total losses in all
𝜆1
conductors of the cable;
is the ratio of losses in the armouring to total losses in all
𝜆2
conductors of the cable.
is the number of equally sized load-carrying conductors in the cable
n
that carry the same load;
𝑈𝑜 is the phase to earth voltage (volts);
Functions/Variables
𝐿(∙), 𝜃𝑐 , 𝜃𝑠 Life measure, conductor temperature and soil temperature
𝐿𝑇 , ∆𝐿𝑇 (∙) Total and hourly thermal loss-of-life
𝐿𝑝 (∙), 𝐿𝑞 (∙) Life in historic period p and planning period q
𝜃𝑐,𝑝 , 𝜃𝑐,𝑞 Conductor temperature in historic period p and planning period q
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 Initial investment cost for DG connection
12
𝑇𝑂𝑇
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 DG investment cost including loss-of-life cost
𝑦𝑛𝑘 Binary decision variable for nodal location of DG k
Binary construction variable to reinforce existing branch ij and to
𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 , 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖 construct new ‘DG branch’ nki
𝐿′′ 𝑇𝑖𝑗 Thermal loss-of-life of cable section ij in planning period
Binary operation variable indicating whether branch ij or nki is
𝑜𝑐𝑖𝑗 , 𝑜𝑐𝑛𝑘𝑖
closed (=1) or open (=0)
Apparent, active, reactive power and current flow in branch ij –
𝑆𝑖𝑗 , 𝑃𝑖𝑗 , 𝑄𝑖𝑗 , 𝐼𝑖𝑗
functions of voltages and angles
Binary decision variable indicating whether the first feeder section
𝑚𝑖
is supplied from a primary (node i is MV bus bar of the substation)
𝑃𝑆𝑖 , 𝑄𝑆𝑖 Active and reactive power supplied by primary substation at node i
𝑦𝑖 Equal to 1 for existing nodes and to 𝑦𝑛𝑘 for new DG nodes
Binary decision variable indicating whether the existing circuit in
𝑐𝑒𝑖𝑗
branch ij is used or not
𝑓𝑖𝑗 , 𝑓𝑛𝑘𝑖 Fictitious power flow in existing branch ij and new ‘DG-branch’ nki
𝐶𝑂𝑆𝑇(𝑡) Hourly cost of loss-of-lives, losses and reliability
𝑐𝑜𝑠𝑡𝐿𝑜𝐿 (∙) Hourly cost of loss-of-lives
𝐿𝐶𝑖 (∙), 𝑐𝑜𝑠𝜑𝑖 Load curtailment and power factor at node i
𝐶𝑂𝑆𝑇1 , 𝐶𝑂𝑆𝑇2 Costs in planning period used for cable ranking
∆𝐶𝐼, ∆𝐶𝑀𝐿 CIs and CMLs associated with a cable section
𝑊𝑐 is the Joule loss per unit length for the conductor insulation (W/m);
is the dielectric loss per unit length for the conductor insulation
𝑊𝑑
(W/m);
Δ𝜃 is conductor temperature rise above ambient (K);
is the alternating current resistance per unit length of the conductor
𝑅𝑎𝑐,𝜃𝑐
at 𝜃𝑐 operating temperature (Ω/m);
𝐶 is the cable capacitance per unit length in (F/m);
𝑡𝑎𝑛 𝛿 is the insulation loss factor at power frequency
𝜆 Failure rate of component based on Weibull pdf
𝜇 Repair rate of component
𝜆𝑒 Failure rate of component based on exponential pdf
13
ABSTRACT
Title: Reliability Assessment of Ageing Distribution Cables for Replacement in

Smart Distribution Systems
Muhammad Buhari, The University of Manchester, June 2016
Doctor of Philosophy (Ph.D.)
Majority of electricity networks have growing number of ageing elements. Critical

network components, such as ageing underground cables, are very expensive to install
and disruptive to replace. On the other hand, global climate changes have made
connection of new low carbon technologies (LCT) into the grids increasingly
necessary. These factors are contributing to the increasing complexity of the planning
and management of power systems. Numerous techniques published on this subject
tend to ignore the impact of LCT integration and the anchoring ꞌSmartꞌ solutions on
ageing network assets, such as underground cables and transformers.
This thesis presents the development procedures of an ageing underground cable

reliability model (IEC-Arrhenius-Weibull model) and cable ranking models for
replacement based on system wide effects and thermal loss-of-life metrics. In
addition, a new concept of LCT integration and distribution network management was
proposed using two optimization models. The first optimizes connection of new wind
sources by minimizing the connection cost and the cost of cable thermal loss-of-lives
in the planning period. In the second stage, the network is optimally reconfigured in
such a way to minimize thermal-loss-of-life of ageing cable. Both optimization
models are formulated as mixed integer non-linear programming (MINLP) problems
applicable to radially operated medium voltage networks. To quantify the reliability
benefits of the proposed approach, Sequential Monte Carlo Simulation (SMCS)
procedure was formulated. Some of the main features of the SMCS procedure are the
IEC-Arrhenius-Weibull model for ageing cable, optimal network reconfiguration,
wind generation modelling using ARMA models and real time thermal ratings.
The final outputs are reliability metrics, cable ranking lists for replacement, savings
due to ‘non-spend’ cable thermal lives, etc. These studies have proven to be important
in formulating an effective strategy for extending the lives of network cables,
managing overall network reliability and planning cables replacement in power
distribution networks.
14
15
Declaration
No portion of the work referred to in the thesis has been submitted in support of an
application for another degree or qualification of this or any other university or other
institute of learning.
16
17
Copyright Statement
The author of this thesis (including any appendices and/or schedules to this thesis)
owns certain copyright or related rights in it (the “Copyright”) and s/he has given The
University of Manchester certain rights to use such Copyright, including for
administrative purposes.
Copies of this thesis, either in full or in extracts and whether in hard or electronic
copy, may be made only in accordance with the Copyright, Designs and Patents Act
1988 (as amended) and regulations issued under it or, where appropriate, in
accordance with licensing agreements which the University has from time to time.
This page must form part of any such copies made.
The ownership of certain Copyright, patents, designs, trademarks and other

intellectual property (the “Intellectual Property”) and any reproductions of copyright
works in the thesis, for example graphs and tables (“Reproductions”), which may be
described in this thesis, may not be owned by the author and may be owned by third
parties. Such Intellectual Property and Reproductions cannot and must not be made
available for use without the prior written permission of the owner(s) of the relevant
Intellectual Property and/or Reproductions.
Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or
reproductions described in it may take place is available in the University IP Policy
(see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual-
property.pdf), in any relevant Thesis restriction declarations deposited in the
University Library, The University Library’s regulations (see
http://www.manchester.ac.uk/library/aboutus/regulations) and in the University’s
policy on presentation of Theses.
18
19
Acknowledgments
In the name of Allah, the most beneficent, the most merciful. “…I desire nothing but
reform so far as I am able, and with none but Allah is the direction of my affair to a
right issue, upon Him do I rely, and to Him do I turn.” (Hŭd: 88). All praise and thank
are most due to Allah and may His peace and blessings be upon Muhammad his
servant and messenger.
I would like to heartily acknowledge my supervisors: Dr Victor Levi and Prof. Jovica
V. Milanović for their guidance, support, advice, and constructive comments over the
course of my program. I would like to extend my gratitude and thanks to Dr Victor
Levi for his invaluable help, empathy, and encouragement throughout this research.
His advices has inspired and motivated me to pursue not only technical research
excellence but also professional best practice.
I would also like to extend these acknowledgments to my sponsors – The Petroleum

Technology Development Fund (PTDF) Nigeria. Their financial support has made
this research possible.
My deepest appreciations go to my caring parents, brothers, and sisters. My parents

have been a constant and inestimable support to me throughout every stage of my life
and certainly no less during the PhD program. To my wife and daughters, big thank
you for your cooperation and unrelenting support. These appreciations extend to all
friends and colleagues too numerous to mention here, who prayed for me so faithfully
during the time of the PhD program. Thanks Mum, Dad, and to everyone.
20
21
CHAPTER ONE
INTRODUCTION
The chapter provides the overview, motivation, aim, and objectives of the research
project and the contributions of the PhD studies. Short descriptions of subsequent
chapters found later in this thesis are also included.
1 Power system reliability studies
Electrical power system is a critical infrastructure that creates and supplies safe and
usable electrical energy to end-users most of whom their livelihood depends on it. The
modern world economy, public health, safety, and security will suffer huge damage
without a reliable supply of electrical energy.
Power systems are perhaps the most complex large-scale engineered systems today
and expected to have the highest level of reliability because of the dependence of
virtually all human activities on them. The average interruption duration in most
power systems does not exceed 2 hours per year, yet their customers expect more
reliability and affordability anytime [1, 2].
For the power system planners, the increasing system size and ageing of the power
network assets, as well as stochastic nature of its numerous components makes it
challenging to manage interruptions and meet up with customer expectations.
Particularly, the uncertain component characteristic further compounded by asset
ageing makes power control and performance prediction a difficult task. Furthermore,
the affordability can be compromised if redundant equipment are incorporated into the
system to improve its reliability. Power system reliability is the incontrovertible
benchmark used by stakeholders to evaluate their satisfaction [1, 3-6]. Consequently,
utilities make major decisions affecting operations and maintenance, development and
investments based on planning standards as well as system reliability – quality of
service to end users.
The importance of system reliability studies to stakeholders has stimulated researchers

into developing numerous techniques and methods of evaluating power system
22
Chapter one: Introduction
___________________________________________________________________________________
reliability and for achieving optimal balance between reliability and cost of supply [1,
4, 6-11].
Power systems reliability studies quantify the capability of the system to satisfy the
customers energy demands in a given period of time [12]. The quantifications are
done in terms of Adequacy and Security assessments. Adequacy assessment is related
to the measured sufficiency of the generation, transmission, and distribution facilities
to satisfy expected load demands without violating the operational constraints of the
systems components within a given period (typically a year). The adequacy
assessments is conveniently carried on a hierarchical level basis as shown in Figure
1-1 [12]. On the other hand, the security studies involve a risk measurement of the
power systems’ ability to withstand disturbances (both dynamic and static overloads
and voltage assessments) arising within the system.
Reliability studies categorization
The reliability studies are classified using the three main subsystems; generation,
transmission and the distribution networks that are connected to the end users. The
reliability analyses of power systems are executed in hierarchical levels for purpose of
convenience and feasibility [12] and this process is illustrated in Figure 1-1 below.
Generation Hierarchical level I

facilities (HL I)
Transmission Hierarchical level II

facilities (HL II)
Distribution
Hierarchical level III (HL III)
facilities
Figure 1-1: Hierarchical levels of reliability studies in power systems
Analysis at HL I is concerned with adequacy of generator(s) output to meet

transmission connected load demands. HL I studies assume that both the transmission
and distribution subsystems are fully reliable and are thus ignored. HL II studies cover
23
___________________________________________________________________________________
both the generation and transmission subsystems and are more complicated than HL I
studies because they involve modelling of transmission networks. HL II studies
involve assessing the ability to meet connected load demands while satisfying
operational constraints (voltage and current flow limits etc.) on the transmission
network.
Lastly, the HL III studies deal with the ability of the power system to meet individual
customer supply demands. HL III studies consider the whole power system including
distribution networks in the reliability analysis. This comprehensive approach makes
the modelling and execution complex and time consuming. Due to the complexity of
performing HL III studies, it is a common practice to use the outputs from HL II as
inputs to the HL III studies in order to reduce modelling requirements and speed up
computations during the analysis [1, 2, 4, 11, 13]. Different reliability indices are used
for the three hierarchical levels in examining the system performances. In this thesis,
HLIII studies involve the examination of the distribution system facilities and utilises
the results of the HLII evaluation as inputs to these studies.
The next section provides more details on the HL III studies, which is central to this
thesis.
1.1 Distribution network reliability
The distribution system is the last part of the electric power supply network. It
manages the transportation, voltage transformation, and regulation of electrical energy
supply (voltage and currents) to customer levels. It is also responsible for majority of
faults and abnormal events occurring on the network, and the network owners must
provide appropriate actions necessary for safety of people and equipment and for
service restoration following any service interruption. Due to the large geographical
span between supply points and the number of (inter)connecting points, numerous
different types of components are connected on typical distribution systems. Main
components include sub-transmission lines, bus bars, several levels of distribution
transformers, overhead lines, underground cables, and protection devices such as
circuit breakers, relays, fuses, feeders etc.
24
___________________________________________________________________________________
The aim of distribution systems planners is to (as economically possible) provide end
users with electricity of adequate reliability – quality as set by the regulators, planning
standards and distribution code. However, the distribution networks have been the
weakest link (in terms of frequency and duration of component failures) in the power
systems supply chain. About 70% of system down time is recorded at this level of the
power supply chain and the interconnecting branches (especially cables) are proven to
be a critical system element [1, 5].
The distribution system reliability studies are carried out to assess the ability of the
system to meet the electricity demands of connected customers when sudden
equipment failures or other disruptive factors reduce the amount of available
electricity [4]. HL III studies are therefore necessary and are carried out using the
reliability analysis in conjunction with ‘adequate’ reliability indices. For this reason, it
is required to model the overall behaviour/characteristics of the distribution system in
as much detail as possible. The following studies are usually done in order to design a
distribution network [14]:
 Power flow study;

 Fault level study; and
 Reliability study.
The power flow and fault level studies are essential for a successful functioning of
any power system. At the distribution system level, a power flow analysis is needed to
compute the steady state voltages and currents in system components in order to
ensure that a proposed design will meet the planning criteria, such as equipment
loading limits, voltage drops, emergency supply etc. It is also used to investigate the
impact of future connections, system expansions, and upgrades. A fault level study is
necessary to determine types and rating of switchgear, typically circuit breakers, and
motorised switches, as well as sizing of protection relays.
However, successful functioning of the power system is only one of two important
parameters to make a design viable. The second essential parameter is the level of
assurance that the system will perform its functions correctly for an acceptable
duration. Thus, a reliability evaluation study is required to design an acceptable
system that can satisfy stakeholders’ requirements. The reliability studies involve
25
___________________________________________________________________________________
assessment/analysis along with the computation of the systems’ performance in terms

of ‘adequate’ reliability indices in order to design systems that meet regulatory
standards [4, 5, 15].
The basic indices used include failure rate λ, failure duration 𝑟, and unavailability 𝑈.
Whilst for the aggregate system (or any group of customers or feeders), the most
common reliability indices applied in distribution system analyses are the number of
interruptions a group of 100 customers are likely to experience (or customer-
interruptions (CIs)) and the average duration of interruptions per 100 customers (or
customer-minutes-lost (CMLs)).
Other indices used include the expected energy not served (EENS), system average
interruption indices (SAIDI) and system average interruption frequency (SAIFI) etc.
[3, 16, 17] etc. These indices represent average system performance and are based on
the perceived two states of the system/components (i.e. operational or failed state);
they were found to be suitable for developing simple distribution system reliability
models.
1.2 Reliability evaluation methods
This thesis attempts to analyse the fundamental issue of ageing failure events in
electric distribution systems based on the sequential Monte Carlo Simulation (MCS)
method (see Figure 1-2). In order to solve this problem, it is necessary to develop
models based on mathematical expressions. In general, a model could be
deterministic, analytical or simulation based. Power system planners use two main
approaches (analytical and or simulation methods) to evaluate the reliability of their
systems in terms of adequacy of energy supply. Using a deterministic approach, a
phenomenon can be modelled with an exact mathematical expression; an example of a
deterministic method is the allocation of percentage reserve in generation. Another
example is planning network capacity based on constructing a minimum number of
circuits to a load group i.e. using the (n-1) or (n-2) criterion, where n is the number of
circuits, the minimum number is based on the maximum demand of the load group.
The analytical approaches represent the power system using mathematical models that
are typically based on Markov models. The solution of these mathematical models is
used to evaluate the system reliability.
26
___________________________________________________________________________________
Alternatively, planners can utilise simulation techniques such as Monte Carlo

simulations to represent the actual process along with its associated uncertainties.
Thus, the random behaviour of components failures can be considered in the model
expression, for example the power discrepancy from a sinusoidal function. The
relationship between the mentioned approaches is depicted in Figure 1-2. In this
thesis, probability theory is applied to evaluate random character of several
phenomena studied.
Power System
Reliability Studies
Adequacy Security
HL I HL II HL III
Simulation approach Deterministic Analytical techniques
Monte Carlo simulation Power flow Machine Learning

(MCS) techniques analysis algorithms
Independent MCS Sequential MCS
Figure 1-2: Reliability assessment approaches (thesis approaches in bold)
In early days, a few analytical techniques and deterministic judgments were used to
assess past or future performance of power supply adequacy at the customer load
points due to their relative ease of implementation [3].
In order to more confidently assess the performance of power systems, past research
has demonstrated a wide variety of analytical techniques, which include cut-set, fault-
tree, Markov chain models etc. [2, 7-10]. Most of these techniques are convenient for
smaller systems with manageable numbers of states under investigation. The
analytical models based on the Markov concepts along with network modelling
concepts such as the minimal cut-set techniques and they provide satisfactorily
accurate results for a wide range of smaller practical distribution systems [2-5]. The
analytical assessment approaches ignores the stochastic nature of components failures
and the uncertainty in operations and control parameters. The analytical methods also
27
___________________________________________________________________________________
become complicated to implement for larger systems with a large number of

system/component states.
For larger systems, simulating the power systems’ structural and operational
characteristics is a realistic way to mimic its actual behaviour and random failure
pattern especially when examining one or more of its many possible states, thus, the
Monte Carlo simulation (MCS) techniques prove to be a more convenient and better
option can be utilised [2, 4, 13, 18, 19]. Some comparable features of the simulation
approaches are listed in Table 1-1 below.
Table 1-1: Comparison of analytical and simulation approaches
Analytical technique Simulation technique

The developed models are often Every event incorporated into a model
simplified. can be simulated.
Solutions ready in short time. Involves extensive computing times.
Output is in the form of average values. Outputs can be re-arranged to generate
the probability distribution functions
associated with the reliability indices.
1.3 Monte Carlo simulation (MCS)
The MCS techniques can be classified into two essential groups: the non-sequential
(independent) MCS and the sequential MCS techniques. However, a hybrid model of
these two MCS techniques can also be developed. Furthermore, the analytical and
non-sequential MCS methods calculate the unavailability or probabilities of
component failures, while the sequential MCS methods considers the alternating
failure and repair cycles along with the time factor, thus providing additional indices
such as frequencies and durations [2]. The main advantage of the sequential MCS
method is modelling of chronological events, which is not possible with independent
MCS method and or analytical methods. Component failures can be either reparable
or non-reparable and these failure modes are considered in any of the three main
approaches discussed above.
At present, the simulation approaches enjoy wide acceptance and this thesis focuses
on simulation methods used in distribution networks to assess the relative merits of
alternative development schemes (construction and operations) including the option of
not doing any intervention at all. Most recently, due to availability of advanced
28
___________________________________________________________________________________
facilities used in computation, numerous applications of MCS methods to investigate

the reliability of different large power system segments from generation to customer
points were developed [4, 5, 11, 15, 18, 20].
More importantly, MCS methods can be used to examine most contingencies and
complex operating characteristics inherent in the power system [18, 21]. The two
main categories of MCS techniques: the non-sequential MCS and the sequential MCS
approaches are briefly discussed below.
1.3.1 Non-sequential MCS technique
Non-sequential MCS (independent MCS) is also known as state-based sampling

approach. Its analysis considers the discretized load demand curves, components
statuses, and snapshots of their combinations. The essence of the method is analysis
of mutually independent system states obtained in a random way. The main difference
between this technique and the sequential MCS techniques is that; the non-sequential
MCS ignores the chronology inherent in the operations of real-life power systems;
hence, it is the simpler of the two main MCS methods. For these reasons, the
independent MCS method is not used in this thesis.
1.3.2 Sequential Monte Carlo simulation (SMCS) technique
The SMCS technique is based on sampling the probability distributions of the

component state durations (also referred to as the state duration sampling technique).
It steps through time; it recognizes that the status of a system component is dependent
of its status in adjacent hours. The SMCS technique can be applied to consider all
contingencies and operating characteristics inherent in the system in a more realistic
fashion. The SMCS techniques can be used to accurately determine the systems’
interruption frequency and duration indices by simulating the actual process (with its
time-chronology element) and random behaviours of the system in a continuous time
[18]. The SMCS methods allow the consideration of the components physical
characteristics during the assessment.
The main requirement of the SMCS is to generate realistic uptimes and downtimes
histories of the relevant network components. These histories will depend on the
29
___________________________________________________________________________________
system failure/repair modes and the reliability parameters of the components. A two-
state model can generally represent distribution system elements. The up state
(Uptimes) indicates that the element is in service, while the down state (Downtimes)
implies that the element is inoperable due to a failure or a scheduled activity such as
routine maintenance etc.
The time during which the component remains in the up state is called time to failure
(TTF). The time during which the element is in the down state is called restoration
time and can be either the time to repair, or the time to replace (TTR) which is the
case for permanent ageing failures (i.e. end-of-life failure) of the component. The
transition process from the up state to the down state defines the component’s failure
process. Transition from the up state to a down state can be caused by the failure of
the considered element or by the removal of the element for maintenance.
The parameters TTF and TTR are random variables that are described by different
probability distribution functions. The Exponential, Gamma, Normal, Lognormal and
Poisson probability distribution functions are most often used to describe the TTF and
TTR variables [2, 4, 5, 20]. Their values are randomly generated within the SMCS
procedure from the uniform distribution applied to the cumulative probability
functions [4, 5, 11, 18]. The system indices are used to evaluate the system
performance and are thus calculated from the basic load point indices. They are
weighted averages of the individual load point values. Distributions of the system
indices can be obtained from the periodic load point indices if the SMCS is repeated
many times for the studied year.
In the SMCS, the number of iterations (𝑁) required to find an accurate solution can
be difficult to determine. This is because too low 𝑁 produces unacceptable error,
while too large 𝑁 results in too long simulation process and period. However, errors
will certainly reduce as the number of simulations increases. Hence, a convergence
criterion can be specified in such a way to determine the optimum number of
simulations so that the results converge within an acceptable error range. The
coefficient of variation (CoV) can be used as a convergence criterion as suggested in
[2, 18, 22]. It is usually applied to the reliability indicator whose convergence is
hardest to achieve.
30
___________________________________________________________________________________
However, the SMCS technique is computationally intensive, requires more

computation time and memory space than the non-sequential MCS approach. This
SMCS attributes occur due to the need to generate and store information on the
chronological state transition processes of all system components in a long time span
(e.g. 8760h in a year). Fortunately, the availability of high-speed computers makes the
SMCS technique a viable option. Other benefits are higher confidence in the
computed results and the whole range of reliability indices calculated within repeated
yearly studies. The method is also very convenient to study large and complex power
systems.
1.3.3 Component (Asset) modelling
One of the primary goals of the system planners is to develop asset management
strategies covering the full life span of system components (assets). The preferred
strategy is one which can result in targeted performance with optimal financial returns
(or, minimized costs) based on the results from the system reliability evaluations [23,
24]. A tree diagram with three main branches depicting the inter-relationship of the
asset assessment strategy for managing power networks is shown in Figure 1-3 below
[23]. The asset management provides a robust, procedural basis to manage risks in
terms of operation and business success [25].
Reliability
Performance Security
Capacity
Customer
Network Equipment Safety
Risks
Assets
Environment
Life Cycle costing

Financial
returns Operational efficiency
Investment & Maintenance planning
Figure 1-3: Network Assets management planning map [23]
From Figure 1-3, it can be deduced that an effective plan will evolve from a holistic
assessment of the assets life cycle that satisfies the stakeholders, regulatory and
31
___________________________________________________________________________________
customer expectations. Ultimately, the main goal of maximising the assets returns
involves a technical planning centred on components performance and operational
risks. The network reliability assessment requires the development of mathematical
models of the network components/assets, as well as of the system impact on
individual assets.
1.3.4 Causes of distribution systems interruption
An equipment lifetime depends on four major factors: design, how the unit is used –
level of exposure to damage from external elements, loading pattern and through-
faults [4]. In distribution systems, customer interruptions occur due to events such as
component/equipment failures, contact with animals and trees, impact of severe
weather and human factors (i.e. car accidents, dig-ins, human errors, scheduled
interruptions, vandalism etc.). These events can be considered in the reliability
modelling by a proper understanding of the nature of customer interruptions and
addressing the root causes. This understanding is crucial for developing an accurate
system reliability model [4, 26]. In general, components and system reliability are
functions of several factors, such as [4-6, 20]:
 Component failure rates (ageing and non-ageing - reparable failures)

 Loads & loading profiles (e.g. electric vehicles, demand response
etc.)
 Design parameters of major components (e.g. cable insulation
material, generator capacity etc.)
 Location and sizes of major components
 Repair and restoration (i.e. switching) durations
 Weather effects on components (e.g. temperature, snow etc.)
The next section describes the most frequently met operating states of the distribution
system/components due to the above factors and how they can be modelled for the
assessment of the system/component performance.
32
___________________________________________________________________________________
1.3.5 Components states modelling
The determination of the system/component operating states signals the beginning of

the simulation (reliability assessment) process and it is an essential input for the
assessment of the overall system behaviour in any reliability modelling technique.
The states can be classified as [2]:
 Normal State,
 De-rated state (e.g. big generators),
 Contingency (i.e. down) state, and
 Scheduled event state such a routine maintenance activity.
A normal operating state of the system occurs when all components are operating as
desired, loading levels are within design limits and the system is configured with its
nominal settings. In this state all system switches are in their ‘default’ positions (i.e.
normally closed-NC or normally open-NO) and protection devices have not initiated
tripping of some components. This is the preferred operating state because all
constraints are met and all customer load demands are satisfied. On the other hand, a
contingency event is any event that will disrupt the normal operating state, whether it
is an unplanned event or a planned event [2-4, 13, 20, 27].
A contingency event is a chance event such as a fault condition that disrupts the
system’s normal operating state, whilst a scheduled event is a planned activity such as
maintenance which results in the interruption of normal system services [2-4, 13, 20,
27]. Therefore, state based reliability assessment methods (i.e. analytical methods,
Markov modelling and MCS), which are sometimes called predictive simulation
techniques, are required for a comprehensive analysis of the system reliability.
1.4 Research motivation
It is important to recognise that ageing is a critical issue inherent in assets based

businesses such as power systems. The main motivation for the research is on
improving the reliability of distribution systems considering the advanced ageing of
most assets, especially the underground cables (UGC). The advanced ageing of the
UGC is due to the current age of the assets and the loading profiles. Electric energy
33
___________________________________________________________________________________
supply in most urban centres is done via UGCs; they cover extensive areas to reach all
customers; they are extremely expensive and disruptive to repair hence they are
critical. In addition, following any failures, the time required to fix them is usually
extensive. Therefore, it is imperative to ensure reliable cable performance for the
provision of the required standard quality-of-supply. Furthermore, utilities are
challenged by limitations imposed on their capital expenditure budget by regulators. It
is therefore necessary to consider new ways of managing asset replacements,
expansion planning (investment deferment), and efficiency in operations.
The second motivation is the fast-evolving ˈsmart gridsˈ manifesting through the
increasing integration of measurement, communication, IT, and software solutions
(so-called ꞌsmartꞌ solutions), as well as low carbon technologies into the existing
power grids. The smart grid (SG) concept is expected to deliver significant,
measurable, and sustainable benefits to the utilities, consumers, the overall economy
and the environment. SG concept potentially alters the existing industry’s approach to
managing network assets for economical delivery of the much desired optimal
functionality [15, 28]. In spite of the envisaged potentials, the SG technologies are
still in their infancy and successful operations of both existing and future grids will
require some of the following strategies:
 Improving the utility’s assets utilisation levels, reliability and the

network security;
 Optimizing the integration of low carbon technologies into existing
electrical networks in a reliable and economical way.
In this thesis, the impacts of an ageing asset on the successful implementation of a

future SG is investigated with the aim of improving the reliability of the system
considering increased power flows due to advanced loading profiles and additional
LCT connections on the networks. The results of these investigations are used to
determine network assets (cable) ranking for replacement planning. The results are in
the form of reliability indices obtained using the SMCS techniques that can provide
more accurate prognosis of future systems reliability levels since both ‘smart’
solutions and new LCTs can be conveniently modelled.
34
___________________________________________________________________________________
1.5 Research question
This thesis presents analysis of existing distribution cable replacement planning

methods and suggests a new approach for future smart grids. The research work aims
to answer the following questions:
 How an ageing network infrastructure impacts the distribution

network reliability?
 How to prioritize critical network components replacement based on
proposed criteria that include cost of reliability?
 How to integrate and manage new LCTs into distribution network
containing ageing assets, with the aim to extend asset lives?
This thesis focuses on developing replacement strategies for ageing underground

cables in distribution networks with newly connected LCTs and implemented ‘smart’
solutions. The overall approach is probabilistic and adequacy studies at the
hierarchical level III are applied (Fig. 1-2). The developed studies can provide a
platform to assess the impacts of inevitable but controllable system events such as
components end-of-life ageing failures. These ageing related failures can be
influenced by network dynamic loading profiles and components inherent design
properties.
In the UK, developers decide the proposed locations of the LCTs based on the
distribution codes [29]. In the studies carried out in this thesis, it is assumed that the
arbitrary chosen LCT connection locations coincide with the choice of the developers,
which is a possibility not investigated previously. Alternatively, incentives could be
provided to developers to choose from the determined network points in order to
boost network reliability.
1.6 Research aim and objectives
The aim of this thesis is to develop a methodology for predictive assessment of the
adequacy of ageing assets (underground cabling system) in smart distribution systems
for replacement planning decisions, whilst improving reliability.
35
___________________________________________________________________________________
The proposed study involves evaluating the reliability of distribution systems with the
proposed ageing cable reliability model (IEC-Arrhenius-Weibull model) in order to
generate cables ranking lists for effective replacement planning. The proposed
reliability model considers the components age, dynamic loading, thermal stresses,
ambient conditions, and asset physical characteristics.
The overall (reliability) model is further expanded to consider optimal management of

wind-DG integration. The developed optimal wind-DG connection strategy is with the
aim of minimizing the connection cost and the cost of cable thermal loss-of-lives in
the planning period. In addition, a new control strategy, optimal dynamic network
reconfiguration for preserving ageing cable life, is also incorporated into the
simulation methodology. These studies are necessary to develop an effective strategy
for extending the lives of network cables and planning cables replacement in power
distribution networks.
The study objectives include:
1. Develop a cable reliability model that considers ageing failures, cable life
evolution and cable thermal modelling.
2. Develop improved simulation procedures that will consider the current age of
components, dynamic loading, thermal stress, and installation characteristics in
the network reliability evaluation.
3. Develop the new methods for ranking of the underground cables for replacement
planning.
4. Develop research grade program of an improved methodology that will investigate
the optimal integration of low carbon technologies into distribution networks to
defer assets investments. Two optimization models to be developed:
i. First, optimize connection of new wind sources by minimizing the
connection cost and the cost of cable ageing in the planning period.
ii. In the next stage, optimally reconfigure the distribution network in such a
way to minimize thermal ageing of cable by reducing the line current
flowing through critical cables.
36
___________________________________________________________________________________
1.7 Contributions
The main contribution of this research work are summarised below. Note that
references given in parenthesis indicate that the work has either been published in a
conference or presented at a peer reviewed international Journal. The publication list
can be found in the attached appendix (List of Publications).
1. The development of a reliability model for ageing underground cables to

be applied in replacement planning of distribution systems. The proposed
technique considers for the first time combined cable thermal model,
Arrhenius ageing model and Weibull stochastic process (i.e. ‘IEC-Arrhenius-
Weibull’ model) for modelling up-times of ageing distribution cables. The
parameters of the Arrhenius model are estimated using a novel technique that
considers the cables’ ages and their historic loading profiles (JP 1, CP1 &
CP2).
2. The integration of unconventional end-of-life failure models of
underground cables into system reliability assessment. This is the first
study that incorporates the unconventional end-of-life model (i.e. IEC-
Arrhenius-Weibull cable failure model) into system reliability assessment. The
failure model considers the thermal stress due to systematic loading conditions
and is characterised by the Arrhenius-Weibull probability distribution. This
study improved the system reliability assessment as it evaluates the impacts of
considering the cables age, historic loading and current loading levels in the
failure models and how this has enhanced the understanding of the influence
of operational factors on system reliability (JP 1 & JP2).
3. The development of an industry-financial ranking index for cable
replacement planning. Within the developed SMCS procedure, reliability
indices such as the customer interruptions (CIs), customer minutes lost
(CMLs) and the numbers of failures are associated with each cable section in
order to investigate each section's contribution towards unreliability. They are
then combined with appropriate unit costs to form a new cable ranking list.
The proposed ranking method is finally compared to the cable ranking using
thermal loss-of-life (JP 1 & JP2).
37
___________________________________________________________________________________
4. The development of a distribution systems planning model, which

involves optimization of both construction and operational characteristics
of the system. Two new optimization models are developed: the first
optimizes connection of new wind sources by minimizing the cost of cable
thermal loss-of-lives and the connection cost in the planning period. Once the
new wind energy sources are connected, a new smart concept is deployed to
optimally reconfigure the network topology in such a way that the thermal-
loss-of-life of ageing cable is minimized. Both optimization models are set as
MINLP problems applicable to radially operated medium voltage networks.
New reliability model of ageing cable, optimal network reconfiguration, wind
generation modelling using ARMA models and real time thermal ratings are
some of the features incorporated into the developed sequential Monte Carlo
simulation procedure. The final outputs are reliability metrics, cable ranking
lists for replacement, savings due to ‘non-spend’ cable thermal lives, etc. (JP 2
& CP 3)
1.8 Thesis structure
In chapter 1, the thesis report begins with an overview of distribution systems

reliability assessments for planning purposes. It then provides a review of the tools
required for implementing system reliability studies based on the SMCS
methodology. The chapter is wrapped up with a list of research questions, aim and
objectives of the studies, and a summary of key contributions from the research work
conducted.
Chapter 2 presents the literature background on the subject of thermal modelling of

cables more specifically based on the IEC cable thermal rating standards. The chapter
details the technical aspects of cable modelling needed for the development of a
proper cable model and how ageing can be analysed using the Arrhenius thermal loss-
of-life model for cables. In this chapter, a novel way of determining the Arrhenius
model parameters A and B is proposed and described because of the costs associated
with the required studies.
Chapter 3 begins with a literature review of network reliability studies that consider
components ageing aspects in the perspective of system reliability studies. Weibull
38
___________________________________________________________________________________
probability function is a popular model used to represent the ageing characteristics of

network components in many works. Thus, details of the development process for the
proposed IEC-Arrhenius-Weibull reliability model for cable ageing are provided in
this chapter.
Chapter 4 presents the step-by-step description of the overall research methodology.

This includes the computer simulation flowcharts and details of how the various
proposed models together with other components of the SMCS methodology are
combined to formulate a robust SMCS model for proper network assessment studies.
The sections present descriptions of the case studies of the research work along with
the details of the test systems and other inputs used in generating the results. The
section concludes with preliminary results from assessments of the described test
systems.
Chapter 5 presents an overview of distribution systems investment planning processes

with emphasis on asset replacement planning. The chapter highlights issues around
system performance indices and regulatory constraints. The chapter also provides the
analysis of the reliability metrics and a comparison of the proposed methods against
conventional methods. Details of the proposed industry-financial cable ranking index
for replacement planning are provided. The final sections conclude with results of
cable ranking studies carried out on the test systems introduced in the previous
chapter.
Chapter 6 describes the concept of future smart grids. The smart grid concept involves
the incorporation of low carbon technologies and ˈsmartˈ control schemes to improve
the overall performance of existing grids. Literature reviews on the assessment of
smart grids performance is provided in this chapter. The chapter then provides details
on the development of two approaches to achieving a smart distribution system. The
approaches consider development of a distribution systems planning model, which
involves optimization of both construction and operational characteristics of the
system along with reliability assessment studies. The details of the proposed SMCS
model which consists of optimal network reconfiguration model; and wind generation
modelling using ARMA form part of this chapter.
39
___________________________________________________________________________________
Conclusions with suggestions for future research work can be found in Chapter 7.
Several remarks were made at the end of each subsection in order to highlight key
observations.
40
41
CHAPTER TWO
THERMAL MODELLING OF AGEING CABLES
This chapter presents the context, background and importance of the thermal
modelling of ageing medium voltage (MV) cables for reliability evaluation in
planning of distribution systems. The main driver behind this chapter is growing
interest amongst power system planners to understand the impact of increased power
flows on thermal ageing of underground cables.
2 Modelling of ageing cables
Underground cables (UGCs) are critical components in electricity distribution because

their failure can produce significant consequences on the end-users due to the high
repair/replacement cost and time. The ageing process of cables differs in each
system/section due to the different utilisation levels and stress conditions. Ageing
occurs in UGCs due to several factors such as thermal stresses (higher operating
temperatures), voltage stresses (surface or internal discharges), mechanical stresses,
chemical contaminants, environmental conditions (heavy humidity, moisture ingress)
etc. In UGCs, temperature (thermal stresses) is the most frequent cause of ageing due
to its continuous duty cycle [6, 19, 30-34].
Power system components can experience one of the two basic failure modes;
repairable and non-reparable failures. Most researchers investigate components
repairable failures in their reliability analysis and tend to ignore critical ageing
process that aggravates cable condition with time. Ageing failures (or end-of-life
failures) are often classified as non-reparable failures [6, 19, 24, 35, 36]. Ageing
failures frequently occur on those components that are experiencing more than
nominal loading for an extended period of time or nearing their end of design life.
Once an end-of-life failure occurs, there will be no repair process to bring it back to
its normal state and a replacement must be arranged. In addition, the occurrence of an
ageing failure on one component could lead to increased ageing related failures on
adjacent in-service network components due to the consequent increased stress they
are likely to suffer as the initial failure result. Therefore, modelling of network ageing
42
Chapter two: Thermal modelling of ageing cables
___________________________________________________________________________________
components is an important factor in reliability analysis and planning of future energy

networks [36-38].
In this thesis, the focus is put on developing a decision making tool for replacement
planning of UGCs that can experience end-of-life failures. To this end, the cable
thermal models are an integral part of the developed sequential Monte Carlo
simulation (SMCS) procedure. The next sections provide a short description of cable
thermal ratings and thermal standards.
2.1 Common cables for medium voltage distribution systems
In the UK, two types of MV cables are usually deployed in the distribution systems
[39]. They are medium voltage paper insulation fluid filled lead covered (PILC) and
extruded dielectric cable in the form of crosslinked polyethylene (XLPE), tree
retardant XLPE (TRXLPE) and the ethylene polypropylene rubber insulation (EPR)
[31, 39-44]. These cable types are installed by direct burial in rural areas and by
placing them in ducts in urban areas. There is an increase in the installation of three
core XLPE cables, single core TRXLPE and EPR cables. Recommended cable
installation practices for the UK and some countries are detailed in [42]. This thesis
focuses on the XLPE cable analysis due to its wide deployment in the distribution
networks. In this thesis, the XLPE and PILC cable types are considered in the
reliability studies carried out. The next section provides a short overview of the cable
thermal modelling studies.
2.2 Brief historical overview of studies on conductor thermal ratings
Pioneering research works in the field of cable conductor thermal ratings can be
traced back to the year 1800 when Michael Faraday conducted theoretical and
experimental studies on conductor thermal ratings [33, 45]. In order to appreciate
these remarkable efforts of the past, some of the earliest works dedicated towards
understanding the behaviour of the conductor thermal rating are given below:
1. The first simple models for calculating the ampacity of underground cables
were developed by Forbes in 1884 and Kennelly in 1889 [46].
43
___________________________________________________________________________________
2. Early publications [33, 46] introduced simple formulas derived from

measurements and experimentations for the estimation of the steady state
current carrying capacity and operating temperatures of electrical fittings.
3. The pioneering work on the derivation of the steady state thermal resistance
formula between a conductor core and a surrounding cable sheath was
authored by Mie in 1905 [47].
4. Booth (1911) derived the formula co-relating conductor cross-sectional area
with permissible steady state current density [48].
5. Atkinson et al. in 1913 pioneered the consideration of transient temperature
rise of conductors [49].
6. In the UK, the history of calculations of current ratings has its early traces in
the 1924 ERA report [33].
7. The publication of the first set of current carrying capacity tables for PILC
underground cables in ducts or air was issued in 1932 [33].
8. Significant number of publications surfaced after the World War II including
the paper by Neher-Mcgrath, which effectively summarised previous works.
The paper generated huge interests and stimulated further research works on
the subject [33, 50].
9. The authors in [50] described a model based on the principles of energy
balance in conductor operations and the analogy between electric current flow
and heat flow. Although the physical principles have been established in
previous works, this paper is viewed as the basis for modern ampacity models
[33].
10. In 1986 Anders et al [51] introduced a computer program based on the
methods in [50] for computing cable ratings more efficiently.
11. As computer technology advanced, the use of numerical models and solvers
became more widely accepted which prompted the publication of a standard
procedure for numerical modelling [34].
The above key historical milestones show the long-standing interest in the
development and adoption of conductor thermal ratings as a standard for planning
electrical energy networks and design of cables. The next section provides a review of
the most important existing standards for the calculation of cable thermal ratings.
44
___________________________________________________________________________________
2.3 Review of cable thermal rating standards
Cable thermal ratings are crucial quantities when considering distribution network
security and reliability. Different ratings can be used during operations for system
normal and contingency conditions. It is important to adhere to the ‘accurate’ ratings
both in emergency and in normal operating conditions to avoid permanent degradation
and enhance cable service life. The UK distribution companies have their own codes
of practices, which specify applicability of different cable ratings under different
conditions. Besides, they use purposely developed software (CRATER software) to
find accurate ratings using the standard described in the next paragraph [52].
Research in the field of conductor thermal ratings has been extensively presented in
the literature and at various stages culminated in the formulation of several national
and international standards. An earlier publication by Neher and Mcgrath in 1957
triggered the greatest interest in this domain. The calculation procedures in each of the
different standards share the same basic principle, which is heat balance and energy
conservation. Consequently, it is possible to compute the different cable current
ratings using the cable models provided in the international and national standards,
such as the:
 International Electrotechnical commission (IEC) [34, 44, 53]

 Institute of Electrical and Electronics engineering (IEEE) [54, 55],
 International Council on Large Electric Systems (CIGRE) [6]
The models provided in the above listed standards are based on law of energy
conservation, which stipulates that there must be a balance between all energy rates at
any instant. The developed heat balance models account for the heat balance between
heats generated from the cable (Joule losses) and the heat dissipated into the
surrounding environment (through insulation and soil) when the cable is energized.
The cable properties, the soil thermal properties and installation conditions are all
considered by these models to determine the cable ampacity. This is because these
factors have significant influence on the heat dissipation and are critical when
determining the permissible thermal ratings. On the other hand, continuous high
current loading can generate excessive conductor temperatures resulting into
irreversible damage to the cable insulation. The current ratings can be determined
45
___________________________________________________________________________________
more precisely by monitoring the network load, using the temperature measurement
sensors on the cable sheath, doing hot spot investigations and by investigating the
surrounding soil thermal properties etc., [44, 56]. Real-time current rating calculation
can be realized by considering the actual temperatures of the cable sheath along with
the dynamic thermal behaviour of the surrounding environment of the cable. It is
evident that the thermal properties of the cables’ surrounding medium has a
significant impact on the cable ampacity [57].
The IEC model is more comprehensive when compared with the other cable thermal
models and is widely accepted in the literature [37, 57, 58]. The IEC 60287 standard
for calculating cable ampacity consists of two parts, namely:
 Calculation of the current rating (at 100% load factor) [44]

 Calculation of the cyclic and emergency current ratings of cables [53]
The current rating computation presented below is applicable to the conditions of

steady-state operation of the cables at all ac voltages, and dc voltages up to 5 kV.
Also, the IEC 60287 standard methods rely on knowledge of the soil temperature at
which drying out of the soil takes place rapidly. The following input parameters are
also required during computation [23, 37, 41, 57-59]:
 Cable properties (cross sectional area, type of sheath bonding, etc.),

 Installation and laying conditions such as installed in air, directly
buried, in duct banks, in cable trenches, in tunnels, backfills, etc.
 Phase arrangement and spacing dimensions (trefoil, flat, etc.),
 Soil thermal resistivity.
However, the IEC 60287 model assumes uniform thermal properties during network
operation along the cable length and in each of the materials surrounding the cable’s
conductor, thus ignoring the issue of hot spots. The deployment of numerical cable
modelling techniques can be used for considering both constructional and hotspot
features more effectively [34]. Notwithstanding the increasing acceptance of
numerical modelling techniques, the IEC Standards 60287 and 60853 are still widely
used by researchers and engineers for the analysis of cable systems and verification of
46
___________________________________________________________________________________
numerical models [31, 33, 37, 57]. The next section provides more details on the
calculation of the cable thermal rating based on the joint BS and IEC standard.
2.3.1 IEC cable thermal model
Methods for calculating conductor thermal ratings for buried cable are divided into
two parts – ‘transient’ and ‘steady state’ ratings. Both parts are covered in IEC 60287
standards and both are based on the principle of heat transfer equilibrium. The thermal
rating model is used to determine the maximum current loading for a specified
temperature or to determine the temperature of the cable conductor for a given current
loading. Although the IEC 60287 method is suitable for simple geometries, it still
enjoys the confidence of system planners, which can be attributed to the following
advantages [31, 33]:
 It is based on the basic principles of heat transfer and laws of energy

conservation that was established around year 1800.
 It is simple to execute and does not need high performance computer
software or hardware.
 It is still widely accepted in the literature as the reference method in
developing the advanced models.
The energy balance equation for underground cables is represented by the expression
shown in (2-1).
𝑊𝑜𝑢𝑡 + ∆𝑊𝑠𝑡 = 𝑊𝑒𝑛𝑡 + 𝑊𝑖𝑛𝑡 (2-1)
where 𝑊𝑒𝑛𝑡 is the rate of energy entering the cable; 𝑊𝑖𝑛𝑡 is the rate of internally
generated energy (heat) in the cable by Joule or dielectric losses; 𝑊𝑜𝑢𝑡 is the rate at
which heat energy is dissipated by conduction, radiation; and ∆𝑊𝑠𝑡 is the rate of
change of energy stored within the cables material. Note that when computing 𝑊𝑜𝑢𝑡 ,
the cable system includes the surrounding soil for buried underground cables.
Heat generated in a conductor by electric current flows is dissipated into the

surrounding medium (such as air, soil etc.) via two methods: conduction and
radiation. Therefore, the surrounding medium properties affect the rate of heat
dissipation. The most important properties affecting the heat dissipation include the
47
___________________________________________________________________________________
insulation material type, soil type (back fill), soil temperature, and the ambient
temperature.
Conduction
The Fourier’s law is usually applied to define the heat flux emitted by conduction
from a material medium using the relation (2-2).
𝑑𝜃 1
𝑞=− ∙ (2-2)
𝑑𝑥 𝜌
𝑑𝜃
where 𝑞 is the heat flux (W/m2); 𝜌 is the thermal resistivity (K. m/W); and − 𝑑𝑥 is the
negative local temperature gradient (K/m).
Radiation
Heat flux characteristics due to radiation obey Stefan Boltzmann’s law, which is given
by relation (2-3):
𝑞 = 𝜀𝜎𝐵 𝜃𝑠4 (2-3)
where 𝜀 is the emissivity of a conductor; 𝜎𝐵 = 5.67 ∙ 10−8W/m2∙ 𝐾 4 is Boltzmann’s

constant; and 𝜃𝑠 is the absolute temperature on the cables’ surface, K. For the
installation condition where there is a non-negative ambient temperature, the heat flux
is calculated using the equation given in (2-4).
4
𝑞 = 𝜀𝜎𝐵 (𝜃𝑠4 − 𝜃𝑎𝑚𝑏 ) (2-4)
The temperature of the underground power cables can be determined from the
expression of temperature rise above ambient for a known current flowing through the
conductor. In order to quantify the heat flow process around a conductor, an electrical
circuit equivalent can be used because there is analogy between the two flow
processes. The analogous circuit is referred to as a thermoelectric equivalent (TEE)
model. For a single buried cable, the typical TEE circuit is shown in Figure 2-1. The
typical underground cable consists of a conductor surrounded by an insulating
dielectric material, outer metallic screen and jacket material, and the surrounding
medium (soil/backfill). The dielectric insulation and metallic layers become more
48
___________________________________________________________________________________
complex with the increase of the cable nominal voltage. Quantities Wc, Wd1, Wd2 and
Ws denote heat losses from the conductor, dielectric material and the metallic screen
respectively. While Cc, Cd1, Cd2, Cj, and Csurrounding represent the electrical capacitance
of the conductor, dielectric material, cable jacket and the surroundings respectively.
T1 T3 T4
Wc Cc Wd1 Cd1 Cd2 Wd2 Cs Ws Cj Csurrounding
Conductor Insulating dielectric Metallic Jacket Surroundings

screen
Figure 2-1: Thermoelectric equivalent (TEE) model of a typical cable.
Calculation of the permissible steady state current rating of cables at all alternating
voltages, and dc voltage up to 5kV, at 100% load factor can be derived from equation
(2-5) or (2-6) below [44]. Equation (2-6) is derived from equation (2-5) by explicitly
expressing the current rating 𝐼 , whilst relations (2-7) to (2-9) are replaced in (2-6):
1
∆𝜃 = (𝑊𝑐 + 2 𝑊𝑑 ) 𝑇1 + [𝑊𝑐 (1 + 𝜆1 ) + 𝑊𝑑 ]𝑛𝑇2 + [𝑊𝑐 (1 + 𝜆1 + 𝜆2 ) +
(2-5)
𝑊𝑑 ]𝑛(𝑇3 + 𝑇4 )
2
∆𝜃 − 𝑊𝑑 [0.5𝑇1 + 𝑛(𝑇2 + 𝑇3 + 𝑇4 ]
𝐼= [ ] (2-6)
𝑅𝑎𝑐,𝜃𝑐 [𝑇1 + 𝑛(1 + 𝜆1 )𝑇2 + 𝑛(1 + 𝜆1 + 𝜆2 )(𝑇3 + 𝑇4 )]
𝑊𝑐 = 𝐼 2 𝑅𝑎𝑐,𝜃𝑐 (2-7)
𝑊𝑑 = 2𝜋𝑓 ∗ 𝑡𝑎𝑛 𝛿 ∗ 𝐶 ∗ 𝑈𝑜2 (2-8)
∆𝜃 = 𝜃𝑐 − 𝜃𝑠 (2-9)
where 𝐼 is current rating (amperes);
𝑇1 is the thermal resistance per unit length between a conductor and the
sheath (K m/W);
𝑇2 is the thermal resistance per unit length of the bedding between cable
sheath and armour (K m/W). it takes a value of zero if the cable does
not have an armour;
𝑇3 is the thermal resistance per unit length of the external serving (or
jacket) of the cable (K m/W);
49
___________________________________________________________________________________
𝑇4 is the thermal resistance between the cable surface and the

surrounding medium which can include the duct and the soil/backfill
(K m/W);
𝜆1 is the ratio of losses in the metal sheath to total losses in all
conductors of the cable;
𝜆2 is the ratio of losses in the armouring to total losses in all conductors
of the cable.
n is the number of equally sized load-carrying conductors in the cable
that carry the same load;
𝑈𝑜 is the phase to earth voltage (volts);
𝑊𝑐 is the Joule loss per unit length for the conductor insulation (W/m);
𝑊𝑑 is the dielectric loss per unit length for the conductor insulation
(W/m);
𝜃𝑠 is the soil ambient temperature (K)
𝜃𝑐 is the conductor temperature (K)
Δ𝜃 is conductor temperature rise above ambient (K);
𝑅𝑎𝑐,𝜃𝑐 is the alternating current resistance per unit length of the conductor at
𝜃𝑐 operating temperature (Ω/m);
𝐶 is the cable capacitance per unit length in (F/m) and;
𝑡𝑎𝑛 𝛿 is the insulation loss factor at power frequency.
The critical parameters include the conductor losses 𝑊𝑐 , the thermal resistances
(𝑇1 , 𝑇2 , 𝑇3 𝑎𝑛𝑑 𝑇4 ), dielectric loss factor 𝑊𝑑 and temperature difference Δ𝜃 due to
their relative magnitudes in equations (2-5) and (2-6). These critical parameters have a
bigger impact on the cable thermal rating as compared to the other ones [31, 41, 60,
61].
Calculation of losses
Relations in (2-7) and (2-8) show the expressions for the two major operational losses
when a cable is energised - dielectric losses and conductor losses.
50
___________________________________________________________________________________
Dielectric losses
The dielectric losses are dependent on the alternating voltage, insulation capacitance,
loss factor and the power frequency; they can be determined from relation (2-8).
Conductor losses
The conductor losses can be determined using the expression (2-7) and are the biggest
source of losses generated within the cable. The conductor losses is a function of two
variables - the operating current and the conductor resistance. The conductor
resistance 𝑅𝑎𝑐,𝜃𝑐 is a function of the surrounding ambient temperature, conductor
physics, material characteristics as well as the level of the operating electric current
[62].
Computation of conductor temperature dependent resistance
In general, conductors have different direct current (dc) and alternating current (ac)
resistance values. The dc resistance reflects the conductor size, material resistivity and
purity, while the ac resistance accounts for additional phenomena such as proximity
effect and skin effect that are functions of the current flowing through the conductor,
operating frequency and magnetic fields of any multiple conductors placed alongside
[44]. The ac resistance per unit length of any conductor 𝑅𝑎𝑐,𝜃𝑐 at an operating
temperature 𝜃𝑐 is calculated using the expression (2-10) [44]:
𝑅𝑎𝑐,𝜃𝑐 = 𝑅𝑑𝑐,𝜃𝑐 (1 + 𝑦𝑠 + 𝑦𝑝 ) (2-10)
where 𝑅𝑑𝑐,𝜃𝑐 is the dc current resistance of the conductor at 𝜃𝑐 operating temperature;

𝑦𝑠 is the skin effect factor; and 𝑦𝑝 is the proximity effect factor.
The skin and proximity effects exist due to alternating magnetic fields, which occur
within the conductors. These effects are influenced by the conductor cross-section,
electrical conductivity of the material, operating frequency and material permeability.
The skin effect arises due to self-inductance, which results in the current density being
higher at the outer surface of the conductor. Proximity effect arises due to mutual
inductance between the conductors of adjacent phases, thus making the current to
flow along one side of the conductor’s cross-section. Cables with cross-sections
smaller than 185 mm2 normally have their proximity effects ignored.
51
___________________________________________________________________________________
The dc current resistance of the conductor at 𝜃𝑐 operating temperature (/m) is

calculated using (2-11):
𝜌0
𝑅𝑑𝑐,𝜃𝑐 = ∙ (1 + 𝑍) ∙ [1 + 𝛼20 (𝜃𝑐 − 20°)] (2-11)
𝑎
where 𝜌0 is resistivity of conductor material at 20 ºC (Ω·m); 𝑎 is conductor cross-

sectional area (m2); 𝛼20 is the conductor temperature coefficient of electrical
resistivity at 20C that is dependent on the material type (𝛼 = 0.00393ºC-1 for copper).
Z is a factor, which allows for cable stranding and for the lay, 𝑍 = 0.03 for single
core cables with cross-sectional area ≤ 500mm2, while 𝑍 = 0.04 for larger cables.
The skin and proximity effects factors 𝑦𝑠 and 𝑦𝑝 of the conductor can be calculated
according to the method provided in [44] and are shown by equations (2-12) to (2-15)
below:
𝑋4
𝑠
𝑦𝑠 = 192+0.8∗𝑋 4 (2-12)
𝑠
2
8𝜋𝑓∗10−7 ∗𝑘𝑠
𝑋𝑠4 = ( ) (2-13)
𝑅𝑑𝑐,𝜃𝑐
𝑋𝑝4 𝑑 2 𝑑 2 1.18
𝑦𝑝 = 192+0.8∗𝑋 4 ∗ ( 𝑠𝑐 ) [0.312 ∗ ( 𝑠𝑐 ) + 𝑋4
] (2-14)
𝑝 𝑝
+0.27
192+0.8∗𝑋4
𝑝
2
8𝜋𝑓∗10−7 ∗𝑘𝑝
𝑋𝑝4 = ( ) (2-15)
𝑅𝑑𝑐,𝜃𝑐
where 𝑑𝑐 is the cable conductor diameter (mm); 𝑘𝑠 and 𝑘𝑝 are constants obtained
from Table 2 of [44]; 𝑓 is the ac current supply frequency (hertz); 𝑠 is the distance
between the conductor axis (mm); and 𝜋 = 3.142.
The determination of conductor resistance at particular temperature is required for

modelling purposes. The calculation method for finding the corrected resistance to
reflect the evolving temperatures can be determined using equations (2-16) to (2-18):
𝑅
𝑅𝑎𝑐,𝜃𝑐 = 𝑘 20 (2-16)
𝜃𝑐
1
𝑘𝜃𝑐 ,𝑎𝑙 = 1+0.00403(𝜃 (2-17)
𝑐 −20)
1
𝑘𝜃𝑐 ,𝑐𝑢 = 1+0.00393(𝜃 (2-18)
𝑐 −20)
where 𝑘𝜃𝑐 ,𝑎𝑙 is temperature correction factor for aluminium conductors; and 𝑘𝜃𝑐 ,𝑐𝑢 is
temperature correction factor for copper conductors.
52
___________________________________________________________________________________
Calculation of thermal resistances for single-core cables
During the network operations, there are four different thermal resistances inside the
cable. These thermal resistances are denoted as 𝑇1 , 𝑇2 , 𝑇3 𝑎𝑛𝑑 𝑇4 . The Figure 2-2
illustrate parameters and geometry of equations (2-12) to (2-24). Details of the values
used in this thesis studies are provided in the subsequent chapters (see Section 4.4.4
and Table 4-2).
Conductor
L ØDa
Ødc
Øds
S1
Figure 2-2: Underground cable geometry
Thermal resistance between one conductor and sheath (𝑻𝟏 )
The thermal resistance between the conductor core and the sheath 𝑇1 is given by the
relation (2-19).
𝜌𝑇 2𝑡1
𝑇1 = ∙ ln [1 + ] (2-19)
2𝜋 𝑑𝑐
where 𝜌𝑇 is the thermal resistivity of insulation (K·m/W); 𝑡1 is the thickness of
insulation between conductor and sheath (mm); 𝑑𝑐 is the diameter of the conductor
(mm).
53
___________________________________________________________________________________
Thermal resistance between sheath and armour (𝑻𝟐 )
The thermal resistance between the sheath and armour (if present) is given by the
equation (2-20).
𝜌𝑇 2𝑡2
𝑇2 = ∙ ln [1 + ] (2-20)
2𝜋 𝑑𝑠
where 𝑡2 is the thickness of the bedding (mm); and 𝑑𝑠 is the external diameter of the
sheath (mm).
Thermal resistance of outer covering (𝑻𝟑 )
The external servings are normally in the form of concentric layers and the thermal
resistance
𝑇3 between them is given by the equation (2-21).
𝜌𝑇 2𝑡3
𝑇3 = ∙ ln [1 + ] (2-21)
2𝜋 𝐷𝑎
where 𝑡3 is the thickness of serving (mm); and 𝐷𝑎 is the external diameter of the
armour (mm).
Thermal resistance between outer covering and soil (𝑻𝟒 )
The thermal resistance between the outer covering and the surrounding medium (soil)
for a single isolated buried cable is given by the relation (2-22) and (2-23).
𝜌𝑇
𝑇4 = 𝑠 ∙ ln [𝑢 + √𝑢2 − 1] (2-22)
2𝜋
2𝐿
𝑢= (2-23)
𝐷𝑒
where 𝜌𝑇 𝑠 is the thermal resistivity of the soil (K·m/W) and 𝑢 is given by the relation
(2-23) where 𝐿 is the distance from the surface of the ground to the cable axis (mm);
and 𝐷𝑒 is the external diameter of the cable (mm).
For installation conditions involving a set of three cables having approximately equal
losses, laid in a horizontal plane and equally spaced, 𝑇𝟒 is given by equation (2-24).
𝜌𝑇 𝑠 2𝐿 2
𝑇4 = ∙ {ln (𝑢 + √𝑢2 − 1) + ln [1 + ( ) ]} (2-24)
2𝜋 𝑠1
where 𝑠1 is the axial separation between two adjacent cables (mm).
54
___________________________________________________________________________________
The next section provides a description of the Arrhenius cable insulation ageing
model which is an essential part of the proposed IEC-Arrhenius-Weibull reliability
model for ageing cable. The cable thermal model will be built into the Arrhenius
model.
2.4 Arrhenius model of ageing cables
One approach to consider the networks’ operational factors in failure-reliability

models is application of the life-stress models such as Arrhenius ageing model [31,
35, 63]. There are two common life-stress models (Arrhenius and Eyring equations)
used to examine the rate of component ageing due to temperature changes (thermal
stresses) [35, 64]. The Arrhenius and Eyring equations can be used to relate any life
measure, e.g. characteristic life, mean, median etc. to the operating temperature. In
this thesis the Arrhenius ageing model is used since it has been widely accepted in the
cable ageing literature and has been used in reliability studies [30-32, 35, 36, 63].
The Arrhenius ageing model is used to determine the life expectancy of cables when
operating at both constant and varying temperatures. For example, the cable
remaining operating life to reach the minimum acceptable performance level can be
obtained using the Arrhenius equation for determining life expectancy [65]. The
Arrhenius model relates time and temperature with the deterioration of material and
thus attributes component failure to changes in the material property, such as
insulation in the case of cables. Relation (2-25) is the essential expression of the
Arrhenius model.
𝐵
( ) (2-25)
𝐿(𝜃𝑐 ) = 𝐴 𝑒 𝜃𝑐
where 𝐿(𝜃𝑐 ) component life time to reach minimum acceptable performance level;
A, B constants determined by the activation energy of chemical reactions;
𝜃𝑐 is the conductor temperature (K)
The quantity 𝐿(𝜃𝑐 ) from the relation in (2-25) can be used as the quantifiable cable
life (years). The cable life is dependent on several factors, among which the cable
loading profile is one of the most influential determinants. For this reason, the cable
life parameter 𝛼𝑙𝑖𝑓𝑒 will be linked with the Weibull distribution scale parameter 𝛼 (see
below). On the other hand, the IEC cable thermal model described in section 2.3.1
55
___________________________________________________________________________________
will be used to link the time varying temperature 𝜃𝑐 and the variable cable loading and
published in [36]. The parameters A and B in (2-25) are component specific constants
derived from empirical studies on cable insulations. The next section describes the
proposed method to estimate these parameters.
2.5 Proposed method for the determination of Arrhenius parameters
The A and B parameters of the Arrhenius model shown in (2-25) are not available for
most network cables because of the high costs associated with the required studies
[31, 63]. A novel way of determining the Arrhenius parameters A and B is proposed
and described in this thesis and published in [36]. Cable manufacturers specify cable
nominal life 𝐿𝑛 under the nominal loading conditions, or constant continuous
temperature 𝜃𝑐,𝑛 . This is often equal to 40 years for the constant temperature of 90C
(or 363K). Replacing these values into Arrhenius model (2-25) gives 𝐿𝑛 (∙) = 𝐴 ∗
𝐵
𝑒 363 , where 𝐿𝑛 is either in years or in hours. Assuming conductor temperature in
hourly interval p=1,2,…,m is 𝜃𝑐,𝑝 , the relative life in interval p is calculated by
dividing life 𝐿( ) from (2-25) by nominal life 𝐿𝑛 to generate the expression (2-26):
𝐵 𝐵
( − )
𝜃𝑐𝑝 363 (2-26)
𝑙𝑝 (∙) = 𝑒
where 𝑙𝑝 ( ) is the Arrhenius relative life in interval p. Equation (2-26) shows that
where conductor temperature 𝜃𝑐,𝑝 is greater than nominal temperature 𝜃𝑐,𝑛 relative life
is less than unity, and vice versa. Summation of relative Arrhenius lives over the
entire historic period p=1,2,…,m gives the total relative cable life (2-27):
𝑚 𝐵 𝐵
( − ) (2-27)
𝜃𝑐𝑝 363
𝑙(∙) = ∑ 𝑒
𝑝=1
End-of-life statistical data from the international study on ageing assets [6] were
further used to estimate parameters A and B. The main characteristics are the mean
and the standard deviation of the PILC cable life estimates found to be 51 and 20
years respectively. In our study, it was assumed that the actual cable age can follow
the life distribution provided in [6] and is shown in Figure 2-3.
56
___________________________________________________________________________________
Figure 2-3: Age distribution and life estimates of underground cables [6]
Assuming Normal distribution is a good fit to the cable end-of-life data, the seven-
step discrete approximation has been adopted and is shown by expressions (2-28) [2,
20]:
𝑃1 (𝜇) = 0.382,
𝑃2 (𝜇 + 𝜎) = 𝑃3 (𝜇 − 𝜎) = 0.242,
𝑃4 (𝜇 + 2𝜎) = 𝑃5 (𝜇 − 2𝜎) = 0.061,
𝑃6 (𝜇 + 3𝜎) = 𝑃7 (𝜇 − 3𝜎) = 0.006 (2-28)
where 𝑃𝑖 (𝜇) is probability, µ is mean and σ is standard deviation. These probabilities
are shown in Figure 2-4.
Figure 2-4: Probabilities associated with normal distribution
57
___________________________________________________________________________________
As can be observed from Figure 2-4, the probabilities decrease as the value of the
random variable increases above µ and at some distance, the probability of an interval
can be approximately zero. Furthermore, more than 0.9973 of probability of the
normal distribution falls within the range (𝜇 − 3𝜎) to (𝜇 + 3𝜎).
Assuming the cable age can be one of the values in (2-28), parameters A and B are
then calculated for each discrete step in turn using (2-29) and (2-30):
𝑚 𝐵 𝐵
1 ( 𝑖 − 𝑖 )
363
𝑙𝑟𝑒𝑙,𝑖 = ∑( ) 𝑒 𝜃𝑐𝑝 (2-29)
8760ℎ ∗ 𝐴𝑐𝑡𝑢𝑎𝑙 𝑎𝑔𝑒
𝑝=1
𝑚 𝐵𝑖
( )
𝜃𝑐𝑝
𝐿𝑖 = 𝐴𝑖 ∑ 𝑒 (2-30)
𝑝=1
where 𝑙𝑟𝑒𝑙,𝑖 , i=1,2,…,7 denotes cable relative life calculated as the ratio of the discrete
statistical life 𝐿𝑖 (i.e. µ, µ± σ, µ± 2σ, µ± 3σ) and the actual cable age. The final
estimates are weighted parameter averages (weights are 𝑃𝑖 ( ) from (2-28)):
𝐴̂ = ∑7𝑖=1 𝑃𝑖 ∙ 𝐴𝑖 & 𝐵̂ = ∑7𝑖=1 𝑃𝑖 ∙ 𝐵𝑖 (2-31)
The main steps of the iterative calculation procedure are outlined below:
1) Input cable historic loading data, or if unavailable, assume relative historic
hourly loading of cables over the historic period (i.e. 8760h * actual cable age
in year). Where no structural changes occurred in the past, the same (relative)
hourly profile can be assumed within all historic years.
2) Initiate three loops a, b and c:
a) The outer loop is over seven discrete cable lives i=1,2,…,7 as defined by
equation (2-28);
b) The mid loop represents iterations h=1,2,… until parameter Bih converges;
and
c) Inner loop is over hourly intervals of the historic period p=1,2,…,m.
3) Take into consideration historic hourly interval p in which relative loading
Irel,p is specified (Irel,p 𝜖{0, 1}). Assume an uncertainty window ωp which is
typically a percentage of the relative loading Irel,p. Determine the cable loading
Ip in Amps from Irel,p, uncertainty window, generated uniform random number
and peak loading in the historic year.
4) Input ambient temperature and determine load temperature of the cable
conductor 𝜃𝑐𝑝 by substituting the previously calculated cable loading Ip into
58
___________________________________________________________________________________
the cable thermal model – (see Section 2.3.1). Other cable data, dependent on
cable and installation characteristics, are used in this model, as inputs to the
cable thermal model. Build the corresponding term of equation (2-29), where
parameter Bi estimate Bih is being used.
5) Increment index p=p+1 and repeat steps 3) and 4) until all historic hourly
periods p are considered, i.e. p=m in (2-30). This gives estimated cable life
lrel,ih in the h-th numerical iteration with the estimated parameter Bih.
6) Set parameter Bi estimate to Bih+1 so that the improved cable life lrel,ih+1 (2-29)
can be obtained. Repeat the steps 3), 4) and 5). Go to the next step if
convergence of the mid loop is obtained. The convergence criterion is lrel,ih+1 –
lrel,i ≤ε , where lrel,i , i=1,2,…7 is defined by the end-of-life statistical data [6]
and ε is assumed tolerance.
7) After having calculated converged Bih+1, find parameter Ai from equation
(2-30).
8) Repeat steps 3) to 7) for each discrete life lrel,i, for i=1,2,…7. After having
completed the outer loop, calculate final parameter estimates 𝐴̂ and 𝐵̂ from
(2-31).
The described method for finding Arrhenius parameters has been applied to derive the
proposed IEC-Arrhenius-Weibull reliability model. This model will be presented in
the next chapter; however, a cable thermal loss-of-life model, which is used for cable
replacement ranking, is given in the next section.
2.6 Cable thermal loss of life model
Cable thermal loss-of-life fractions during periodic load cycles can be computed using
the Miner’s theory for time to failure inference in the presence of time varying
stresses [66, 67]. Miner suggested that ageing accumulates over a component life
cycle and it follows a process whereby short periods of severe degradation may be
compensated by long periods of tolerable degradation. The theorem stated that, at
failure point, the sum of all losses of life fractions is equal to unity [65]. This
postulate is generally used as a tool to analyse the life of devices subjected to time
varying stresses. The total cable thermal loss-of-life is calculated from relation (2-32).
59
___________________________________________________________________________________
𝑚
𝑡𝑝
∑ = 1, (2-32)
𝑡𝑓𝑝
𝑝=1
where 𝑡𝑝 is the time spent at absolute temperature 𝑇𝑝 ; 𝑡𝑓𝑝 is the life value at a
constant absolute temperature 𝑇𝑝 , based on the Arrhenius model; 𝑚 is the number of
periods (fractions) of the total life.
The exposure to higher temperatures leads to chemical changes in cable insulation and
reduced life. The cable life can be related to the conductor temperature using the
Arrhenius relationship (2-25). Based on this model, it has been shown in [65-67] that
the (relative) thermal loss-of-life at variable conductor temperatures can be expressed
as shown in (2-33):
𝑀 𝑁
𝑡𝑝 𝑡𝑞
𝐿𝑇 = ∑ +∑ = 𝐿′ 𝑇 + 𝐿′′ 𝑇 (2-33)
𝐿𝑝 (𝜃𝑐,𝑝 ) 𝐿𝑞 (𝜃𝑐,𝑞 )
𝑝=1 𝑞=1
where the first sum LꞌT denotes loss-of-life in the historic period (M intervals) and
second sum LꞌꞌT in the planning period (N intervals). The subscripts p and q denote
historic and planning intervals, 𝑡𝑝 and 𝑡𝑞 are durations of interval p and j in which
conductor temperature is constant, and Lp and Lq represent lives at temperatures θc,p
and θc,q, obtained from the Arrhenius model (2-25).
Equation (2-33) represents a summation of fractional loss-of-life increments in the

considered historic and planning periods. The sum approaches unity (or, the nominal
number of periods) when the cable reaches the end of thermal life (i.e. replacement is
required). The model developed in the thesis suggests that, those cables with the
highest thermal loss-of-life are the most aged and most prone to failures. In this study,
thermal loss-of-life technique will be used for ranking of network cables and for the
comparison with the proposed ranking schemes. In addition, the study only considers
the thermal effect on ageing under steady state power flows and ignores transient
analysis for reasons of simplification. An advanced method could be formulated to
consider the dielectric stress and related ageing of insulation under the HV
surge/overvoltage transients.
60
___________________________________________________________________________________
2.7 Chapter summary
This chapter presented the details of thermal modelling of ageing MV cables and its
importance for effective reliability evaluation in planning of distribution systems.
This study is needed to understand the impact of increased power flows (through
proper quantification of cable ageing effects and utilisation flexibility) on
underground cables. Accuracy of methods used for planning of distribution networks
is crucial particularly when the network is operated under high load conditions. The
need for incorporating the thermal modelling is a step towards improving the accuracy
of planning decisions.
The inclusion of cable thermal properties enabled the computation of the cable thermal
loss-of-life. A cable ranking based on the cumulative cable thermal loss-of-life that can
be used as a guide for prioritised replacements of network assets is therefore presented.
The cable thermal model and Arrhenius ageing model will be combined with the
Weibull distribution to get the proposed reliability model of ageing cable.
61
CHAPTER THREE
RELIABILITY ASSESSMENT OF AGEING CABLES
3 Reliability assessment considering end-of-life failure
Mathematical modelling of component failures is one of the central problems in

reliability analysis. Modelling and reliability assessment of ageing network assets
(especially cables and transformers) is not new; however, a detailed quantitative
approach is yet to be developed.
This chapter provides details of some important elements incorporated into the
proposed reliability assessment technique that is applied to modern distribution
systems. The aim is to develop a reliability model of ageing components that will
consider the physical aspects of components, as well as impact of individual
component ‘health’ on system indices. This section presents the reliability models for
modelling ageing components. The first part of the chapter gives basic definitions and
expression from the reliability theory and ends up with the description of Weibull
probability distribution function and renewal process; the second part presents the
proposed IEC-Arrhenius-Weibull model for modelling of uptimes of ageing cable.
This model will be incorporated into the proposed reliability assessment methodology
and used to predict the reliability performance of the planned distribution system.
3.1 Literature review
The existing literature on how ageing affects reliability is scarce due to conventional
methods for failure statistics reporting. The purpose of modelling the reliability of
ageing components is to predict future failure behaviour with some level of
confidence. In general, power system components exhibit one of two failure states;
repairable and non-reparable (or end-of-life) failures. Most references provide
network analysis based on reparable failures and thus tend to ignore the impacts of
non-reparable failures. Ageing failures form part of the non-reparable failures and
occur mostly on those components nearing their end of design life, or on components
that have experienced onerous operating conditions, such as overloading for an
extended period.
62
Chapter three: Reliability assessment of ageing cables
___________________________________________________________________________________
Once an ageing failure occurs, it is likely that a replacement would have to be

arranged. In addition, the occurrence of an ageing failure of one component could lead
to a cascading ageing related failure on other network components due to the
additional stresses they are likely to be exposed to. This type of events makes it
imperative to consider ageing failures in planning of future energy networks. In power
systems reliability evaluations, the failure rate function of a component can be
illustrated using the bathtub curve given in Figure 3-1 [3]. The bathtub curve consists
of three phases; early life is characterised with high but decreasing failure rate, useful
life has constant (low) failure rate and wear out (or end-of-life) stage with steeply
increasing failure rate.
Early life Useful life Wear-out

Failure rate
Time
Figure 3-1: Bathtub Curve
Figure 3-1 shows that reliability modelling of a component shall include an early life,
useful life and wear out phases. Early life failures are mostly caused by several
reasons such as manufacturing defects, poor processes, poor quality control, human
errors etc. during the components initial years of usage. The constant failure rate
phase covers the period during which the component failures occur randomly and
constantly over substantial time periods. The potential causes of failures in this period
include, low safety factors, undetectable defects, human errors, abuse, natural failures,
higher random stress than expected, etc. In the wear-out failure period, higher
frequencies of failures occur mainly due to the wear caused by ageing of components.
Other factors include poor maintenance, incorrect overhaul practices, corrosion, and
short designed in-life of the item, etc. The end-of-life failures discussed in this review
concern failures occurring at the final stage of the bathtub curve.
63
___________________________________________________________________________________
In a given system, the ‘appropriate’ failure (or hazard) model should be specified
consistently with the period in which it is assessed. In traditional reliability evaluation
techniques, the ‘useful life period’ of all components is considered. The component
time to failure (up times) and time to repair (down times) are represented by the
exponential distribution function [20]. In this case, time-homogeneous Markov
process is used to evaluate the system reliability. However, during the final stages of
component life - wear-out period, the failure rate is rising, thus the use of Markov
processes is no longer valid. In this case, the process becomes non-Markovian. One of
the ways to represent non-Markovian processes in reliability evaluations is to use of
other probability distributions for uptimes, such as Log-normal, Weibull distribution,
etc. [2, 11, 19, 20, 68, 69].
In this thesis, focus is on end-of-life (ageing) failures for reasons of simplification. In

addition, following a maintenance process, the concept of virtual aging is incorporated
into the computation of the component’s future state - minimal and general repair
types (see Figure 3-5). An advanced approach could be to analyse the whole
component lifetime as described by the bathtub curve of Figure 3-1 and consider all
the three types of repair processes.
The literature review of methodologies that incorporate ageing of distribution

components within network planning assessments can be classified into three
categories:
 Planning techniques that address issues related to component end-of-

life failure modelling,
 End-of-life modelling incorporated into distribution systems
reliability assessments, and
 Applications in replacement planning
The conventional reliability evaluation techniques that consider components ageing

are based on historic statistics, and the probabilistic distribution functions (pdf) are
determined from historical failure data [20, 69-73]. The derived pdfs can be
determined using parametric approaches [20, 74, 75]. Alternatively, non-parametric
approaches have also been used to develop components failure functions.
64
___________________________________________________________________________________
When the parametric approach is used, several probabilistic distribution models are
tried and the parameters obtained from the ‘best fit’ model are calculated using the
goodness-of-fit tests. The historic failure data are modelled using the parametric
methods, such as Weibull pdf or Log-normal pdf to represent the end-of-life failure
instead of the exponential pdf because of its memory-less property [20].
The parametric methods assume that uptimes as well as failures are independent and
identically distributed based on the theory of stochastic point processes; this
assumption may be inconvenient when dealing with modelling of very rare events
within the component repairable failure states [20, 75-78]. The authors of [76, 78, 79]
used statistical hypothesis testing of historical failure data of several distribution
system components, and concluded that if the inter-arrival times for the uptimes show
neither improvement nor deterioration, then a zero-ageing trend (i.e. renewal process)
can be safely adopted. In the opposite case if an increasing failure rate is recorded, the
component is said to exhibit a positive-ageing trend (i.e. a deterioration of the
component ‘health’), otherwise it is considered to exhibit negative-ageing trend for a
fairly improving inter-arrival times (i.e. decreasing failure rate). The authors of [43]
focused on the impacts of ageing of several MV-distribution network cables on other
neighbouring cables and the network reliability indices by developing a failure model
for each of the different modes of failure (e.g. active failures, earth faults, multiple
earth faults, etc.) from the historical failure data. Furthermore, active failures were
represented using exponential pdf failure rates while the earth faults and double earth
faults with Weibull pdf in the reported studies.
An attempt to improve the modelling of the ageing failure states (i.e. inter-arrival
times of the up states) within parametric methods was presented in [19]; it is based on
introduction of two unavailability parameters in a non-sequential MCS, one for
reparable and one for ageing failures. According to this method, two random numbers
are drawn to determine if the component is unavailable due to a reparable or a
permanent ageing failure. The unavailability due to ageing is consistently updated
after a pre-set time period (e.g. ten simulation years). A recent advancement of the
method developed in [19] considers component failures due to ageing using the state
duration sampling technique and three variant methods [68]. Component (generator)
ageing failures are modelled using the non-homogenous Poisson process (NHPP), and
65
___________________________________________________________________________________
in particular the power law process. Consideration was also given to the sequential
time dependence of ageing failures and its impact on the determination of generating
system states within the reliability evaluation. However, the components modelled in
the reliability evaluation are ranked using the historical failure data in order to
establish the critical assets that need replacement - reinforcement.
The methods from the second group, the non-parametric approaches, do not depend
on the parameters derived from a predefined pdf. They consist of infinite (growing
number) parameters and their predictions depend on existing parameters as well as
current data. In addition, the non-parametric approaches capture the subtle aspects of
the component lifetime data (relatively more accurate); are relatively more flexible;
and have more degrees of freedom. However, more computation time is required.
The failure rate function is estimated using estimation functions, such as additive
model, neural networks, multiple additive regression splines and Kaplan-Meier’s
estimator, and also the survival function is found from component lifetime data [80].
Most studies involving this technique have focused on improving the estimation
accuracy [74, 75]. However, in the absence of additional assumptions, the likelihood
of a future event may depend on historical events in both type and duration of the
occurrence. This can be inconvenient because historical records can be unreliable (or
even corrupt) due to inherent incompleteness caused by the various forms of data
collection, classification, misinterpretation, etc. [74, 75, 80]. Thus, the non-parametric
estimators seem to be not so good alternative to the models based on parametric
approaches.
To this end, a novel approach based on the Arrhenius-Weibull distribution was

proposed and applied to the end-of-life characterisation of power transformers [35,
81]. This technique is based on the fact that power system components have a very
long life and there is a lack of sufficient failure data that cover all possible failure
scenarios and operational conditions. These operational conditions, such as thermal
stresses due to high loading and voltage surges caused by switching transients, play
significant role in determining the components failure models represented with
‘appropriate’ pdfs. This was the first study to integrate Arrhenius ageing model
(unconventional end-of-life models) into system reliability assessments. This new
66
___________________________________________________________________________________
failure modelling method is characterised by Arrhenius-Weibull distribution as

described by the expression (3-1) and shown in [81].
𝑡
𝐹(𝑡) = 1 − 𝑒𝑥𝑝 (− ) (3-1)
𝐵
𝐴 ∙ exp(Θ )
𝐻𝑆𝑇
where Θ𝐻𝑆𝑇 is the hot spot temperature of the transformer winding in kelvin,  is the
shape parameter of the Weibull pdf.  is estimated by fitting the historical temperature
data to a Weibull distribution. A and B are empirical constants (Arrhenius parameters)
estimated from historical transformer life/age (); IEC 60076-7 Standard and;
historical temperature data.
Several assumptions were adopted on the parameters used in the studies based on the
relation (3-1): Gamma function was used to estimate the Weibull parameter  and; the
corresponding values of transformer life parameter  are computed based on the
average hot spot temperatures (at different loading levels). The Weibull parameter 
was then assumed a constant equal to 5 throughout the subsequent reliability studies.
However, since A, B and  are considered to be fixed for the transformers, this means
that only  needs to be calculated (for different loading levels), thus allowing only a
small data set to be analysed in the study. Furthermore, the reliability evaluation
technique deployed is based on non-sequential MCS, thus ignoring the impacts of
events chronology, which is a key characteristic of component ageing evaluations [35,
81].
To the best knowledge of the author, current research did not consider the combined
impact of the dynamic loading (i.e. time varying load), thermal stress model, along
with the current age and remaining component life in determining the state durations
within the reliability evaluation. These factors cannot be ignored when realistic
conditions need to be considered in the system reliability evaluation for planning of
future networks. More specifically, planning of modern networks requires application
of comprehensive and robust predictive reliability models such as SMCS. This is
based on the recently introduced regulatory quality-of-supply incentives which
impose requirements to meet specific reliability targets; this in turn requires to
67
___________________________________________________________________________________
optimize ‘reliability’ spending based on the cost per reliability benefit (i.e. reliability
cost/worth) [4, 11, 24].
The proposed IEC-Arrhenius-Weibull approach involves defining the scale parameter

of the Weibull distribution using the Arrhenius thermal model and tuning the shape
parameter. The Arrhenius ageing model takes into account physics of the component
as well as the thermal stresses in terms of the component temperature [65]. The latter
can be determined from the IEC thermal model that defines functional relationship
between the component temperature and loading [44]. In this way, IEC thermal model
is incorporated in the Arrhenius ageing model, which is in turn substituted in the
Weibull distribution. Other parameters of the Arrhenius model are derived from the
components available historical operations data. This resulted in the combined IEC-
Arrhenius-Weibull model, which is used to determine up times of ageing components
and is fed as an input into the probabilistic reliability evaluation methodology and
published in [36].
This thesis examines more closely the modelling of cable ageing in the reliability
evaluation of medium voltage (MV) urban distribution networks. The thesis proposes
a new methodology based on the SMCS, which incorporates the effects of ambient
conditions, dynamic loading, thermal stress, and current components age in order to
determine the durations of up-states of ageing network components. The proposed
approach is applied to different types of underground cable which are the major
component in MV urban networks.
3.2 Basic features of reliability distribution functions
Prediction of a component’s (e.g. cable) uptime (which can be lifetime in case of non-
reparable failure) is a fundamental issue, which requires application of the probability
– reliability theory. The lifetime of a component can be modelled by a random
variable 𝑋 and a probability distribution function used to define the probability that
the lifetime is less or equal to 𝑥, where x is an observation of X. Below is a description
of the essential properties of the probability – reliability functions.
68
___________________________________________________________________________________
Probability functions
There are four essential probability – reliability functions:

1. The probability distribution (density) function (p.d.f.), f(x), indicates the point
in time when a failure is more or less likely to occur, thus:
∞
(3-2)
∫ 𝑓(𝑥)𝑑𝑥 = 1
0
The probability that the lifetime (or uptime) X is between the periods of
special interest, a and b, is then defined as:
𝑏
𝑃(𝑎 ≤ 𝑋 ≤ 𝑏) = ∫ 𝑓(𝑥)𝑑𝑥 (3-3)
𝑎
2. The cumulative distribution function (cdf), is given as:

𝑥
𝐹(𝑋) = 𝑃(𝑋 ≤ 𝑥) = ∫ 𝑓(𝑢)𝑑𝑢 (3-4)
0
The cdf is the probability that a failure occurs prior to the time instant x.
3. The reliability function is defined as:
∞
𝑅(𝑥) = 𝑃(𝑋 ≤ 𝑥) = ∫ 𝑓(𝑢)𝑑𝑢 (3-5)
𝑥
This is the probability that no failure occurs prior to time x.

4. The failure rate function (or hazard function) 𝜆(𝑥) is given by:
𝐹(𝑥 + Δ𝑥) − 𝐹(𝑥) 𝑓(𝑥)
𝜆(𝑥) = lim = (3-6)
Δ𝑥→0 Δ𝑡 ∙ 𝑅(𝑥) 𝑅(𝑥)
The relation (3-6) is the conditional probability of experiencing a failure in the

next instant of time x+Δx, subject to the component survival at time x. Hence,
the shape of the failure rate function is representative of the nature of failures
[70].
From the above definitions, if any of the four probability functions (3-2) to (3-6) is
known, the remaining can be determined [82]. In reliability theory, t is a stochastic
variable typically used to represents component up time or downtime. Since this is the
case, the cumulative distribution function 𝐹(𝑡) represents the unreliability (or failure)
69
___________________________________________________________________________________
function. The connection between reliability function (or survival function) 𝑅(𝑡) and
the unreliability – failure function 𝐹(𝑡) is described by:
𝑅(𝑡) = 1 − 𝐹(𝑡) (3-7)
It can be shown that the failure rate function (or hazard function) 𝜆(𝑡) is given by
[20]:
𝑓(𝑡)
𝜆(𝑡) = (3-8)
𝑅(𝑡)
Several probability distribution functions are used for modelling stochastic nature of
components in reliability studies. These distribution functions include Exponential,
Weibull, Normal, Log normal etc. [11, 20, 71]. The next section describes the Weibull
pdf, which is commonly used for modelling of ageing components in reliability
studies.
3.3 Exponential probability distribution function
The exponential pdf is the most common distribution function used in reliability
evaluations. It is specified by a single parameter, lambda (𝜆𝑒 ) - event rate, and it has a
memoryless property because the probability of a failure event depends on the current
time and not on previous failure events. The exponential pdf can exhibit positive and
or negative characteristics. The negative exponential distribution can be considered as
a special case of another distribution function (Poisson distribution). Thus, it is used
in the analysis of Poisson processes – a process in which events occur independently
and continuously at a constant average rate as described by equation (3-9) also shown
in Figure 3-2. It is applied in modelling reparable components that are considered to
be within their useful life cycle. In this thesis, exponential pdf is applied in evaluating
the reliability of non-ageing components of the distribution system.
𝑓(𝑥) = 𝜆𝑒 𝑒 −𝜆𝑒𝑥 (3-9)
70
___________________________________________________________________________________
Figure 3-2: Effects of different failure rates on exponential pdf
3.4 Weibull probability distribution function
In order to quantify the risk of end-of-life failures, it is important to assess the

components’ end-of-life failure probability. In reliability assessment studies, the
random variable is often time and the probabilities of occurrence of time dependent
events (failures or repairs) are defined using probability distribution functions.
Different functions can be used for up-times (i.e. component is in service and it fails
at random instant) and downtimes (i.e. component is out of service and the repair is
completed at a random moment).
i. One way of modelling the stochastic behaviour of a component is to assume

that its uptime can be represented by a Weibull distribution function [35, 68,
71, 83, 84]. The Weibull distribution function can have several shapes which
is dependent on the selected function parameters. It has three parameters –
location (shift or threshold parameter), shape parameter, 𝛽 and scale
parameter, , all of which provide great flexibility to model a wide range of
data sets. The region observed in the ꞌbath tubꞌ curve between the useful life
and end-of-life can be analysed using the Weibull probability distribution
function [2, 20, 35, 68].
The final phase of the standard bathtub curve (end-of-life failures) can be used to
represent the ageing behaviour of the component within reliability modelling (see
Figure 3-1). During the ageing phase, the failure rate 𝜆(𝑡) is expected to increase with
time. The Weibull probability distribution function (pdf) is a parametric model that
71
___________________________________________________________________________________
can suitably represent this behaviour [2, 85]. The Weibull pdf is shown in equation
(3-10):
𝛽𝑡𝛽−1 (−( 𝑡 )𝛽)

𝑓(𝑡) = 𝑒 𝛼 (3-10)
𝛼𝛽
where 𝑡 is the age of the component; β is the shape parameter; and α is the scale
parameter of the Weibull distribution. β parameter is also the Weibull slope
(dimensionless), thus different values of the parameter can impact the behaviour of
the distribution. The effects of different values of β parameter on the Weibull pdf is
shown in Figure 3-3: (a) – (d). Typical shapes that can be produced for the Weibull
distribution are presented in Figure 3-3; Figure 3-3(a) presents plots of Weibull’s
failure rate for variable  parameter and constant values of  while Figure 3-3 (b)
presents the variation plots when  parameter is varied and  parameter is kept
constant. Similarly approach is used to show the effects of variations of both
parameters on the hazard rate function in Figure 3-3 (c) and (d). Some deductions
from Figure 3-3 (c) and (d) include:
 A constant failure rate is obtained when  = 1 (identical to

exponential pdf);
 An increasing failure rate is obtained when  > 1;
 A decreasing hazard rate is obtained when  < 1;
 The variable feature of  and  parameters makes the Weibull pdf
adaptable to many real life processes.
72
___________________________________________________________________________________
Figure 3-3: Effects of different failure rates on Weibull pdf
In this thesis, α is represented by the component lifetime 𝐿(𝜃𝑐 ) of equation (2-25).

The shape parameter β needs to be β > 1.0 in order to use the Weibull pdf for
modelling of ageing failures.
The next section provides further details on the effects of the Weibull parameters used
in the subsequent analysis.
3.5 Weibull renewal process
There are two types of stochastic processes that can be represented by the Weibull
distribution: the Weibull renewal process and the non-Homogenous Poisson process
(NHPP). The processes are inspired by statistical investigations into reparable
systems/components. When upon the failure of a component it is repaired to a state
‘same as new’, a Weibull renewal process best models the stochastic characteristic of
such component. On the other hand, if the component’s reliability does not change
after a repair process – has the same age, then the appropriate modelling approach is
the NHPP model [85].
73
___________________________________________________________________________________
The survivor function 𝑅(𝑡) and time-dependent failure rate (hazard rate) related to the
NHPP [85, 86] are given by equations (3-11) and (3-12) respectively:
∞ 𝑡 𝛽
(−( ) )
𝑅(𝑡) = ∫ 𝑓(𝑡)𝑑𝑡 = 𝑒 𝛼
(3-11)
𝑡
𝑓(𝑡) 𝛽𝑡𝛽−1
𝜆(𝑡) = = (3-12)
𝑅(𝑡) 𝛼𝛽
This implies that the time-dependent failure rate in (3-12) can also be described by
equations (3-13) and (3-14) in line with the NHPP model:
𝜆(𝑡) = 𝜆𝛽𝑡𝛽−1 (3-13)

1
𝜆= (3-14)
𝛼𝛽
where 𝜆 represents the components’ historic failure rate parameter; 𝛼 is a life scaling
parameter and can be determined by calculating the life expectancy of the component.
The failure rate (3-10) or (3-11) is not constant over time and depends on the values
of the shape parameter  and scale parameter  as shown in Figure 3-3. Weibull scale
parameter  values reflect intensity of cable ageing; they increase quite rapidly
(particularly SAIDI) when  increases and; they also show the rate of change of the
failure rate (thus, can be an indicator of cable age). Practical choice of  should be
based on comparison with real-life data on historic performance.
The next section describes the development of the new IEC-Arrhenius-Weibull

model.
3.6 Developed IEC-Arrhenius-Weibull reliability model for ageing cable
In reliability assessment of power systems, a network component can be modelled as

a repairable component and the inter-transition times between each component down
time due to repairs as a stochastic point process {𝑁(𝑡), 𝑡 ≥ 0}. Here, 𝑁(𝑡) is the
number of up time events whose individual duration is time to failure (TTF). The up-
times are denoted as 𝑥𝑖 in the considered time interval, and the arrival time is 𝑡 as
shown in Figure 3-4.
74
___________________________________________________________________________________
Figure 3-4: A graph of stochastic point process
For more accurate modelling of ageing in reparable systems, the concept of virtual
age can be introduced to represent the actual (operational) age. The Kijima model
developed in [87] and successfully applied in [68] to model ageing generators is also
used here to represent the functional ageing characteristics of ageing cable. The
Kijima model states that a repair adjustment factor 𝑞 affects the status of the
component 𝑉𝑘 each time it undergoes a repair process. In that respect, the ageing
effect at the 𝑘 𝑡ℎ repair stage can be modelled by equation (3-15):
𝑉𝑘 = 𝑉𝑘−1 + 𝑞𝑥𝑘 = 𝑞𝑡𝑘 , 𝑓𝑜𝑟 𝑘 = 1,2,3 …. (3-15)
When 𝑞 = 0, it means that the component is repaired and returns to service as new
(i.e. unaffected by ageing), while 𝑞 = 1 means that the repaired component is only as
good as just before the previous failure. Any value 0≤q≤1 indicated a general repair
that improves the condition of the component. The effects of the various values of q
on the repair time are described in Figure 3-5 . A minimal repair can be a minor
maintenance action (such as damaged insulation because of accidental dig on the
cable insulation) that returns the component (cable) to its state of health as just before
the minor repair action. A renewal repair returns the component to ‘as good as new’
state while a general repair improves the system state (such a scheduled maintenance).
In this thesis, the minimal and general repairs were considered in the analysis as this
is assumed the most frequent cases in the industry.
75
___________________________________________________________________________________
Renewal
time
Minimal
repair
failure rate
General
repair
Minimal
repair
time t
Figure 3-5: Minimal and general repair
From Figure 3-4 and Figure 3-5, if a failure occurred at time 𝑡𝑘 , the time to next
failure at time 𝑡𝑘+1 can be determined using equation (3-16).
𝑡𝑘+1 = 𝑡𝑘 + 𝑥 (3-16)
where 𝑥 is the sum of all previous times to failure before 𝑡𝑘 .
The Weibull cumulative distribution function (cdf) for the inter-transition time (i.e.
inter-failure time) between states (𝑘 + 1) and (𝑘) is given by the relation in (3-17);
𝛽 −(𝑞𝑡 )𝛽 ]
𝐹𝑡𝑘+1 (𝑥) = {1 − 𝑒 −𝜆[(𝑞𝑡𝑘 +𝑥) 𝑘 } (3-17)
The next step is to determine a random inter-failure time 𝑥𝑘 by using a randomly

generated number from the Uniform distribution (0, 1]. The number is then applied to
the Weibull cdf (3-17), which is given by relation (3-18). Recognizing that random
number 1-U has also uniform distribution in the same range and can be used instead,
the last part of relation (3-18) can be established;
76
___________________________________________________________________________________
𝑈 = 𝐹𝑡𝑘+1 (𝑥)
𝛽 −(𝑞𝑡 )𝛽 ]
𝑈 = {1 − 𝑒 −𝜆[(𝑞𝑡𝑘+𝑥) 𝑘 }
𝛽 −(𝑞𝑡 )𝛽 ]
𝑈 = { 𝑒 −𝜆[(𝑞𝑡𝑘+𝑥) 𝑘 }
(3-18)
By taking the natural logarithm of left and right hand sides of equation (3-18) and re-
arranging the terms will result into equations (3-19) and (3-20).
1⁄
ln 𝑈 𝛽
𝑥𝑘 = − ( ) , 𝑓𝑜𝑟 𝑘 = 1 (3-19)
𝜆
1⁄
𝛽 𝛽
𝑘−1 𝑘−1
ln 𝑈
𝑥𝑘 = ( {𝑞 (∑ 𝑥𝑖 )} − ( )) − 𝑞 (∑ 𝑥𝑖 ) , 𝑓𝑜𝑟 𝑘 > 1 (3-20)
𝜆
𝑖=1 𝑖=1
In order to incorporate asset physical characteristics and operational conditions, a new

reliability model for ageing cable is proposed in this thesis. The new method
acknowledges that the network cable life depends on the dynamic loading and its
physical properties when determining the time to failure (or up-time). The Arrhenius
model, whose parameters are determined by the parameter estimation method
proposed in Section 2.4, is incorporated in the methodology for determining the state
durations of the cable circuits. The new IEC-Arrhenius–Weibull cable model is
developed as an improvement to the models proposed in [35, 68]. It is based on the
physical meaning of the Weibull scale parameter 𝛼 and the cable temperature-loading
characteristic.
The parameter 𝛼 of the Weibull pdf is defined as the life scaling parameter
determined by calculating the life expectancy of the component – relation (2-25).
According to the authors of [35] the 𝛼 parameter can be related to the Arrhenius
ageing model, and the Arrhenius ageing model is used to determine the life
expectancy of cables when operating at both constant and varying temperatures. For
example, cable remaining operating life to reach the minimum acceptable
performance level can be obtained using the Arrhenius equation for determining life
expectancy [65].
77
___________________________________________________________________________________
The Arrhenius model attributes component failure to changes in a particular electrical

or mechanical property of the component, such as insulation in the case of cables. The
relation (2-25) gives the Arrhenius equation. The Weibull parameter 𝛼 can be defined
by (2-25) - can be used as the quantifiable component (cable) life in years [35, 36].
For this reason, the cable life parameter 𝛼𝑙𝑖𝑓𝑒 will be linked with the Weibull’s
distribution scale parameter 𝛼. The cable life is dependent on many factors, among
which cable loading profile is one of the most important determinants [31, 36, 63].
The IEC cable thermal model described in Section 2.3.1 also is therefore used to
determine the time varying temperature 𝜃𝑐 and as published in [36]; this gives the
IEC-Arrhenius–Weibull model:
𝑡
𝐹(𝑡) = 1 − 𝑅(𝑡) = 1 − 𝑒𝑥𝑝 (− ) (3-21)
𝐵
𝐴 ∙ exp( )
𝜃𝑐 (𝐼)
The IEC-Arrhenius-Weibull model (3-21) is used for sampling the up-times of aging
cable within the SMCS procedure. To this end, equations (3-19) and (3-20) are still
applicable, with λ=1/αβ and α=A·exp(B/θc(I)), where θc(I) is conductor temperature –
current relationship obtained from the cable thermal model (2-5) - (2-6). More
specifically, during the SMCS simulations incorporating the IEC-Arrhenius-Weibull
model, the value of 𝜃𝑐 in equation (2-25) is calculated from the IEC cable thermal
model by substituting the cable currents obtained from load flow computations into
the relations (2-5). The resulting value (𝑖. 𝑒. 𝛼 = 𝛼𝑙𝑖𝑓𝑒 ) from equation (2-25) is
plugged into the relations in (3-19) and (3-20) in order to determine the time varying
life values used in establishing the cable state durations. This procedure will ensure
that the cable thermal exposure and operational status are considered in determining
the duration of the components uptime.
3.7 Chapter summary
This chapter provided a review of the available reliability analysis methods that
consider ageing properties of network components. The review highlights a gap in
this area due to tendency to ignore ageing of components in system reliability analysis
and the absence of reliability analysis models that consider physical properties of the
78
___________________________________________________________________________________
network components (especially underground cable) and their dependence on

operating and environmental conditions.
A new IEC-Arrhenius-Weibull model is developed by considering underground cable

physical properties and impact of cable loading on thermal stress within the Weibull
based ageing model. The model also considers the cable time varying thermal
properties, ambient conditions and loading profiles. Besides, variations of the ageing
stress parameter  of the Weibull based model can be used to quantify the level of
ageing stresses in the simulations of the ageing failures. Also, details of the repair
process as deployed in the SMCS procedures are presented in this section
Finally, the proposed IEC-Arrhenius-Weibull model is capable of providing

additional insights for assessing future smart distribution systems during the planning
stage. Importance of the proposed models lies in the ageing network infrastructure and
application of modern solutions to prolong the asset lives. A few algorithmic tools
required for the proposed assessments have been developed; they are presented in the
subsequent chapters.
79
CHAPTER FOUR
CABLE AGEING STUDY
This section provides details of the techniques used to develop the proposed
distribution system planning and operation approach. The entire proposed
methodology consists of several main building blocks, which are for convenience
presented below:
• Line/cable thermal models
• IEC-Arrhenius-Weibull model for ageing cable
• Sequential Monte Carlo simulation (SMCS) model (presented below),
A high level description of the developed SMCS procedure is as follows. System

elements (cables) are classified as either at the ‘useful life’ or ‘wear-out’ stage of their
life cycle. Standard two-state model based on exponential distribution of up and
downtimes is used to model components from the first group. Aging distribution
cables are modelled with the aid of the proposed IEC-Arrhenius-Weibull model for
up-times and exponential model for down-times. Integration of wind-DG units and
dynamic network reconfiguration with the objective to reduce cable ageing are
incorporated into the SMCS procedure. Some of the procedure outputs, such as
customer interruptions (CIs), customer minutes lost (CMLs) and the numbers of
failures are associated with each cable section, because they clearly show each
section’s contribution towards unreliability. The proposed cable ranking based on the
conventional thermal loss-of-life model (equation (2-33)) is also presented. Several
results and discussions are provided at the end of the chapter based on the provided
input data, test systems and the described methodology.
4 Details of SMCS model
Sequential Monte-Carlo simulation (SMCS) technique is used to simulate time

chronological events that characterise power system operation over a stipulated
planning period. The results (in the form of several types of indices, as well as their
probability distributions) of the simulation provide an insight on the expected
performance levels of the investigated system. However, in order to distinguish
80
Chapter five: Investment planning
___________________________________________________________________________________
between ageing and non-ageing cable and to determine cable ageing parameters (i.e.
Arrhenius constants), it was necessary to simulate both historic and planning period
using ‘similar’ SMCS procedures. All components are considered to be at the ‘useful
cycle’ of their lifetimes in the historic period, so that random sampling is drown from
exponentially distributed uptimes and downtimes [2, 3]. Following the completion of
simulation of the historic period, cable sections are classified as ageing and non-
ageing and the proposed IEC-Arrhenius-Weibull model is used in the second SMCS
run to simulate planning period.
Brief descriptions of all major SMCS building blocks and developed algorithms are
presented below.
4.1 Sampling time to next transition
Conventional SMCS approaches assume that durations of the components up and

down states follow an exponential probability distribution function. On the other
hand, reliability model of ageing cable(s) is based on the developed IEC-Arrhenius-
Weibull model (Section 3.6) which is applied to component’s up-time; exponential
model is still used for the down-time.
The overall system chronological state transition process is a combination of the

individual components state transition processes. Once the individual transition
processes are determined, analysis of system behaviour over hourly periods can be
performed.
4.1.1 Transition time for non-ageing elements
The time to failure (TTF) - uptime is the period during which the component remains
in operation. Whereas, the time during which the element is inoperative is restoration
period and it is called time to repair or replace (TTR) - downtime. The transition
process from up states to down states defines the component’s failure process. The
transition process can be caused by the failure of the considered element, or by the
removal of the element for maintenance, or by completion of the repair/maintenance.
Further details of the repair/maintenance process as adopted in this thesis are provided
in Section 3.6.
81
___________________________________________________________________________________
In case of analysis of systems comprising of non-ageing elements, the time to failure

is assumed to follow an exponential distribution function. An exponentially
distributed TTF(R) has the follow function:
𝑓(𝑡) = 𝜆𝑒 −𝜆𝑡 (4-1)
where 𝜆 is the mean value of the exponential probability distribution and t is either
TTF or TTR. Hence, its cumulative probability distribution function (i.e. failure
function) is obtained from relation (4-2).
𝐹(𝑡) = 1 − 𝑒 −𝜆𝑡 (4-2)
Using the inverse transform method, the randomly sampled TTF is given by (4-3).
1
𝑇𝑇𝐹 = − ln(1 − 𝑈) (4-3)
𝜆
where 𝑈 is a randomly generated number from the uniform distribution (0-1] applied
to exponential cumulative distribution function (6-2). Since 1 − 𝑈 is also distributed
uniformly in the same range as 𝑈, the latter can be used instead; the expression (4-3)
can be simplified into (4-4).
1
𝑇𝑇𝐹 = − ln(𝑈) (4-4)
𝜆
In the SMCS process, if the present state of the element is the up state, 𝜆 represents
the failure rate of the component. Conversely, if the present state of the element is the
down state, 𝜆 is replaced with the repair rate 𝜇 of the component as given by (4-5).
ln(𝑈)
𝑇𝑇𝑅 = − (4-5)
𝜇
4.1.2 Transition time for ageing elements (cables model)
Ageing network cables are modelled with the aid of Weibull distribution for uptimes
and exponential model for down-times. Scale parameter in the Weibull distribution
represents life expectancy of the considered component and is defined using the
Arrhenius model. Dominant element of the Arrhenius model is the conductor
82
___________________________________________________________________________________
temperature which is further linked to the cable’s physical properties and loading in
accordance with the cable thermal model. This approach led to the proposed ‘IEC-
Arrhenius-Weibull’ model for modelling up-times of aging distribution cable (see
equations (3-19) to (3-20)) in which parameters of the Arrhenius model are estimated
using a novel technique that considers historic cable loading (see Section 2.5).
4.2 Coefficient of variation (CoV)
Distribution system reliability is evaluated using reliability indices calculated from the
basic load point indices. They are weighted averages of the individual load point
indices. Statistical distributions of the system indices can also be obtained from the
periodic load point indices if the SMCS is repeated many times for the studied period
(typically, several thousand times). According to the standard given by IEEE
reliability task force [88], system load is modelled by hourly load duration curve.
Hence, one full sample of SMCS simulations consists of planning period in year times
8760 hours, and this is repeated until convergence criterion is satisfied.
In power system reliability analysis, different reliability indices have different

convergence speeds. Empirical studies have shown that the expected energy not
served (EENS) index requires the longest convergence time and it is therefore used to
determine simulation stopping rule to ensure convergence of all other indices. The
permissible tolerance on CoV is usually set in the range of 1%, because the recorded
variation of reliability index EENS can be achieved in reasonable time. Typically, in
order to satisfy a prescribed level of CoV, it is necessary to update the estimated mean
𝐸̂ (𝐹) and estimated standard deviation 𝑆𝐷 (𝐸̂ (𝐹)) of the expected value of the
system reliability indices [2]. The relations for determining these expected value and
standard deviation are shown in (4-12) and (4-7).
𝑁
1
𝐸̂ (𝐹) = ∑ 𝐹(𝑥 𝑖 )
𝑁 (4-6)
𝑖=1
(𝐸(𝐹)𝑖 − 𝐸̂ (𝐹))2
𝑆𝐷(𝐸̂ (𝐹)) = √ (4-7)
𝑁−1
where, 𝑁 represents the total number of simulated hours; 𝐹(𝑥 𝑖 ) is the system
83
___________________________________________________________________________________
reliability index at 𝑖𝑡ℎ simulation hour; 𝐸̂ (𝐹) is the estimated mean value of the
reliability index; 𝑆𝐷(𝐸̂ (𝐹)) is the estimated standard deviation of the estimate 𝐸̂ (𝐹);
and 𝐸(𝐹)𝑖 is the expected reliability index value at the 𝑖 𝑡ℎ hour. If the stopping
criterion is met the procedure is stopped, otherwise, it is returned to the first SMCS
simulation step. CoV measures the estimated uncertainty:
𝑆𝐷(𝐸̂ (𝐹))
𝐶𝑜𝑉 = (4-8)
𝐸̂ (𝐹)
Alternatively, the required number of samples 𝑁 can also be applied as the stopping
criteria [2]. The higher the value of 𝑁, the lower the coefficient of variation will be,
hence, an improved accuracy of the Monte Carlo simulation.
4.3 Simulation methodology
Previous sections of this chapter have covered the basics of the sequential Monte
Carlo simulation procedure. The next sections provide the algorithms steps of the
developed SMCS procedures.
4.3.1 Conventional SMCS procedure with non-ageing components
The conventional SMCS procedure is shown by the algorithmic flowchart Figure 4-1.
This model is applicable to the analysis of systems without ageing elements. The
procedure was used to simulate historic period, and the results so generated are the
inputs to the SMCS procedure that simulates planning period (typically, next price
control period). Figure 4-1 consists of 5 main steps that are described below. These
steps are repeated in an iterative way until the convergence criterion on the (CoV) of
the EENS is satisfied.
84
___________________________________________________________________________________
Define network, load properties cable data &

Installation properties
First hour of
the simulation Set the innitial line thermal ratings based on θs
block
Randomly select component uptimes and
downtimes from exponential distribution and find
the component with the minimum uptime
Next hour when the
Run AC Load flow with SMCS using the hourly
network does not
load curve change
Calculate conductor temperature and resistance θc &

Rθc using loading from the current hour
θc
NO converged ?
Step 2
Apply new time varying cable ratings to find load

curtailments in all branches
Compute reliability indices: SAIDI, SAIFI, ENS.
Hour with no
NO
network change
over ?
Next hour within End of the

the simulation NO simulated 5 years Step 3
block ?
SMCS
NO
converged?
Step 4
Compute cable thermal loss-of-life and final indices
Figure 4-1: SMCS model with non-ageing elements
 Step 1; Initialization:
i. Define the network topology, dynamic (hourly) load profile, cable design
specifications, and installation conditions.
ii. Set all components initially in the up state. The number of load loss events 𝑑
and energy not supplied 𝐸𝑁𝑆 is zero (i.e. 𝑑 = 0, 𝐸𝑁𝑆 = 0).
iii. Assume the initial cable conductor temperature to be
𝜃𝑐𝑖𝑛𝑖𝑡𝑖𝑎𝑙 = 𝑆𝑜𝑖𝑙 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 (𝑖. 𝑒. 𝜃𝑠 ) from the soil ambient temperature, use
85
___________________________________________________________________________________
it to correct the line resistances, and compute the time varying thermal
ratings. Update and adjust the subsequent hourly conductor resistances to
reflect the conductor temperatures based on the method in section 2.3.1.
 Step 2; Generation of components time to failure (TTF):
i. All components times to failure (𝑇𝑇𝐹) are determined by applying the

uniform distribution to the exponential cumulative probability distribution
function based on the method described in section 4.1.
ii. Specify the component with the lowest 𝑇𝑇𝐹 and use it to mark the beginning
of its repair period (i.e. to find TTR); also compute the residual times of the
other components based on it. The TTR of the components is derived based
on the method described in section 4.1.
iii. Run AC load flow to determine the line currents based on the hourly load
profile. Then use the calculated cable currents to evaluate the new cable
conductor temperatures (𝜃𝑐𝑛𝑒𝑤_𝑖 ) from the thermal rating equations (2-5).
Repeat until the temperature difference between two consecutive iterations is
≤1C; the converged temperature in the considered hour is found.
iv. Aggregation of component states:
During the interval until the closest 𝑇𝑇𝐹, all components are in-service and the
following steps are executed:
a. The bus loads for each hour are updated to the current simulation hour,
b. New time varying cable thermal ratings are computed using the converged
conductor temperatures and they will be used to compute the required load
curtailments.
c. If there is no line failure(s), the network does not change and the
simulation process proceeds to the next hour within the simulation block
(block of 5 years) otherwise;
o A restoration action is invoked to close normally open (NO) switches

of the nearest suitable tie-line(s). The action is applied in order to
reconfigure the network and supply disconnected customers pending
86
___________________________________________________________________________________
the repair of the faulted section and possible disconnection of some

customers.
o Next, randomly select components uptimes and downtimes for the
remaining simulation hours within the block (i.e. repeat step 2).
 Step 3; Data recording:
At the end of the load flow analysis, frequency and duration of losses (i.e.
curtailments) of load and energy are recorded, and added on to the previously
obtained results for each simulation block as follows:
i. Increment the quantities of load loss frequency 𝑑 and energy not supplied
𝐸𝑁𝑆.
ii. Compute reliability indices SAIDI (CML), SAIFI (CI), and ENS as shown by
relations (4-9) to (4-11). The ENS index accounts for all cases of energy
curtailed due to thermal and voltage constraints in the planning period. In this
thesis, all indices are computed based on 5-year planning period.
∑ 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐶𝑢𝑠𝑡𝑜𝑚𝑒𝑟 𝑖𝑛𝑡𝑒𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑠

𝑆𝐴𝐼𝐹𝐼 = (4-9)
𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐶𝑢𝑠𝑡𝑜𝑚𝑒𝑟𝑠
∑ 𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝐶𝑢𝑠𝑡𝑜𝑚𝑒𝑟 𝑖𝑛𝑡𝑒𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑠
𝑆𝐴𝐼𝐷𝐼 = (4-10)
𝑡𝑜𝑡𝑎𝑙 𝑒𝑛𝑒𝑟𝑔𝑦 𝑐𝑢𝑟𝑡𝑎𝑖𝑙𝑒𝑑
𝐸𝑁𝑆 = (4-11)
 Step 4; Repeated simulations:
Steps 1 to Step 3 are repeated until convergence criterion on CoV of the EENS is
satisfied. The cables cumulative thermal loss-of-lives are computed and the
average values of the recorded reliability indices from the simulated blocks of 5
years are reported as the system indices at the end of the simulations.
Finally, it is important to note that some improvements to the conventional SMCS

method described in [2] have been included in the method described in Figure 4-1,
more specifically:
(a) Computation of conductor temperature dependent resistances,
87
___________________________________________________________________________________
(b) Consideration of the cables physical properties (i.e. cable thermal

properties),
(c) Computation and setting of time varying thermal ratings.
(d) Computation of cumulative thermal loss-of-life of cable.
These additions were introduced to improve the accuracy of the reliability

assessments by considering the physical properties of components and for comparison
with proposed methods.
Application of the conventional SMCS procedure to the simulation of the historic

period is briefly explained below. In this case, some of the output results are also used
as inputs into the developed SMCS procedures applied for simulation of the planning
period.
4.3.2 Simulation of the Historic Period
Probabilistic simulation of the network operation within the historic period is done
with the aid of a SMCS procedure that takes into account various uncertainties that
can occur with time. The chronological phenomena relevant to this thesis are load
curtailments, wind generations, real-time thermal ratings and cumulative thermal loss-
of-lives. The simulation of the historic period follows the mean of the cable life
estimate which was found to be 51years.
The main features of the conventional SMCS procedure are:
 All distribution network components and traditional generation units

are at the ‘useful life’ stage of the life cycle, so that standard two-
state Markov model based on exponential distribution of up- and
down times is applied [2].
 Arrhenius parameters are determined using the following proposed
approach in Section 2.5.
 Relative hourly profiles are used to determine nodal loads in the
studied time step and as published in [36]. Load varies in a window
around this profile and random sampling is applied to get the nodal
relative temporal values. These values are then multiplied by historic
peak values to get absolute loads.
88
___________________________________________________________________________________
 Real time thermal ratings are modelled using the IEC cable thermal
model [36, 44].
 Modelling of wind generation is described in Section 6.2.2.
4.3.3 Proposed SMCS procedure with ageing cable
Prioritization of cable sections for replacement is based on the extensive probabilistic

analysis of the considered network in the envisaged planning period. Sequential load
and demand profiles, network characteristics including available controls (e.g.
switching devices), cable physical and installation properties, as well as thermal and
aging models of cable are built into the developed SMCS procedures. The main
building blocks of the developed methodology are:
 Simulate historic loading of cable and estimate parameters of the Arrhenius model
using iterative numerical procedure (see Section 2.5). Also classify cable as either
non-aging or aging using the thermal loss-of-life criterion. This is done using the
conventional SMCS procedure as explained in the previous section.
 Initiate the SMCS procedure for the simulation of the planning period and
randomly select component states. Loads and generations are sampled within a
window around the specified annual load/ generation profiles. Exponential
distribution is used to generate up- and down-times of non-aging network
components and down-times of aging components. Up-times of aging components
are determined from the proposed ‘IEC-Arrhenius-Weibull’ model – Section 3.6.
 Calculate CIs, CMLs, energy not served (ENS) and thermal loss-of-life in all
system states. Allocate reliability indices to the individual cable section(s). Find
thermal loss-of-lives of all other cable sections caused by an individual faulted
section. Create cable ranking lists using the two proposed approaches and the total
thermal loss-of-life.
The developed SMCS procedure is supposed to cover a planning interval which is

equal to the regulatory period – typically 5 to 8 years. In that respect, one full
simulation will consider 5(8)*8,760 hours and it will be repeated several thousand
times until convergence is achieved. Coefficient of variation of the expected energy
not served (ENS) was used as the stopping rule. The global flow chart of the entire
89
___________________________________________________________________________________
probabilistic methodology that consists of two SMCS procedures is shown in Figure

4-2; its main steps are summarized below.
Step 1: Input the network data including historic and forecast load profiles; envisaged
new demand/generation connections; existing and future network controllers and
control concepts; input cable data including physical and construction properties - soil
temperature, thermal resistivity, historic failure data, etc.; input parameters of the
load-flow.
Input network data, demand and generation profiles,

cable physical & installation properties, failure data
Iterative calculation of Arrhenius parameters in the three loops:

1. Outer: Over discrete life times
2. Mid: Over numerical solutions of B
3. Inner: over historic hourly intervals Step 2
Calculation of cable thermal loss-of-life in the historic period and classification of

cables as non-aging or aging
Step 3 Set the initial hourly period of the SMCS
Randomly select component up-times and down-times using:

1. IEC-Arrhenius-Weibull model for up-times of aging cables
2. Exponential model for down-times and up-times of non-aging cables
Step 4
Random selection of loads & generations from windows around profiles
Run AC load flow: Record data for rel. indices and thermal-loss-of-life
Next hour when Next hour when the
network does network does not
change change
Next hourly interval is the end of simulation YES
period ?
NO
Next hour is with

YES network change ? NO
Step 5
The SMCS has

NO
converged?
YES Step 6
Compute final reliability indices & ranking lists
Figure 4-2: Flow chart of the SMCS for cable ageing analysis
Step 2: Consider the historic period and calculate Arrhenius parameters A and B in
accordance with the methodology described in Section 2.5. Calculate the cable
90
___________________________________________________________________________________
relative thermal loss-of-lives – see Section 2.6. Upon the convergence of the
procedure, classify individual cable sections as either non-ageing or ageing. An
empirical threshold of 0.875 has been used to classify the cables as ageing or non-
ageing. This choice was based on the relative performance of the network cables and
the assumption that any cable that lost at least 35 years of its 40-year thermal life
should be closely monitored.
Step 3: Initiate the SMCS procedure over the envisaged planning period (typically 5
years, with hourly granularity).
Step 4: Analyse the current hourly period with the aid of the load-flow. Loads and
generations are randomly generated from a window around the specified profile
values. Where appropriate, change the status of network component(s) using the data
from the previous sampling of these components; then sample the component(s) from
the IEC-Arrhenius-Weibull model if the ageing component has just been repaired, or
the exponential model in all other cases.
Step 5: Update the data for the calculation of reliability indices SAIFI (or CIs),
SAIDI (or CMLs) and expected energy not supplied (EENS). Allocate the reliability
indices to the cable section which had caused supply disruption:
 In case of total loss of continuity, all CIs, CMLs and EENS are
attributed to the faulty cable section.
 In case of the partial loss of continuity, a proportion of the
load/customers is shed and allocated to the cable section.
Step 6: After having completed simulations over the planning period, steps 4, 5 and 6
are repeated until convergence of the SMCS procedure is reached. Final reliability
indices and cable ranking lists using the thermal loss-of-life and the proposed methods
are then calculated.
4.4 Test systems
This section provides details of the network input data, simulation results, as well as
analyses of results of the distribution network reliability assessment and replacement
planning using the three SMCS methodologies described above. The section also
91
___________________________________________________________________________________
provides discussions on the sensitivity studies done to find the impact of some critical
parameters on the study results; for example, impact of different operation schemes,
impact of varying the shape parameter  in the Weibull distribution, etc. In the studies
conducted in this thesis, focus was only the effects of ageing UGCs for reasons of
simplification. Advanced studies could assess the impacts of ageing in all system
components altogether during the planning period.
Developed test systems and input data are presented first, which is followed by
presentation and analysis of results including sensitivity studies.
Developed test systems
The network schematic diagrams, network data, and component details required for
the developed methodologies are presented in this section.
4.4.1 IEEE 33 bus single feeder test system
A single feeder IEEE 33 bus test network shown in Figure 4-3 is used as the test
network for all studies [89]. This medium voltage (MV) test system consists of 33
buses interconnected by 32 lines and 5 normally open tie-lines, making 37 lines in
total; in-service circuits are represented with full lines, whilst dashed lines are used
for normally open sections. All studies are based on the network nominal voltage of
22kV and 100MVA base apparent power. The peak load demand of the distribution
network is 3.715MW and 2.3MVar and the (equivalent) utility generator connected to
Bus 1 supplies it (i.e. Bus 1 is a slack node).
92
___________________________________________________________________________________
Substation represents Bus

L: represents cable lines
L1 Tie-lines: represented by dashed
L18 (100,60)
lines
(90,40) L2
(90,40) L22
L3 (90,50)
(120,80)
L4 L23
(60,30) (420,200)
L5 L24
L19 (60,20)
L33 (420,200)
(90.40) L6
L20 (200,100)
(90,40) L7
(200,100)
L25
L8
L21 (60,20)
(90,40) L9 L37
(60,20) (60,25)
L10 L26
(45,30) (60,25)
L35 L11
L27
(60,35)
(60,20)
L12 L28
(60,35) L34
L13 (120,70)
(120,80) L29
L14 (200,600)
(60,10) L30
L15 (150,70)
(60,20) L31
L16 (210,100)
(90,20) L32
L17 (60,40)
(90,40)
L36
Figure 4-3: IEEE 33 bus single feeder test system
Network data that describe the distribution network in Appendix - A 1 consist of line
resistances, reactance, and the bus load demands, as shown in Appendix A 1. The 33
bus network data are provided in 38 rows and 6 columns; fields in the first row give
the table headings while rows 2 to 38 contain line identifiers, resistances (),
reactance (), active power (kW) and reactive power demands (kVAR) at the end of
the lines. However, the network data provided in Appendix A 1 do not contain any
further details such as line types, design and installation characteristics, etc. A few
assumptions were made with regards to missing network data.
It was assumed that the 33 buses are interconnected by 32 underground cables and 5
normally open tie-lines, making 37 cable circuits in total. It was also assumed that the
cable circuits are switched via sectionalizing switches connected at both ends of each
cable section. The conventional IEEE RTS-96 hourly chronological load profile was
93
___________________________________________________________________________________
used as the annual relative loading profile at all nodes [88]. The network underground
cable (UGC) sections were assumed 1km in length, which were deemed appropriate
for urban areas for each of the 37 circuits.
Reliability data
Failure rates and mean repair times are basic inputs required for SMCS analysis. For
reasons of simplification the three main distribution network components are
considered in this thesis for components in their useful life period. Other network
components such as distribution transformers, circuit breakers etc. are considered to
be 100 % reliable. The Table 4-1 provides the details of the reliability data for the
main network components applied to the test systems for use in the subsequent
studies. In addition, the data provided
Table 4-1: Reliability data for main network components [90]
Power component Failure rates (failure/year) Mean time to repair, h

buses 0.08 140
cables 0.034/km 201
sectionalizing switches 0.004 140
The next section presents the main features of the second test network used in the
studies.
4.4.2 The IEEE 69 bus one feeder test system
The IEEE 69 node 22kV radially operated test network is shown in Appendix - A 2.
The normally open tie-lines are marked with dashed lines [36, 91]. The peak demand
is 3.34MW and 2.77MVar in the base year. The IEEE 69 bus network data are
presented in Appendix A 2 in 73 rows and 7 columns; all fields in row 1 give the table
headings, while rows 2 to 73 contain line identifiers, resistances (), reactance (),
active power (kW) and reactive power demands (kVAR) connected at the line end
terminals. Line sections are labelled as L1, L2 … L74 in Figure 4-4. All loads of
distribution substations are represented as (kW, kVAr) quantities next to each node.
94
___________________________________________________________________________________
Substation
represents Bus
L1 Cable line
Tie-line
L27 L2
L3
L35
L4
(26,18.55)
L46 L36
L5
(26,18.6) (26,18.55)
L28 (2.6,2.2)
L47 L6 L37
(26,18.6) (79,56.4) (40.4,30)
L29 L48 L38
(384.7,274.5) L7 (24,17)
L30 L49 (75,54) L39
(384.7,274.5) L8 (24,17)
L72 (30,22) L50
L31 L9 L40
(40.5,28.3) (1.2,1)
L32 L52 (28,19)
L51 L41
(14,10) L10 (3.6,2.7)
L33 (4.35,3.5) (145,104) L42
(19.5,14) L11 L69 (6,4.3)
L53
L34 (26.4,19) (145,104) L43
(6,4) L54 L12
(24,17.2) (8,5) L44
L65 L13 (39.22,26.3)
L55 L45
(18,13) (8,5.5) (39.22,26.3)
L56 L66 L14
(18,13)
L57 L15 L71
(45.5,30)
L58 L16
(100,72) (60,35)
L59 L67 L17
(28,20)
L68 (60,35)
L60 (28,20)
(1244,888) L18
L70
L61
(32,23)
L62 L19
(1,0.6)
L63 L20
(227,162) (114,81)
L64 L21
(59,42) (5,3.5)
L22
L23
(28,20)
L24
L25
(14,10)
L26
(14,10)
L73
Figure 4-4: IEEE 69 bus single feeder test system
Similarly, to the previous test system, the circuit details were not provided in [89, 91].
It was also assumed that 69 nodes are interconnected by 68 underground cables and 5
normally open tie-lines, making 73 cable sections in total.
The conventional IEEE RTS-96 hourly chronological load profile was also used as the
operational loading profile [88]. All underground cable sections were assumed to be
1km in length for each of 73 branches deemed appropriate for some urban networks.
Each cable section is also assumed to be connected to the corresponding buses at both
its ends via sectionalizing switches.
95
___________________________________________________________________________________
The next section presents the loading profile used for both networks in the studies.
4.4.3 Network load model
A dynamic – chronological hourly load profile over 1-year period is needed for the
developed sequential SMCS procedures. The hourly varying relative load model
widely used in reliability evaluation studies is presented in [88] and it is used in this
study. Appendices: A 3, A 4, and A 5, present the relative hourly, daily, and weekly
peak loads in (%), respectively. Therefore, the periodic (hourly) power demand L(t)
can be determined for any system node using expression (4-12) and the values
provided in A 3, A 4, and A 5.
𝐿(𝑡) = 𝐿𝑃 × 𝑃𝑤 × 𝑃𝑑 × 𝑃ℎ (𝑡) 𝑡 = 1,2, … .8760 (4-12)
where 𝑃ℎ (𝑡) is hourly peak load in percent of daily peak; 𝑃𝑑 is daily peak load in
percent of weekly peak; 𝑃𝑤 is weekly peak load in percent of annual peak; and 𝐿𝑃 is
the annual peak load value. Week 1 starts from the 1st day of the calendar year and
loads at different buses of the test systems are varied around arbitrary values of
±30% in the SMCS process.
Depending on the value of the peak load at each load point (Appendix A 1and A 2),
the annual power consumption profile can be generated with hourly granularity. This
model captures daily, weekly, and seasonal variations of demands at distribution
substations. Week 1 load profile for bus 2 of the 333 bus test system is illustrated by
the load curve in Figure 4-5.
96
___________________________________________________________________________________
Figure 4-5: Typical load profile (week 1) for bus 2 of the 33 bus test system
4.4.4 Cable data and operating conditions
The two network data sources [89, 91] did not provide any further details on cable
design and installation characteristics. In this thesis, the chosen cable type is similar to
the one presented in [92]; Table 4-2 presents the cable details.
Table 4-2: Cable data
S/No. Cable characteristic Value

1 Axial spacing 30mm
2 Cable arrangement Trefoil
3 Conductor CSA 95mm2
4 Conductor diameter 12mm
5 Conductor material Aluminium
6 Conductor outer diameter 64mm
7 Depth of burial (L) 100mm
8 Design temperature 90oC
9 Insulation diameter (excluding screen) 28mm
10 Insulation type XLPE
11 Kp (from Table 2 IEC 60287-1-1:2006) 0.80
12 Ks (from Table 2 IEC 60287-1-1:2006) 1.00
13 Number of phases/cores 3
14 Operating frequency (f) 50 Hertz
15 Operating voltage 22kV
16 Rsc (cable screen resistance) 0.00089808 /m
17 Screen material Copper (25mm2)
18 Soil ambient temperature 3oC -15oC (UK)
19 Soil thermal resistivity 1.2 K m/W (UK)
97
___________________________________________________________________________________
S/No. Cable characteristic Value

20 T1 0.908 K m/W
21 T2 0 K m/W
22 T3 0.0986 K m/W
23 T4 (PVC duct installation) 1.3332 K m/W
24 Wd 0.088987 W/m
25 𝜆1 ratio of losses in metallic sheaths to total conductor losses) 8.12E-05
26 𝜆2 ratio of losses in armour to total conductor losses) 0
27 Cable life time 40 years
Cable properties and data presented in Table 4-2 are required to compute current
ampacities – permissible cable ratings using the IEC thermal model [44]. Another
application is substitution of the IEC thermal model in the IEC-Arrhenius-Weibull
uptime model for cable ageing. The real time thermal ratings so calculated are then
used within power flow studies and reliability assessment models. Cable parameters
presented in Table 4-2 also show that it is possible to study different cable designs
and installation conditions. For example, conductor material type can be copper or
aluminium, conductor cross sectional area and other associated dimensions can be
varied, cable insulation type can be paper, ethylene propylene rubber (EPR), etc., duct
type and its associated installation properties can be modified, and different soil
thermal characteristics can be used.
Cable ambient conditions
During network operation, cable conductor temperature will vary not only due to the
dynamic loading, but also due to time varying soil ambient temperatures and the
seasonal soil thermal resistivity (items 18 and 19 in Table 4-2). A time varying soil
temperature profile, shown in Figure 4-6), was adopted in the sequential hourly
simulations (item No. 18 in Table 4-2). The profile is characteristic for the UK and it
was adopted from [93].
98
___________________________________________________________________________________
Figure 4-6: Time varying soil temperatures used in the SMCS over one year
Figure 4-6 shows that the maximum soil temperatures at 100 cm under surface were
8°C in winter, 27°C in summer, 17°C in spring and autumn, whilst the minimum
temperature was 2°C. Soil thermal resistivity is used as an input for the calculation of
T4 in the IEC cable thermal model – equation ((2-5) and (2-6)). It is dependent upon
the soil/backfill type, moisture content, particle size and distribution, etc. In this
study, ‘typical’ seasonal values of soil thermal resistivity were adopted and are
presented in Table 4-3. Soil thermal conductivity is highest during winter months,
when there is a lot of rainfall. Winter resistivity was 0.9 K m/W (weeks 1-8 & 44-52),
summer resistivity 1.2 K m/W (weeks 18-30), and spring/fall resistivity was 1.05 K
m/W (weeks 9-17 & 31-43), as shown in Table 4-3.
Table 4-3: Time varying soil thermal resistivity
Soil thermal resistivity (K m/W)

Winter Spring Summer Autumn
0.9 1.05 1.2 1.05
The cable characteristics presented in Table 4-2, Table 4-3 and the ambient soil
temperature profile depicted in Figure 4-6 were utilised in the chronological network
analysis involving ageing cables. The developed approach enabled more realistic
modelling of the impact of cable design and environmental conditions on network
operation.
99
___________________________________________________________________________________
The next section presents the study cases and SMCS simulation results based the
network and reliability input data provided in the previous sections.
4.5 Study cases and presentation of results
The two main study cases and analysis of results presented in this section include:
 Impact of ageing distribution cable on system reliability and cable

replacement planning.
 Optimal integration of wind-DG and dynamic network
reconfiguration to reduce cable ageing and defer replacement.
4.5.1 Impact of ageing distribution cable on reliability and cable replacement

planning
This section presents the calculated parameters of the Arrhenius ageing model,
reliability indices for the two test systems with non-ageing and ageing cable, as well
as cable ranking using the thermal loss-of-life and the first proposed cable ranking
criterion. The section also provides results from the sensitivity studies carried out to
find the impact of variation of shape parameter  in Weibull distribution (i.e. Weibull
ageing stress parameter) on the simulation results.
Determination of Arrhenius (A and B) parameters
The estimated Arrhenius model parameters A and B were obtained using the proposed
iterative method within simulations of the historic period (see Section 2.5) for two
cable types, namely, polymeric – XLPE and paper insulated. The results are given in
Table 4-4 and further presented through the normalized Arrhenius thermal life graph
(Figure 4-7), which is expressed in per unit of 40 years (i.e. cable rated life).
Table 4-4: Estimated Arrhenius parameters for different cable types
Cable type Arrhenius parameters

A B
Paper type 0.1536 2020
XLPE 2.4114 1020
100
___________________________________________________________________________________
The assumed ageing cable mean age and standard deviation were, respectively, 51 and
20 years, which is in line with the results of the international study on cable lifetime
data [6].
Figure 4-7: Normalized Arrhenius thermal graph
Figure 4-7 shows a figure containing plots of the ratio of cable life (at service
temperature) to the rated life (at rated temperature) versus service temperature for
XLPE and paper (PILC) cables. As would be expected, Figure 4-7 suggests that the
actual life of the cables could exceed rated life to some degree depending on the
thermal loading cycle. On the other hand, when the cables are overloaded, typically
due to faults when the network reconfiguration and back-feeding are being done,
conductor temperature can exceed 90C which results in the loss of relative life
greater than unity. Besides, it is clear that XLPE cables are more resilient than the
paper cables at higher temperatures. The Arrhenius parameters A and B for XLPE
cable type were used in further studies, as a part of the IEC-Arrhenius-Weibull
reliability model within the SMCS technique. It is important to note that the values of
A and B and subsequent reliability indices depend on the cable age at the beginning of
the planning study.
101
___________________________________________________________________________________
System-wide reliability indices and impact of Weibull shape parameter
Prioritization of ageing underground cables for replacement under financial

constraints was done using the developed SMCS procedure – second SMCS algorithm
in Section 4.3.3. The system-wide indices SAIDI and SAIFI are calculated for the
considered 5-year planning period and they are shown in Table 4-5 (a) - (b). To
quantify the impact of cable ageing and thermal stresses, SMCS studies were run for
different values of the Weibull stress parameter . Base case =1 means that there is
no cable ageing (i.e. Weibull distribution becomes exponential), whilst the highest
=1.8 implies very high ageing stress level. This range was selected in line with the
research carried out in [35, 68]. The repair adjustment factor in the Weibull stochastic
process q is set to 1 in order to account for the full impact of ageing on the repair
process. This means that the cable age following a repair is exactly the same as before
the repair indicating that so-called ‘minimal’ repair has been done (as opposed to
‘general’ repair when 0 < q < 1.0).
Table 4-5: Test systems reliability indices

(a) IEEE 33 bus test system
IEEE 33 bus test system

SAIDI (hours / customer. 5yr)
( 1 1.2 1.4 1.8
3.821 (1 pu) 25.686 (6.7 pu) 153.108 (40.1 pu) 586.6 (153.5 pu)
SAIFI (interruptions / customer. 5yr)
( 1 1.2 1.4 1.8
0.385 (1 pu) 6.576 (6.5 pu) 15.315 (39.7 pu) 58.369 (151.5 pu)
(b) IEEE 69 bus test system
IEEE 69 bus test system

SAIDI (hours / customer. 5yr)
( 1 1.2 1.4
1.025 (1 pu) 41.124 (40.1 pu) 305.957 (298.4 pu)
SAIFI (interruptions / customer. 5yr)
( 1 1.2 1.4
0.094 (1 pu) 4.044 (42.8 pu) 30.509 (322.9 pu)
Table 4-5 (a) – (b) presents the SAIDI and SAIFI results for the two test systems and
several  values; the per unit values are in relation to the non-ageing SMCS results
102
___________________________________________________________________________________
=1. The total number of customers connected to the 33 bus and 69 bus test systems
was, respectively, 1600 and 3400 (50 customers per bus with no distinction of
customer types - for reasons of simplification). It can be deduced that ignoring the
impact of cable ageing can lead to gross underestimation of the system performance in
future. More specifically, as parameter  increases the system reliability worsens
considerably. From the results,  can be interpreted as an indicator of ageing intensity,
component age, and rate of change of failure rate. Therefore, the practical choice of
parameter  should be based on the comparison with the real-life company data on
historic network/circuit performance.
It is expected that the parameter  is likely to be around 1.2; this value was used in all
subsequent studies. In addition, the higher values of reliability indices observed in
Table 4-5 can be also attributed to the assumptions (made on network topology) of
having a 1 kM cable section connected between buses in both test systems. This
simplified assumption meant that it is not possible to repair one cable section while
the other cable sections are unaffected by the faulted section.
Cable ranking for replacement planning (based on Thermal loss of life)
Classification of cables to non-ageing and ageing was done by simulating the historic
period (51 years) and using the thermal loss-of-life criterion. The results of the cables
thermal loss of life in the historic period for the 33 bus and 69 bus test systems are
presented in Table 4-6 and Table 4-7. The simulation of the historic period based on
the method described in Section 4.3.2 resulted in the values shown in columns of both
tables below.
ageing)
Cable ranking list based on the thermal loss of life criterion – 33 bus network
Historic Historic Historic
Line ID Line ID Line ID
period period period
Line 1 0.9308 Line 12 0.8604 Line 23 0.8638
Line 2 0.9296 Line 13 0.8602 Line 24 0.8603
Line 3 0.8868 Line 14 0.8604 Line 25 0.8653
Line 4 0.9257 Line 15 0.8659 Line 26 0.8761
Line 5 0.9316 Line 16 0.8649 Line 27 0.8708
Line 6 0.8720 Line 17 0.8610 Line 28 0.8661
103
___________________________________________________________________________________
Historic Historic Historic
Line ID Line ID Line ID
period period period
Line 7 0.8960 Line 18 0.8616 Line 29 0.8614
Line 8 0.8897 Line 19 0.8603 Line 30 0.8603
Line 9 0.8626 Line 20 0.8605 Line 31 0.8604
Line 10 0.8605 Line 21 0.8604 Line 32 0.8604
Line 11 0.8606 Line 22 0.8658
ageing)
Line ID Historic period Line ID Historic period Line ID Historic period
Line 1 0.8610 Line 25 0.8611 Line 49 0.8627
Line 2 0.8610 Line 26 0.8607 Line 50 0.8604
Line 3 0.8607 Line 27 0.8605 Line 51 0.8609
Line 4 0.8630 Line 28 0.8622 Line 52 0.8618
Line 5 0.8775 Line 29 0.8636 Line 53 0.8617
Line 6 0.8784 Line 30 0.8611 Line 54 0.8619
Line 7 0.8659 Line 31 0.8634 Line 55 0.8618
Line 8 0.8623 Line 32 0.8652 Line 56 0.8690
Line 9 0.8683 Line 33 0.8646 Line 57 0.8646
Line 10 0.8617 Line 34 0.8622 Line 58 0.8619
Line 11 0.8652 Line 35 0.8610 Line 59 0.8696
Line 12 0.8635 Line 36 0.8656 Line 60 0.8726
Line 13 0.8620 Line 37 0.8662 Line 61 0.8603
Line 14 0.8614 Line 38 0.8626 Line 62 0.8603
Line 15 0.8610 Line 39 0.8603 Line 63 0.8605
Line 16 0.8613 Line 40 0.8766 Line 64 0.8621
Line 17 0.8602 Line 41 0.8679 Line 65 0.8603
Line 18 0.8605 Line 42 0.8616 Line 66 0.8602
Line 19 0.8604 Line 43 0.8603 Line 67 0.8605
Line 20 0.8605 Line 44 0.8618 Line 68 0.8602
Line 21 0.8603 Line 45 0.8602
Line 22 0.8607 Line 46 0.8605
Line 23 0.8614 Line 47 0.8671
Line 24 0.8624 Line 48 0.8784
In Table 4-6 and Table 4-7 the critical cable sections are indicated using italicized
fonts and the subsequent discussions will focus on these components. Figure 4-8 (a) –
(b) shows the relative losses of cable lives for both studied systems. An empirical
threshold for loss-of-life 0.875 (=35yr/40yr) has been used to classify the cables as
ageing. This choice was based on the comparison of the relative performance of the
network cables and the assumption that any cable that had lost at least 35 years of its
thermal life requires closer monitoring.
104
___________________________________________________________________________________
Thermal Loss of Life (p.u.)

Thermal Loss of Life (p.u.) 0.935 0.879
0.925 0.879
0.878
0.915 0.878
0.905 0.877
0.877
0.895
0.876
0.885 0.876
0.875 0.875
L5 L1 L2 L4 L7 L8 L3 L26 L6 L48 L5 L40
Aging cables/lines Aging cables/lines
[a] [b]
Figure 4-8: Cable ranking based on thermal loss of life in the historic period:
(a) IEEE 33 Bus test system (b) IEEE 69 Bus test system
Figure 4-8(a) – (b) shows that cables L1-L5, L7, L8 and L26 of the 33 bus test
system, as well as L5, L6, L40 and L48 for the 69 bus network will be classified as
ageing and will require monitoring of their impact on system reliability and
economics. The next chapter provides analysis on the investment planning required in
order to develop an effective replacement strategy for the identified critical cables in
Figure 4-8.
4.6 Chapter summary
This chapter presented the details of the development procedure for the proposed
distribution system planning model. The proposed model is formulated in a manner to
consider the impacts of an ageing underground cable on the reliability of the
distribution network. The standard IEC cable thermal model, Arrhenius equation for
ageing effects, Weibull pdf and SMCs are the methods incorporated into the new
reliability (cable ageing) model – IEC-Arrhenius-Weibull cable reliability model. The
IEC-Arrhenius-Weibull model is used for the calculation of the TTF of cables
classified as ageing in the SMCS process. Details of the input parameters required for
the SMCS process such as details of test systems, load profiles, cable data and
installation properties, were provided in this chapter.
This chapter also provided results on the application of the IEC-Arrhenius-Weibull

cable reliability model in the SMCS assessment of the test period within a 5-year
planning period. The results of the studies included sensitivity studies on the impact
of variation of shape parameter in Weibull distribution (i.e., Weibull ageing stress
parameter - ) on the simulation results. From the results, it was observed that the
105
___________________________________________________________________________________
main factors affecting the frequency of ageing failures include , loading cycle, and
the network topology during contingency.
To this end, for more accuracy, subsequent studies on future networks would require a
robust planning technique that considers not only the network component
performance but also on operational economics, system operations (such as network
reconfiguration etc.) along with impending connections of new equipment (such as
LCTs). The next chapter provides analysis on the investment planning required in
order to develop an effective replacement strategy for the identified critical cables.
106
CHAPTER FIVE
INVESTMENT PLANNING
This section presents an overview of investment planning for power distribution

networks with the focus on asset reinforcement and replacement planning. The section
also discusses the importance of system performance indices in the development of
future distribution networks. More specifically, details of a new asset (cables) ranking
models for replacement planning based on the developed system financial indices are
presented. The new cable ranking models make it easier to locate areas (cable
sections) where reliability needs to be improved.
5 Distribution system investment planning
Distribution network operators (DNOs) normally manage large and evolving

electricity distribution networks and they are asset rich companies. In general, power
system planning requires identification of best new equipment, along with finding its
location, time of construction and operating schedule [24]. Distribution system
development plans have high investment costs. Therefore, DNOs are constantly faced
with the challenge to decide whether to retain existing assets or
retire/replace/reinforce them with newer ones that will better and more economically
serve current and future requirements. The frequency of making these decisions is
dependent on several factors such as the rate of load growth, technology
advancements and trends in business dynamics including regulatory regimes [94].
Unfortunately, the investment decisions are often financially constrained because
either the regulatory or the DNO internal business plans determine the available
funds.
Therefore, a common goal of many planning studies performed within utilities is to

decide whether to replace existing assets, or for how long in future to retain the
currently installed assets. Addition of new assets ‘in parallel’ with existing ones is
called reinforcement and it is done whenever the planning standards are violated. The
process of obtaining an optimal planning solution can be influenced by numerous
factors, which can be classified as technical, social, political, and economic
considerations etc. However, system planners consider technical and economic
108
___________________________________________________________________________________
aspects as the dominant factors and a basis upon which most of the issues can be
resolved.
Since ageing is unavoidable and irreversible in electrical components, some measures

can be taken nonetheless to mitigate its effects, slow it down, or merely predict its
effects more precisely. Mitigating the effects of aging can reduce the consequences of
the failures. Slowing down the deterioration extends the lifetime and reduces the
probability of outage due to deterioration. Besides, predicting deterioration failures
more accurately can help to manage impending failures more effectively.
The main types of investment undertaken by utilities to comply with planning

standards, control equipment failures, and quality-of-supply (QoS) performance
include reinforcement, replacement, refurbishment, and maintenance. Reinforcement
and replacement are the most common investment actions applicable to the
underground cables. In this thesis, reinforcement is used in planning the optimal
connection and management of LCTs in distribution network while a ranking list for
cable replacement is generated from the reliability assessment studies of the planning
period.
The next section summarizes the major steps deployed by utilities to satisfy planning
standards and reduce the likelihood of end-of-life failures of ageing components.
5.1 Reinforcement planning (load related planning)
In general, reinforcement planning is necessary when deciding to install new

equipment in order to increase the capability or capacity of existing (retained)
equipment.
Within the medium voltage section of the distribution system, utilities carry out
network reinforcement planning in three main stages. The first planning task involves
‘general reinforcement’ in which changes in general load and overall compliance with
the national planning standards are studied. This activity is done typically on a yearly
basis and known replacement interventions are taken into account. Studies of
connection applications for both generation and load are considered side by side on a
daily basis. The second major planning stage is the quality of supply (QOS) studies
109
___________________________________________________________________________________
triggered by fault statistics (observed poor performance) or regulatory incentives. This

planning study aims to determine construction activities such as new switchgear
installation, automation of existing switchgear, lying of short cable links, etc. The
final stage involves optimal network reconfiguration to minimize losses and improve
reliability.
5.2 Replacement planning
Replacement of an older unit with a new unit is an effective way of improving the
system reliability by way of reducing the number of failures and associated
downtimes. Replacement can be classified as replacement on failure or preventive
replacement. The high costs of preventive replacement may be justified when the
consequences of failures are extremely expensive.
A replacement study is done to decide whether to substitute or retain assets based on

technical/economic reasons. The need for replacement could be due to several factors
such as physical impairment, inadequacy, reduced performance, altered requirements,
obsolescence, anticipation of future requirements, risk of unplanned replacement or
catastrophic failure etc.
In the analysis conducted in this thesis, cable (component) replacement is carried out
when it is estimated that the value of the cable thermal loss-of-life is  1 in the
planning period. Whilst reinforcement is considered when an increase capacity for
power flows is required due to impacts of a planning decision (such as chosen wind-
DG connection point and/or smart solution) on the distribution network. Furthermore,
reinforcements are modelled as branch capacity (in MVA) that can be added in
parallel to an existing section (to accommodate increasing power flows) while
replacements are deemed as a new cable with equivalent capacity installed in place of
a cable that has its thermal-loss-of-life value equal to 1.
5.2.1 Underground cables replacement planning
The continuity of energy supply is significantly affected by the asset and network
characteristics. The underground cables (UGC) are a very important component in the
distribution networks based on outreach, cost and impact on SAIDI (especially at the
110
___________________________________________________________________________________
MV level) [1, 4, 6, 39]. In Europe, the average age of the UGC for the MV networks
is 51 years and a small group of cable is more than 60 years old [1, 6, 39]. The main
criterion used to decide on any replacement is either history of service failures or
physical degradation (e.g. detection of partial discharges, growth of electrical trees
etc.). To this end, it would be very useful to determine the remaining life of the
underground cables (UGC), which is most often affected by the following types of
failures [31, 39]:
 Manufacturing defects transforming into permanent failures;

 Corrosion of the metallic parts resulting into water ingress;
 Partial discharges within the insulation; and
 Thermal runaway failures due to hot spot temperatures.
The severity of the above listed failure modes depends on the physical, constructional,
and operational characteristics of the cable component and the failures could occur at
any section of the UGC network. In any case, the expected lifetimes of the cables and
any other critical components is inextricably linked with their loading patterns and
voltage surges - transients.
In this thesis, focus is on decision making related to replacement planning of ageing

UGC, which can experience end-of-life failures. The developed UGC end-of-life
failure model (i.e. IEC-Arrhenius-Weibull model) forms a very important part in the
developed SMCS procedure.
5.3 Regulatory investment constraints
The power industry has recognized that equipment ageing is a significant problem and
thus equipment lifetimes must be effectively managed [24, 25]. However, available
financing budgets for effective management of assets is a challenge considering the
present economic trends [24, 25]. Furthermore, some national electricity regulators
(e.g. Ofgem in the UK) set a limit on each DNO’s capital income to ensure that their
customers are not subjected to high charges (overcharged) for use of the network
facilities. Even though each DNO owns the distribution network infrastructure in its
particular region, the regulator still has to take a view on how much investment is
111
___________________________________________________________________________________
needed to maintain and develop the system assets in order to prevent over-investment,
which would reflect negatively on customer bills.
In the UK, following the privatisation of the power system, an asset volume, and unit
cost based regulation concept was applied to control DNO’s revenues by setting up
limits on each utility’s revenue [95]. This limitation has exposed the DNO’s to a
broad range of conflicting objectives such as: improve reliability and quality of
service, reduce operating costs, offer services at lower cost, and reduce capital
expenditure. This set of objectives has led to the development of construction and
operation schemes that tend to maximize utilization of existing assets and often defer
capital investments. Consequently, the replacement schemes for distribution circuits
are driven by investment costs and available (squeezed) investment budgets.
5.4 Ranking for cable replacement
The proposed cable-ranking criteria consider company financial performance in the

form of quality-of-supply costs, operational costs linked to repair, as well as cost of
‘cable ageing’. The ranking criteria take some of the reliability indices as input
parameters.
5.4.1 System performance indices
System performance indices can be used for various purposes in the distribution
system business. There are several different types of indices that are used to assess
network performance, efficiency of distribution business investment and operation,
etc. One of possible categorisations of system performance indices can be as follows:
 Basic component/point and system-wide reliability indices

 Momentary reliability indices
 Cost benefit indices
 Power quality indices
The basic component or point reliability indices are related to specific assets or points
of supply in the network. They are, for example, annual failure rate, average outage
duration, annual failure frequency, etc. On the other hand, basic system-wide
reliability indices are an aggregation of the basic component-point indices and they
112
___________________________________________________________________________________
reflect the overall system performance. Some of these indices are used by the
regulator to assess DNO performance, and they are often called quality-of-supply
metrics.
Momentary reliability indices are seldom used because they represent momentary
interruption effects which are currently not considered by the regulator. However,
these indices have to be counted and reported to the regulator which indicates their
possible inclusion in the forthcoming price control mechanisms.
The cost benefit indices provide cost benefit assessment of some activities on the
studied system. The investment indices can be used to compare investment
alternatives when new assets are deployed in a network, or to assess a retro-fit
investment programme on the existing network. Other examples are comparison of
quality-of-supply investments with the cost of reliability indices as defined by the
regulator, cost of maintenance programme versus cost of faults etc.
The power quality indices measure the quality of power served to customers. These
indices are gaining increasing attention by system planners due to the sensitivity of
modern digital electronic devices to voltage/current wave distortions. In connections
planning, each development plan has to be assessed against power quality standards
(i.e. levels of harmonic injections) and appropriate filters installed where necessary to
bring distortion levels under threshold values.
Distribution systems reliability performance can be defined using a number of

reliability indicators, as presented in [3, 16]. The UK regulator, Ofgem, has selected
two widely used indicators, SAIFI (System Average Interruption Frequency Index)
and SAIDI (System Average Interruption Frequency Index) [1, 2] as the basic
quantities in the Interruption and Incentive Scheme (IIS) that defines reward/penalty
regime for DNOs. SAIFI measures the number of sustained interruptions an average
customer will experience over a specified period (typically one year) and when it is
computed per 100 customers, it is usually called customer interruptions (CIs). SAIFI
can be improved by reducing the number of faults, e.g. installing more on-network
circuit breakers or parallel paths so that fewer customers are affected by faults. SAIDI
provides a measure of the duration of interruptions an average customer experiences
over a specified period (typically one year); and when it is computed per 100
113
___________________________________________________________________________________
customers, it is usually called customer-minutes-lost (CMLs). It can be improved by

reducing the fault interruption durations, for example by deploying a faster post fault
system reconfiguration to restore services to customers more rapidly. Therefore, an
improvement in SAIFI will result in the corresponding improvement in SAIDI, but
not necessarily vice-versa. Hence, both SAIDI and SAIFI are considered in this study
in conjunction with incentive rates. Energy not served (ENS) is energy (kWh) lost by
an average customer during the planning period. In this thesis, a 5yr planning period
is used. The next section further shows why SAIDI and SAIFI are important to DNOs.
5.4.2 Utilities performance drivers
The vast majority of DNO revenues come from the load and generation customers in
the form of connection and distribution-use-of-system charges. However, a part of the
revenue is exposed to risk because it is driven by incentive regimes that are
established to either reward or penalize a utility. The quality-of-supply incentive
scheme (called Interruption Incentive Scheme – IIS) is one of the most important
because it sets benchmarks for the quality of service delivery. As already mentioned,
the indices explained above are used in the IIS scheme in conjunction with incentive
rates to calculate a DNO quality-of-supply reward (or penalty). At present, DNOs
compare the actual interruption frequency and duration indices with the regulatory
targets - benchmarks. The UK regulator currently makes use of the three major types
of metrics:
 Reliability indices targets,

 Performance based incentive rates, and
 Guaranteed standard of performance.
In some countries and parts of the United States, regulators provide reliability index
targets founded on historical index values and desired future trends. The targets are
typically geared towards maintaining or improving SAIFI and SAIDI indices after
certain periods in time (e.g. several years). This method may not appeal to all the
different classes of customers, since each might desire a different reliability level
based on their ‘importance’ (e.g. hospitals, schools, precision manufacturing, etc.).
114
___________________________________________________________________________________
Reliability index targets (benchmarks) projected in future are also applied in the UK
combined with performance-based rates (PBRs) which are main statutory regulations
for rewarding or penalizing distribution utilities for good or poor performance,
respectively. The performance assessment is based on customer interruptions (CIs)
and customer minutes lost (CMLs) within a certain period (typically a year). Given
the extrapolated targets for reliability indices and known PBRs within a price control
period, utilities try to optimise trade-offs between system reliability reward and
investment cost. More specifically, they compare the pounds worth earned or saved
(from penalties) due to improved reliability to that incurred by additional investments
[4, 84]. On the other hand, since system reliability is affected by many factors such as
equipment type and age, and changing operational practices, the PBRs have to be set
by the regulator following an extensive economic analysis that includes both
willingness to pay and willingness to be paid [4]. This is necessary for the successful
implementation of the incentive regime.
Guaranteed standard of performance is another type of performance drivers imposed

by the regulator and applied by utilities. Guaranteed standard of performance in the
UK are in the form of ˈGuaranteed Standardsˈ and ˈWorst Served Customersˈ. In the
UK, the Worst Served Customers are those who have experienced an average of 3
incidents of higher voltage interruptions per year or 5 over a three year period. An
additional incentive is set aside for these customers by the regulator. On the other
hand, the Guaranteed Standards are set by the regulator and backed by a guarantee –
customers receive a payment, either via the electricity supplier or directly from the
DNO. These are guaranteed standards of performance as defined in the electricity
regulations which cover relevant regulations on issues such as distributor’s fuse,
supply restoration (under normal and severe weather conditions), investigation of
voltage complaints, etc.
From the above discussions, it is obvious that different benchmarks need to be set by
utilities because certain aspects that affect distribution system reliability need to be
considered. For example, historic network build and performance, geographical -
weather, vegetation and population densities; technical (e.g. overhead versus
underground cabling), etc. However, CIs and CMLs are the basis for performance
regulation within distribution companies worldwide. UK DNOs face challenging
115
___________________________________________________________________________________
requirements stipulated by the Regulator (Ofgem) related to the allowable investment

funding, quality-of-supply (i.e. reliability), losses incentive regimes, and security
levels, etc. Consequently, the task of planning modern distribution networks has
become much more complicated, which has to take into account several different
aspects such as, nature of generation and load profiles of new customer types, smart
solutions, LCT connections and so on.
The next section provides details of the developed cable ranking indices in which CIs
and CMLs are used.
5.5 Proposed cable ranking index
Utilities replacement programmes are usually based on ranking the assets for
replacement using a pre-specified criterion. Over the last decade, DNOs in the UK
have moved towards ‘condition based replacement strategies’ that are based on asset
condition monitoring and asset health index calculations [24, 25]. The UK Regulator
has recently recognized that replacement and reinforcement activities should be
funded from a single budget [96]. Significance of considering system aspects in
replacement planning and vice versa is also illustrated in Table 5-1 through the
comparison of costs of copper losses and loss-of-lives for 1km of 11(6.6)kV cable
with the following data: loading 50%; energy price £0.1/kWh; replacement cost
(excavate & lay) £100,000/km; nominal life 40yr; annual maintenance rate 2%/yr. It
should be noted that the ratio of replacement to reinforcement costs is typically 5(4):1
in UK.
Table 5-1: Losses vs replacement costs for MV cable
Cost of 1 h
Cable Type Rating (A) Resistance (Ω/km) Cost of 1h Life (£)
Losses (£)
3c 95 Cu 300 0.194 0.44 0.51
3c 240 Al 390 0.128 0.48 0.51
3c 300 Cu 560 0.064 0.50 0.51
However, asset ranking can also be based on the financial implications for a DNO in
the forthcoming regulatory – planning period. Distribution companies are regulated
monopolies and they get capital allowances for ‘load’ and ‘non-load’ related
investments for the next regulatory period. In that respect, distribution companies
have a pre-specified ‘replacement budget’ and they need to make prioritization lists of
116
___________________________________________________________________________________
different types of assets to be replaced. Prioritization (or ranking) can then be based
on the DNO financial exposure in the regulatory period. The following costs can be
considered:
 Cost of failure repairs, and
 Lost revenue or incurred penalty under the quality-of-supply (QoS)

incentive regime. It usually consists of two parts; the first is related to
customer-interruptions (or SAIFI) and the second to customer-
minutes lost (or SAIDI).
Using this approach, ranking of cables for replacement can be based on the financial
criterion proposed and published in [36]. This can be expressed as:
𝐶𝑂𝑆𝑇1 = ∆𝐶𝐼 ∗ 𝐶𝐼𝐼𝑛𝑐 + ∆𝐶𝑀𝐿 ∗ 𝐶𝑀𝐿𝐼𝑛𝑐 + 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐹𝑎𝑖𝑙𝑢𝑟𝑒𝑠 ∗

£ 𝑝𝑒𝑟 𝑓𝑎𝑖𝑙𝑢𝑟𝑒 (5-1)
where ∆CI and ∆CML are customer interruptions (SAIFI) and customer-minutes-lost
(SAIDI) caused by the considered cable section in the studied regulatory period (5 to
8 years). CIInc and CMLInc are monetary values specified in the QoS regulatory regime
(i.e. PBRs) and £ per failure is the repair (minimal repair) cost of a failure due to
ageing. The average number of repairs during the planning period is used to compute
the indices.
Ranking criterion (5-1) represents direct financial costs which a DNO incurs due to
unreliability of a cable section. However, it does not completely reflect importance of
a cable section within the network and does not consider the additional cost incurred
due to lengthy failure duration for reasons of simplification. Criterion (5-1) can be
extended with an ‘indirect cost term’ that gives impact of the considered cable section
(i.e. its fault) on all other cables in terms of additional ageing caused by increased
currents. This additional impact factor on other network cables is considered in the
proposed ranking criterion in (5-2):
𝐶𝑂𝑆𝑇2 = 𝐶𝑂𝑆𝑇1 + ∑ ∑ (𝑐𝑜𝑠𝑡𝐿𝑜𝐿 (𝐼𝑖𝑗𝑝𝑜𝑠𝑡 (𝑡)) − 𝑐𝑜𝑠𝑡𝐿𝑜𝐿 (𝐼𝑖𝑗𝑝𝑟𝑒 (𝑡))) (5-2)

𝑡 𝑖𝑗∈Ω𝑏𝑟
where summation goes over all hours t in the planning period and branches ij ϵ Ωbr
other than the considered one, 𝑐𝑜𝑠𝑡𝑖𝑗𝑟𝑒𝑖𝑛 is cost of reinforcing branch ij, and ΔLTij(t) is
relative loss-of-life increment in branch ij hour t, calculated from the post-fault and
117
___________________________________________________________________________________
pre-fault currents using model (2-33) where Arrhenius model (2-25) and IEC thermal
model θc(I) are substituted. The second term can be dominant in criterion (5-2) due to
the high cost of reinforcement. Finally, cable sections can be ranked by using the
relative loss of life (2-33).
5.6 Cable ranking for replacement planning (based on the first proposed ranking
scheme)
Performance of the individual cable sections within the planning period was then
assessed with the second developed SMCS procedure (Section 4.3.3) and ranking
profiles developed using the proposed financial based ranking criterion – equation
(5-1). The changes in system reliability indices - SAIDI and SAIFI caused and
allocated to each cable section are shown in Figure 5-1 (a) – (d), whilst the cost
assumptions are presented in Table 5-2.
25 23.02 =1.2 =1.4 =1.8 2.5

SAIDI (hours /cust. 5 yrs)
=1.2 =1.4 =1.8

SAIFI (intrr./cust. 5 yrs)
21.66
19.01 20.94 19.71 20.09
20 18.45 17.37 2.0
15 1.5
6.44
10 1.0
0.65 6.11
1.36 5.78
1.27 5.71
5.09
5.01
4.92
0.79 4.95
5 0.5
0.82
0.75
1.23
1.32
0 0.0
L1 L2 L3 L4 L5 L7 L8 L26 L1 L2 L3 L4 L5 L7 L8 L26
Aging cable/line Aging cable/line
[a] [b]
10 8.97 =1.2 =1.4
1
SAIDI (hours /cust. 5 yrs)
=1.2 =1.4
SAIFI (intrr./cust. 5 yrs)
8.70
8 0.8
6.28
6 5.68 0.6
4 0.4
2 0.94 0.2
0.84 0.82 0.89
0 0
L5 L6 L40 L48 L5 L6 L40 L48
Aging cable/line Aging cable/line
[c] [d]
Figure 5-1: Ranking of critical cables for replacement (a) ∆𝑆𝐴𝐼𝐷𝐼 33 Bus system (b)
∆𝑆𝐴𝐼𝐹𝐼 33 Bus system (c) ∆𝑆𝐴𝐼𝐷𝐼 69 Bus system (d) ∆𝑆𝐴𝐼𝐹𝐼 69 Bus system
It can be observed that the ageing cables L1-L5, L7, L8 and L26 of the 33 bus test
system, as well as L5, L6, L40 and L8 of the 69 bus network contribute ~30% and
~10% of the total system interruptions and the durations, respectively. This means
118
___________________________________________________________________________________
that improving or replacing these cables would enhance the system reliability
considerably.
The most important factors affecting the frequency of ageing failures and system
reliability indices are the loading cycle, the degree of cable ageing (i.e. ) and the
network topology during contingencies. The loading cycle combined with the time-
varying cable thermal ratings can have substantial impact on the reliability indices in
certain intervals of the planning period (e.g. winter, when temperature and thermal
resistivity are low), which coincides well with the contingencies. The impact of
parameter  on reliability indices is even higher, because it directly affects the length
of time to failure of the cable. For example, cable L5 on the 33 bus network (Figure
5-1.a) will contribute 1.36 hours of loss of supply over 5 years when  = 1.2, 5.78
hours when  = 1.4, and 23.02 hours when  = 1.8. Similarly, cable L6 on the 69 bus
network (Figure 5-1.c) will contribute 0.84 and 8.97 hours when  = 1.2 and  = 1.4,
respectively. In addition, those outages which can be restored from adjacent feeders
have lesser impact on SAIDI (switching time is shorter than repair time). Finally, it
can be deduced that the test systems performance for values of  = 1.2 appear to show
clearer distinction between cable effects on network performance. Thus, replacing the
critical cables identified in Figure 5-1 will help to significantly improve the reliability
of the test systems.
Table 5-2: Cost assumptions
£ per SAIFI £ per kM cable £ per new

SAIDI incentive*
failure* incentive* reinforcement* cable*
£17.85 £10,000 £5.84 £20,000 £30,000
*SAIFIInc and SAIDIInc are monetary values specified in the QoS regulatory regime
and £ per failure is the cost to repair a failure. We have used CIInc=17.85 £/cust,
CMLInc=5.84 £/min and £ per failure=£10,000 in our studies based on the values used
by a DNO (ENWL) [95]. While assumed values of an additional £10,000 and £20,000
on £ per failure were used for £ per kM cable reinforcement and £ per new cable as
shown in Table 5-2.
The proposed financial-based cable ranking was then computed by using the
SAIFI/SAIDI values from Figure 5-1, allocated numbers of faults and cost figures of
119
___________________________________________________________________________________
Table 5-2; the results are outlined in Figure 5-2. It provides a logarithmic bar chart of
the new ranking of ageing cables in the 33 bus (Figure 5-2.a) and 69 bus test systems
(Figure 5-2.b), for the different ageing stress levels (i.e. parameters ).
=1.2 =1.4 =1.8

£0.72 £0.66 £0.74
£0.61 £0.64 £0.61 £0.59 £0.57
Industry standard indices
£0.193
£0.192
£0.192
£0.177
£0.166
£0.157
£0.153
£0.148
(in millions)
£0.20
£0.042
£0.035
£0.035
£0.034
£0.026
£0.025
£0.023
£0.020
£0.02
L1 L2 L3 L4 L5 L7 L8 L 26
Aging cables/lines
[a]
£3.02
£3.12 £2.97 =1.2 =1.4
Industry standard indices
£2.62
£2.14
(in millions)
£2.12 £1.91
£1.62
£1.12
£0.297
£0.292
£0.290
£0.290
£0.62
£0.12
L5 L6 L 40 L 48
Aging cables/lines
[b]
Figure 5-2: First proposed cable ranking scheme for replacement:

(a) IEEE 33 bus test system (b) IEEE 69 bus test system
Figure 5-2(a) shows that the cables best candidates for replacement on the 33 bus
system are L5, L2, L3, L4, L1, L7, L8 and L26. This is the same set of cable sections
as those obtained using the loss of life criterion (Figure 5-2.a), but the ranking order is
different. Similarly, cable sections best candidates for replacement on the 69 bus
system are L6, L5, L40 and L48, which is a different priority order to the ranking
based on the thermal loss of life (Figure 5-2.b).
120
___________________________________________________________________________________
Finally, from the discussions thus far it can be concluded that the effective
prioritization of distribution cable for replacement has to take into account overall
system modelling, cable ageing, physical characteristics of components, as well as
envisaged network operating conditions in the planning period. From the results thus
far, it can be concluded that the test systems performance for values of  = 1.2 appear
to be more practical (with clearer distinction of cable ranking). In addition, replacing
the critical cables identified in Figure 5-2 will help to significantly improve the
reliability of the test systems.
The next section provides a summary of the topics discussed in this chapter.
5.7 Chapter summary
The chapter provides insights into investment planning of distribution networks.

Investment planning is categorised as reinforcement (load related), replacement (non-
load related) and quality-of-supply planning. General system performance indices are
then discussed followed by utility performance indices, which are related to the
quality-of-supply incentive regime. Two new cable ranking methods using the
industry-based approach are developed for effective replacement planning. The new
cable ranking models make it easier to locate areas (cable segments) where reliability
needs to be improved and areas that have high impact on the entire network.
121
CHAPTER SIX
SMART DISTRIBUTION SYSTEMS
This section presents optimal integration of low carbon distributed generation - DGs
in distribution systems, as well as optimal management of distribution networks with
DG. Wind power and photovoltaic distributed generation rank highest among the
newly installed DGs in recent years; the integration of wind-DGs is further
investigated in this section. Once the DG is connected to the network, it should be
operated in an optimal way. To this end, a ˈsmartˈ solution, dynamic network
reconfiguration, is examined. Following the review of available literature, a new
concept and methodologies for optimal connection of low carbon technologies (wind-
DG) and optimal operation of distribution networks is proposed. The main objective
of both methodologies is to extend cable lives by optimally managing new wind
sources and network reconfiguration in near real time. Both approaches make use of
the physical aspects of ageing assets (cables) that are used in probabilistic studies to
optimally integrate wind-DG and manage the network in order to reduce asset (cable)
ageing and postpone their replacement.
6 Smart power networks
The renewable DG directly connected to distribution networks continues to rise in

response to global climate changes and international agreements to improve fuel
diversity. The mix of diverse electricity energy sources is difficult to manage due to
several factors. The anticipated future electricity networks will have large central
generators operating at relatively constant output together with a large number of
intermittent distributed low carbon generating plants that depend on uncontrollable
environmental parameters such as solar irradiation, wind speed etc. In addition,
increasing levels and complexity of load demands require new systematic approaches
in dealing with completely new network uncertainties.
Well-sited DGs can improve the services of existing electricity networks, enhance
reliability, relieve network congestion, provide ancillary services, and so on [97-100].
In the UK, DNOs cannot influence developers’ choice of locating LCTs.
Nevertheless, in this thesis arbitrary optimal LCT connection locations were
122
Chapter six: Smart grid planning study
___________________________________________________________________________________
determined and examined based on the assumption that incentives can be provided to
encourage developers to choose LCT connection points that have been established to
mitigate ageing effects on network performance. In addition, a scenario whereby the
chosen location by the developer and the DNO preferred (optimal) connection point
coincides could happen, thereby enabling the application of the technique discussed in
this thesis.
Despite substantial technical and economic benefits attributable to DG integration, a

rapidly ageing distribution network can make it difficult to realise the full potential of
the DG [24]. Thus, the increase in the DG penetration into ageing distribution
networks raises the questions of infrastructure capacity and replacement needs.
Furthermore, increased integration of DGs often requires implementation of novel
operation and control methods, which are often called ‘smart’ solutions. In that
respect, networks with integrated low carbon technology DGs and ‘smart’ solutions
are often named ‘smart grids’ (SG).
SG is a type of electricity network that can intelligently integrate all elements

necessary to deliver secure, cost-effective, and sustainable electricity supplies. The
SG concept supports the transition to sustainable low carbon generation, facilitates
changes in customer/consumer behaviours, enables novel operation and control
concepts, and provides more efficient and higher quality services. Moreover, to
successfully implement a SG, there is need to monitor and replace some ageing
(older) equipment [24, 101]. Thus, it becomes imperative to develop ꞌsmartꞌ ways of
balancing energy supply and demand along with managing ageing infrastructure
replacements. This fact is well supported by the current capital investment in
distribution asset replacement, which is typically 4-5 times higher than the
reinforcement programme.
Asset replacement planning needs to develop in a way in order to extract maximum

benefits from SG (integration of DGs and application of ‘smart’ solutions). The
‘smart’ solutions studied in this thesis are dynamic network reconfiguration (i.e.
rerouting of power flows through less loaded circuits or non-ageing components) and
application of real time thermal ratings. There are other ‘smart’ solutions, for example
energy storage and demand side management, which can also contribute to the
reduction in asset replacement on distribution networks.
123
___________________________________________________________________________________
6.1 Smart measures to reduce replacement cost
More DG, especially wind and solar units, are expected to be connected to the
distribution systems in future due to existing government incentives, technological
improvements and installation cost reductions [26, 28, 100, 102-108]. A common
challenge for countries experiencing high DG penetration levels is how to optimally
integrate DG into the network with acceptable costs and manageable risks. Where
distribution companies own the DG units, their connection to the distribution network
can be varied, and an important element- of a viable integration strategy is placing the
DG units at the right location and at the right time [102]. It was assumed in this study
that such flexibility in connecting DG units exists, which increased complexity of the
problem. Where this is not the case, i.e. where DG units are in possession of third
parties, their location is known and optimal connection branches to the network need
only to be determined.
Network planning can be typically based on the 5-7 year period into the future.
Distribution network planners are faced with uncertainties due to rising DG
penetration levels and ageing of existing assets, and this was the motivation behind
developing new approaches to DG integration planning. Even if all DG units are
owned by third parties, the DNOs have limited information on where and when the
DG developer will apply for connection, which further compounds the uncertainties
for network planners. The examination of connection plans and network management
schemes have been the subject of most recent publications [98, 109-113].
The introduction of smart measures to manage power systems planning and

operations will present a good opportunity to investigate the potential investment cost
reduction in distribution networks. The measures will ensure more efficient assets
utilization, secure, and sustainable electricity supply, at the least possible cost. Hence,
incorporating optimal network management schemes is essential for planning the
integration of DG (such as smart metering, conservation voltage regulation, integrated
volt/var control etc.).
124
___________________________________________________________________________________
6.2 Optimal wind-DG integration
Advancements in wind turbine technology have made it possible to install small-sized

wind turbine generators within distribution networks. Due to the relatively lower cost
of small-sized wind generators, private investors are increasingly interested in
electricity generation on lower voltage levels typical for distribution system.
Distributed wind generators (wind-DG) can be utilized as a source of electrical energy
as well as back up to enhance distribution system reliability through the possibility of
operating in an islanding mode, as is done in some countries. The islanding scheme on
a DG is initiated in a timely manner when a fault occurs which is dependent on the
fault location and the actual status of the distribution network. During the islanding
mode, power supply to the isolated feeder is fully dependent on the installed DG. In
the UK distribution networks, wind-DG integration is on the rise; however, islanding
operation is not utilized yet.
In general, wind generation are installed in power networks for two main reasons: to
replace traditional energy generation sources or to complement existing energy
sources. When large-scale wind generation is installed, the characteristic uncertainty
and randomness of its output may result in significant system stability issues due to
generation intermittency and power forecasting errors. Consequently, higher amounts
of expensive energy storage systems or operating reserves are needed to keep system
reliability at acceptable levels; a trade-off between achieving good operational
economics and acceptable reliability is required. Consequently, researchers have
suggested several techniques for achieving optimal network operational planning and
management [97, 101, 110, 113, 114].
Literature review on DG integration in distribution systems
The impact of wind-DG integration on network performance is discussed in numerous

papers [28, 106, 107, 114-118]. Two main challenges are identified from relevant
literature:
1. How to choose an optimal DG location (locating and sizing)? and

2. How to determine an optimal DG integration strategy considering more future
connections and necessary reinforcements of the existing lines?
125
___________________________________________________________________________________
There is a consensus amongst researchers that optimal location and sizing of any DG
is key to extracting maximum benefits. Most distribution systems were designed for
radial operation and a small amount of ‘sensibly’ connected distributed generation can
have a positive effect on distribution system reliability [115].
The impact of different wind generator operational modes and the probability of
successful operations including ‘islanding’ during contingencies have been
investigated in [114-116, 119, 120]. The research shows that a ‘sensibly’ located
wind-DG connection is most likely to enhance the system operation with respect to,
electrical losses, voltage profiles, load balancing, and system reliability. During
system normal operation, the system performance can be optimized for enhanced
assets utilization and reliability while during contingency periods, the system can be
managed to reduce the effects of overloading (accelerated aging) of assets. The
models proposed in this thesis for optimal integration of wind-DG and optimal
operation of distribution networks are set without considering islanding operations,
because it is not permitted in the UK industry.
Common objectives in optimisation problems involving distribution systems include

minimisation of system (copper) losses, minimisation of investment and operational
costs, minimisation of environmental impact and enhancement of reliability metrics.
With the advent of improved optimization solvers, several shortcomings of the
previous research contributions are receiving attention. Some of the shortcomings
include: ignored or inappropriately applied voltage constraints; radiality constraints;
ignored or incorrectly quantified reliability issue; missed budgetary constraints;
ignored existing assets age and condition and so on [97, 110, 113]. New models that
are developed for distribution expansion planning though the DG integration do not
still properly address ageing assets in the DG integration planning. From
examinations carried out on available literature, research on design and planning of
electricity networks can be classified into three major areas:
 Strategic/longer-term planning of investments and main network

configuration(s);
 Network planning of individual asset investments for the near future;
and
126
___________________________________________________________________________________
 Construction design of each network component to consider available

options.
In [97, 110, 113] DG optimal planning methods are analysed and reviewed. The
reviews focus on two key perspectives: the objective function with constraints and the
optimization formulation with solution. The objective functions define the goal of the
planners and it is considered along with the design constraints relevant to the pursued
aim. The distribution system optimisation problem is defined as a highly constrained,
highly dimensional, mixed integer, nonlinear, multi-modal optimisation problem with
a large number of local optima. Key highlights from the studies include:
i. Recent literatures are dominated by techniques for determining optimal location

and capacity of feeders, DG, and substations [97, 110].
ii. DG integration planning is mostly viewed from the angle of assisting to meet load
growths by deciding ‘sensible’ location and capacity. In some countries, the
operators cannot contribute much in decision making concerning DG location or
capacity [16].
iii. Network power losses; (re) configuration; budgetary constraints; and reliability,
issues receive the most attention in the problem formulation. In addition, not all
connection constraints are considered altogether, e.g. radiality and full
connectivity etc. [98, 121].
iv. Few publications have applied dynamic approach to DG connection planning – it
can be used to determine optimal planning solution that considers an entire
planning period.
v. The majority of power distribution utilities still follow traditional planning
approaches without considering advanced concepts from the literature.
vi. Some important objectives given far less attention include [97, 110]:
a. Demand side management
b. Equipment ageing & cost of wear
c. Integration of electric vehicles
d. Coordinated planning approaches
vii. Fewer publications discuss practical experiences of applied techniques such as the
impact of intelligent distribution management system – DMS [28].
127
___________________________________________________________________________________
From the perspective of optimization formulation with solution, the algorithms used
for network planning solutions can be categorised as mathematical programming
methods and heuristic methods. Mathematical programming methods solve a
mathematical formulation of the planning problem by finding an optimum solution.
The major drawback of the mathematical methods is that the final solution is largely
affected by the initial input values due to the use of approximations. The heuristic
techniques are methods that generate, evaluate, and select alternative solutions step by
step, while the metaheuristic algorithms use similar approach but without the users
help. Common metaheuristic methods include, genetic algorithms (GA), simulated
annealing (SA) etc. The major drawback of heuristic techniques is that an optimum
solution cannot be guaranteed. Highlights of the three main approaches reported are
provided below:
i. Classical optimization methods that involve the use of deterministic algorithms

[122, 123]:
a. Dynamic programming
b. Linear programming
c. Nonlinear programming
d. Mixed integer programming (MINLP is the most complex)
ii. Technical/heuristic methods apply technical switching strategies [89]:
a. Branch exchange method is the most common method
iii. Metaheuristic methods give better performance and more efficient solution
strategy [124, 125]:
a. Metaheuristic algorithms used in the literature adopt a default parameter
setting that controls the computations [110, 123].
From the above reviews, it is pertinent to note that, a comprehensive model that
considers all major elements of today’s real life distribution planning (such as ageing
assets in replacement planning) is still lacking, even though many aspects have been
individually addressed. In addition, premature convergence to local optima instead of
global optima is the main issue for all approaches in different degrees. In this thesis,
optimal network management models are developed with the goal to reduce cable
ageing and achieve effective replacement planning; cable ageing is expressed in terms
of cable thermal-loss-of-life. As part of the proposed methodology, two new
128
___________________________________________________________________________________
optimization models are developed: the first optimizes connection of new wind
sources by minimizing the connection cost and the cost of cable thermal loss-of-lives
in the planning period. In the next stage, the network is optimally reconfigured in such
a way to minimize thermal-loss-of-life of ageing cable(s). Both optimization models
are applicable to radially operated medium voltage networks.
6.2.1 Wind generator modelling for reliability assessment
For proper planning of power systems operations, it is important to accurately predict

wind generator outputs [102]. Several methods have been developed for computing
wind-DG generator outputs in reliability studies; they can be broadly categorized into
two groups. The first group is based on the physical characteristics, such as terrain,
obstacle, pressure and temperature as model inputs [126]. The second major category
is the statistical methods such as ARMA models, artificial intelligence techniques etc.,
which considers the use of raw wind speed data with relatively reduced computational
efforts in handling large data [127].
In reliability assessment studies involving wind power generators, some of developed

methods use the measured chronological wind speed data that are used to calculate
wind generator outputs [127, 128]. In some cases, the chronological wind speed
profiles are obtained from direct field measurements, while in others the profiles are
calculated or simulated. However, simplified and modified models are used in the
network reliability assessments in many cases due to difficulties to obtain the wind
speed data or data unavailability [127-129]. In some cases, the proposed
simplifications are based on the use of truncated data sets or on approximations,
which disregard the chronology of wind speeds [128].
Application of calculated chronological data sets involves the use of wind speed
simulators such as numerical weather prediction (NWP) to compute and develop wind
profiles in the absence of wind data. The NWP is a popular approach, which uses the
average wind speeds and auto-correlation coefficients from the wind site data to
forecast the wind speeds with reasonable accuracy. The authors in [130] applied NWP
technique to simulate wind velocities using a wind speed time series model based on
the auto regressive moving average (ARMA) technique. Some advantages of the
NWP approach include:
129
___________________________________________________________________________________
 It avoids the need to collect extensive measured data

 It avoids the need for handling large data measurements
 It reduces computational efforts in handling large data
Another popular technique for modelling and determining generator outputs form
wind speeds is ARMA stochastic process; it can be used to simulate the time and
location dependent wind speeds considering the wind speed’s inherent stochastic
characteristics. It is for this reason that the ARMA time series model is adopted in this
thesis to simulate periodic (hourly) wind speeds. References [127, 131, 132] provide
more details on ARMA model development; the main steps are given below.
Billinton compared the accuracy of several wind speed modelling techniques using
the wind speed data from two different locations in Canada (Boavista and Swift
current sites) [127]. The techniques compared include the hourly observed data,
hourly mean observed wind speeds, ARMA time series, MA time series, normal
distribution (with a goodness-of-fit test) and Markov chain models. Some key
findings that can be extracted from the studies conducted in [127, 128, 130, 132-135]
include:
 ARMA technique is comparatively most suitable and enjoys wider

acceptance in reliability studies.
 The accuracy of the final reliability indices is dependent on certain
parameters especially the accuracy of the observed mean wind speed.
 The closer the observed mean wind speed is to the rated output of
wind generator, the lesser the amount of wind speed data required to
predict future wind speeds when using ARMA technique.
 ARMA model thrives on determining seasonal and cyclic trends
within a dataset.
 Ignoring the auto-correlation of the wind speed with time can lead to
producing less accurate results.
Apart from the ARMA technique, other methods have also been used to characterize
the wind speed distributions such as artificial intelligence techniques, e.g. artificial
neural networks [128, 136, 137]. Several probability distribution functions such as
Pearson, Chi2, Rayleigh, Johnson, and Weibull distribution functions have been found
130
___________________________________________________________________________________
to be adequate in some cases. A summary of common wind speed forecasting

methods along with their key features and limitations is presented in Table 6-1
below[136, 138].
Table 6-1: Common wind speed forecasting methods
Wind forecasting method Key features Suitability

i. Weibull Probability
Simple, imprecise, and ignores Long term wind speed
Distribution
time chronology predictions
ii. Auto Regressive
Moving Average Mid and Short term wind
Involves time-series analysis
(ARMA) speed predictions
iii. Artificial Neural Complex, more accurate and

Short term wind
Network (ANN) influenced by nature of initial
prediction
data
The relation (6-1) below shows the fundamental mathematical expression of the
ARMA stochastic model:
n m
yt = ∑ φi yt−i + αt + ∑ θj αt−j (6-1)

i=1 i=1
Parameter yt is the value of the wind speed at time t while the coefficients of function
yt include autoregressive parameters φi for i = 1, 2, 3 … . . . n and moving average
parameters θj for j = 1, 2, 3 … . . . m. {αt } is a white noise process normally
independently distributed (NID) with a zero mean and standard deviation 𝜎. The
standard deviation can be estimated as the root mean squared (RMS) error of the fitted
ARMA model. Thus, the main effort towards the development of ARMA time series
model (6-1) involves determination of order 𝑛 and 𝑚 before fitting the model and
calculation of the AR, MA and data noise parameters.
The Box and Jenkins method provides empirical guidelines for the determination of
the 𝑛 and 𝑚 values if one of the parameters is zero [131]. However, it has been shown
that ARMA model of order (𝑛, 𝑛 − 1) can be used to approximate most stationary
stochastic systems, as suggested in [139]. Thus, solving for (𝑛, 𝑚) is reduced to
determining only 𝑛 parameter. The procedure suggested in [139] is based on F-
criterion and has also been successfully applied in [132] with satisfactory accurate
131
___________________________________________________________________________________
results. For this reason, the same process as given by Billinton et al [132] for
determining the order of ARMA models is adopted in this thesis.
In this thesis, it was assumed that the directly driven permanent magnet synchronous
generator (DDPMSG) is the wind-DG technology and has failure and repair rates of
1.501/year and 0.0081/year respectively [140, 141]. ARMA technique is developed
using historical wind speed data from a UK city (Canterbury site, 12 meters above sea
level) [93]. ARMA technique is applied to forecasting wind speeds in the planning
period. The ARMA technique procedure can be summarized as follows:
 Normalize historic wind speeds at hour t using the mean and standard
deviation at the same hour t.
 Estimate parameters of the ARMA(n, m) stochastic process fitted to
the normalized historic wind speeds [142], where n is order of auto-
regressive model and m is order of the moving-average model.
 Transform back the forecast normalized wind speeds using the mean
and standard deviation in the period t.
 Determine wind turbine active power output in the considered hour t
using the nonlinear relationship [127].
6.2.2 Wind generation forecast
Wind speed forecasting is an important component in wind farm design and planning.
The wind energy outputs need to be evaluated before the wind generator is
constructed. In this thesis, the wind speed forecast is required for system operation
simulation in the planning period. ARMA wind speed forecasting technique for wind
generator output is used to develop the wind generator’s model. The output of the
wind turbine is determined from its nonlinear relationship with the wind speed.
For illustrative purposes, the relationship between wind speed and wind generator
electrical output for older wind turbine generators can be represented by the sample
wind power curve as shown in Figure 6-1. It is based on the model developed in
reference [129].
132
___________________________________________________________________________________
Figure 6-1: Typical wind generator output characteristic
The output power of a wind generator unit is expressed as a function of wind speeds
in a piecewise format as given by the expression (6-2).
0 0 ≤ 𝑉𝑤 < 𝑉𝑐𝑖
(𝐴 + 𝐵𝑉𝑤 + 𝐶𝑉𝑤2 )𝑃𝑟 𝑉𝑐𝑖 ≤ 𝑉𝑤 < 𝑉𝑟 (6-2)
𝑃(𝑉𝑤 ) {
𝑃𝑟 𝑉𝑟 ≤ 𝑉𝑤 < 𝑉𝑐𝑜
0 𝑉𝑤 ≥ 𝑉𝑐𝑜
where
𝑉𝑤 Wind velocity/speed
𝑃𝑟 Rated power output of the wind turbine
𝑉𝑐𝑖 Cut-in wind speed
𝑉𝑟 Rated wind speed
𝑉𝑐𝑜 Cut-out wind speed
The constants 𝐴, 𝐵 and 𝐶 are given by equations (6-3), (6-4) and (6-5), respectively.
1 𝑉𝑐𝑖 + 𝑉𝑟 3
𝐴= [𝑉 (𝑉 + 𝑉 ) − 4(𝑉 𝑉 ) ( ) ]
(𝑉𝑐𝑖 − 𝑉𝑟 )2 𝑐𝑖 𝑐𝑖 𝑟 𝑐𝑖 𝑟
2𝑉𝑟 (6-3)
𝐵= )
[4(𝑉𝑐𝑖 + 𝑉𝑟 ( ) − (3𝑉𝑐𝑖 + 𝑉𝑟 )] (6-4)
(𝑉𝑐𝑖 − 𝑉𝑟 )2 2𝑉𝑟
133
___________________________________________________________________________________
𝐶= [2 − 4 ( ) ] (6-5)
(𝑉𝑐𝑖 − 𝑉𝑟 )2 2𝑉𝑟
Figure 6-1 shows that for wind velocities (𝑉𝑤 ) slower than the cut-in wind speed (𝑉𝑐𝑖 )
of a given wind generator, no power output is produced until the cut-in wind speed is
attained. The cut in speed provides the minimum torque required to overcome the
static inertia of the generator blades. Figure 6-1 also shows that, for all wind speeds
between the cut-in wind speed and the rated wind speed (𝑉𝑟 ) of the wind turbine, the
turbine power output increases rapidly following a quadratic behaviour shown in
relation (6-2). However, as the power output reaches the limiting wind speed (𝑉𝑟 ), the
maximum output limit of the wind turbine is activated and no further rise in power
output can be delivered. In such cases, the angles of the turbine blades are usually
adjusted to keep the power output at a constant level. When the wind speeds are too
high and there is a risk of turbine damage, a braking system is deployed to bring the
turbine to a standstill. This point is called cut-out wind speed (𝑉𝑐𝑜 ),
6.2.3 Optimal integration of wind generation
Optimal connection of wind-DG into distribution networks can provide ˈhealthyˈ

technical benefits in the form of voltage profile and stability improvements, reliability
increase, and network upgrade deferrals. However, inappropriate location and
operations of DG resources may lead to significant variable losses, accelerated asset
ageing, reverse power flows and lowered system stability [105, 118, 143, 144].
DG planning approaches that consider various economic, environmental, and

technical benefits have been developed and analysed over the last decades. Network
upgrade deferral is one of highly desired benefits due to its potential to alleviate
economic constraints. It is the ability to defer the required investment on
reinforcing/replacing system assets such as feeders, line sections, and transformers
into the near future. Studies in [106, 118] showed that depending on DG technology
mix incorporated into a network, DG units can have diverse impacts on the network
deferral. For instance, wind-DGs provide substantial benefit as reactive power
sources, while CHP units are better suited to relieve the overloads on power networks
because their output is not intermittent. However, just like in most other studies, the
issue of assets replacement was not considered within the DG integration planning.
134
___________________________________________________________________________________
6.2.4 Developed optimization models
To develop a mixed integer nonlinear programming (MINLP) model, it is required to

express the desired objective(s) and design constraint(s) as functions of discrete and
continuous unknowns (also known as design variables or decision variables). MINLP
models are developed in this thesis to address problem of optimal connection of new
wind-DG to the distribution network. Two such models are presented below, the first
with the ‘standard’ objective function and the second with the improved objective,
that considers cost of cable loss-of-lives. Application of these models is given in
Chapter 6.
a) Nomenclature
A list of symbols is presented first to facilitate understanding of MINLP models.
Sets
Ω𝐷𝐺 Set of DG units
Ω𝑛𝑜𝑑 , Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 Sets of nodes and potential nodes for DG unit k
Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘 Set of potential ‘DG branches’ for location nk
Ω𝑏𝑟 Set of existing branches
Ω𝑠𝑢𝑏 Set of primary substations
Parameters
𝐴, 𝐵 Parameters of the Arrhenius model
𝑡𝑝 , 𝑡𝑞 Duration of historic period p planning period q
𝑐𝑜𝑠𝑡𝑛𝑛𝑒𝑤
𝑘𝑖
Cost to construct new ‘DG branch’ nki
𝑟𝑒𝑖𝑛
𝑐𝑜𝑠𝑡𝑖𝑗 Cost to reinforce existing branch ij
𝑆𝑖𝑗𝑟𝑎𝑡 , 𝐼𝑖𝑗𝑟𝑎𝑡 Nominal rating of branch ij in MVA and A
𝑃𝐷𝐺𝑖 , 𝑄𝐷𝐺𝑖 Active and reactive power supplied by DG, node i
𝑃𝐷𝑖 , 𝑄𝐷𝑖 Active and reactive demand at node i
𝑒𝑥 𝑟𝑒𝑖 𝑛𝑒𝑤
𝐼̅𝑖𝑗 , 𝐼̅𝑖𝑗 , ̅̅̅̅̅
𝐼𝑛𝑘𝑖 Rating of existing, reinforced and new branch
𝑉𝑖𝑚𝑖𝑛 , 𝑉𝑖𝑚𝑎𝑥 Minimum/ maximum voltage limits at node i

𝑁𝐷𝐺 , 𝑁𝑏 Number of DG units and existing buses
𝑐𝑜𝑠𝑡𝐸𝐸 , 𝑐𝑜𝑠𝑡𝐸𝑁𝑆 Cost of delivered and non-delivered kWh
135
___________________________________________________________________________________
𝑅𝑖𝑗 Resistance of branch ij

𝐾𝑖 Fictitious load of 1.0 at DG nodes; 0 otherwise
𝐶𝐼𝐼𝑛𝑐 , 𝐶𝑀𝐿𝐼𝑛𝑐 Incentive rates for CIs and CMLs
𝑁𝑂𝑓𝑎𝑖 , 𝐶𝑓𝑎𝑖𝑙 Number of failures and cost per failure
Functions/Variables
𝐿(∙), 𝜃𝑐 Life measure and conductor temperature
𝐿𝑇 , ∆𝐿𝑇 (∙) Total and hourly thermal loss-of-life
𝐿𝑝 (∙), 𝐿𝑞 (∙) Life in historic period p and planning period q
Conductor temperature in historic period p and planning period
𝜃𝑐,𝑝 , 𝜃𝑐,𝑞
q
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 Initial investment cost for DG connection
𝑇𝑂𝑇
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 DG investment cost including loss-of-life cost
𝑦𝑛𝑘 Binary decision variable for nodal location of DG k
Binary construction variable to reinforce existing branch ij and
𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 , 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
to construct new ‘DG branch’ nki
𝐿′′ 𝑇𝑖𝑗 Thermal loss-of-life of cable section ij in planning period
Binary operation variable indicating whether branch ij or nki is
𝑜𝑐𝑖𝑗 , 𝑜𝑐𝑛𝑘𝑖
closed (=1) or open (=0)
Apparent, active, reactive power and current flow in branch ij –
𝑆𝑖𝑗 , 𝑃𝑖𝑗 , 𝑄𝑖𝑗 , 𝐼𝑖𝑗
functions of voltages and angles
Binary decision variable indicating whether the first feeder
𝑚𝑖 section is supplied from a primary (node i is MV busbar of the
substation)
Active and reactive power supplied by primary substation at
𝑃𝑆𝑖 , 𝑄𝑆𝑖
node i
𝑦𝑖 Equal to 1 for existing nodes and to 𝑦𝑛𝑘 for new DG nodes
Binary decision variable indicating whether the existing circuit
𝑐𝑒𝑖𝑗
in branch ij is used or not
Fictitious power flow in existing branch ij and new ‘DG-
𝑓𝑖𝑗 , 𝑓𝑛𝑘𝑖
branch’ nki
𝐶𝑂𝑆𝑇(𝑡) Hourly cost of loss-of-lives, losses and reliability
𝑐𝑜𝑠𝑡𝐿𝑜𝐿 (∙) Hourly cost of loss-of-lives
136
___________________________________________________________________________________
𝐿𝐶𝑖 (∙), 𝑐𝑜𝑠𝜑𝑖 Load curtailment and power factor at node i

𝐶𝑂𝑆𝑇1 , 𝐶𝑂𝑆𝑇2 Costs in planning period used for cable ranking
∆𝐶𝐼, ∆𝐶𝑀𝐿 CIs and CMLs associated with a cable section
b) Conventional objective function
This is a minimum cost investment model in which costs of connecting new DG units
and potential reinforcement of medium voltage network are minimized. It was
assumed that each DG plant can be built at several locations, and that each DG
location can be connected to the existing network via several new branches. The
investment cost COSTINV is:
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 = 𝑚𝑖𝑛 ( ∑ ∑ 𝑦𝑛𝑘 ∑ [𝑐𝑜𝑠𝑡𝑛𝑛𝑒𝑤

𝑘𝑖
∙ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
]
𝑘∈Ω𝐷𝐺 𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 𝑛𝑘 𝑖 ∈Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘
𝑟𝑒𝑖𝑛
+ ∑ [𝑐𝑜𝑠𝑡𝑖𝑗 ∙ 𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 ]) (6-6)
𝑖𝑗 ∈Ω𝑏𝑟
where yn is binary decision variable indicating whether potential location n is used for
DG siting, crninew is binary variable for construction of new circuit between potential
DG location n and network node i, whilst crijrein is binary variable for reinforcement of
existing branch ij. The first term in describes costs of new ‘DG-branches’ required to
connect several DG units to the network. The first summation goes over all DG units
k ε ΩDG, the second covers all feasible locations nk ε ΩDG-nod,k for DG unit k, whilst the
third sum is over all potential ‘DG-branches’ nki ε Ωnew-br,nk connecting feasible DG
location nk and existing node i for each such location nk and each DG unit k. The
second term in gives investment costs of reinforcing existing branches ij. Where third
parties own the DG units, a DG location nk is known, there is no decision variable ynk
and the second summation can be omitted.
c) Constraints
The following logical constraints on binary decision variables need to be added to the
problem formulation:
∑ 𝑦𝑛𝑘 = 1 ∀𝑘 ∈ Ω𝐷𝐺 (6-7)

𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘
137
___________________________________________________________________________________
∑ ∑ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
=1 ∀𝑘 ∈ Ω𝐷𝐺 (6-8)
𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑘
(1 − 𝑦𝑛𝑘 ) ∑ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
=0 ∀𝑛𝑘 ∈ Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 ∀𝑘 ∈ Ω𝐷𝐺 (6-9)
𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑘
𝑐𝑒𝑖𝑗 + 𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 = 1 (6-10)

Condition (6-7) specifies that a DG plant can be located at only one site n, whilst
expression (6-9) provides that there are no new circuits constructed in branches
towards non-used sites (i.e. where yn=0). Constraint (6-8) specifies the permissible
number of new circuits connecting the new DG plant to the existing network. These
relations imply ‘non-firm’ connection (i.e. DG is teed-off with one circuit), whilst
‘firm’ connection (i.e. DG is looped in) is modelled by setting the right-hand side to
be equal to 2. Finally, equation (6-10) specifies that either the existing circuit or the
new one is used in existing branch ij.
d) Proposed objective function
In the proposed approach, the main goal of DG integration was to improve

performance of the existing network in terms of extending lives of (old) cable sections
and deferring cable replacement. Total cable thermal loss-of-life is defined by
equation (2-33) in which the second term 𝐿′ 𝑇 𝑖𝑗 is calculated within the main SMCS
procedure. Quantities 𝐿′ 𝑇 𝑖𝑗 ∈ {0 − 1} were used as a weighting factor that multiplies

𝑇𝑂𝑇
replacement cost of existing cable sections, so that the total cost 𝐶𝑂𝑆𝑇𝐼𝑁𝑉 is:
𝑇𝑂𝑇 𝑟𝑒𝑖𝑛
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 = 𝑚𝑖𝑛 (𝐶𝑂𝑆𝑇𝐼𝑁𝑉 + ∑ [𝑐𝑜𝑠𝑡𝑖𝑗 ∙ 𝑜𝑐𝑖𝑗 ∙ 𝐿′′ 𝑇 𝑖𝑗 ]) (6-11)
𝑖𝑗 ∈ Ω𝑏𝑟
Closer inspection of objective function (6-11) shows that the minimum cost plan will
involve cable sections with largest loss-of-lives to be open (open sectionalizing switch
and or close tie-line switches to alter the power flow pattern) in the optimum solution
- the network configured in such a manner as to reduce the current flowing through
those sections. However, since the overall objective is to minimize ageing of all cable
sections (both open and closed), cost model for individual branches often applied in
marginal pricing of transmission and distribution assets is used [145]. Then the
objective function (6-11) can be re-written as:
138
___________________________________________________________________________________
𝑟𝑒𝑖𝑛
𝑇𝑂𝑇
𝑐𝑜𝑠𝑡𝑖𝑗
𝐶𝑂𝑆𝑇𝐼𝑁𝑉 ≈ 𝑚𝑖𝑛 (𝐶𝑂𝑆𝑇𝐼𝑁𝑉 + ∑ [ ∙ 𝑆𝑖𝑗 (∙) ∙ 𝑜𝑐𝑖𝑗 ∙ 𝐿′′ 𝑇 𝑖𝑗 ]) (6-12)
𝑆𝑖𝑗𝑟𝑎𝑡
𝑖𝑗 ∈ Ω𝑏𝑟
Summation in (6-12) goes over sections with existing cable. The replacement cost is
specified as unit cost in £/MVA multiplied by (peak) power flow, binary decision
variable ceij and relative loss-of-life in the planning period. The set of logical
constraints (6-7) to (6-10) also applies.
An appropriate operating regime should be selected for the DG connection. The vast
majority of distribution networks is demand dominated, which means that maximum
power flows are experienced during winter peaks. This also applies to 11kV networks
where ‘big’ DG units are often connected to 11kV bus bars of the primary substations.
In such cases, the selected operating regime is winter peak. On the other hand, in
generator dominated networks the maximum power flows are experienced during
summer minimum loading regime. Then it is appropriate to study the summer
minimum operating regime.
The presented approach is based on the assumption that existing branches can be
reinforced if their capacity is insufficient, which can be expressed by 𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 = 1 in
(6-10). It is possible to use a different approach, in which existing branches are
optimally reinforced following the first solution of the conventional DG connection
problem. If this is the case, only the capital costs of DG connections and loss-of-life
costs of ageing cable are minimized (i.e. there is no second term in equation (6-6)),
and different set of constraints needs to be set.
e) Power flow constraints
Nodal AC active and reactive power balance equations need to account for power
flows over existing closed branches, as well as over newly constructed closed
branches. These equations are specified for all nodes including new locations n for all
DG plants, which can be expressed as:
139
___________________________________________________________________________________
[𝑚𝑖 ∙ 𝑃𝑆𝑖 + 𝑃𝐷𝐺𝑖 − 𝑃𝐷𝑖
− ∑ 𝑜𝑐𝑖𝑗 ∙ 𝑃𝑖𝑗 (∙)

𝑖𝑗∈Ω𝑏𝑟
−∑ ∑ ∑ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
∙ 𝑜𝑐𝑛𝑘 𝑖 ∙ 𝑃𝑛𝑘𝑖 (∙)] ∙ 𝑦𝑖
𝑘∈Ω𝐷𝐺 𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘
𝑛𝑘 𝑖∈Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘
=0 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-13)
[𝑚𝑖 ∙ 𝑄𝑆𝑖 + 𝑄𝐷𝐺𝑖 − 𝑄𝐷𝑖
− ∑ 𝑜𝑐𝑖𝑗 ∙ 𝑄𝑖𝑗 (∙)

−∑ ∑ ∑ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
∙ 𝑜𝑐𝑛𝑘𝑖 ∙ 𝑄𝑛𝑘𝑖 (∙)] ∙ 𝑦𝑖
𝑘∈Ω𝐷𝐺 𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘
𝑛𝑘 𝑖∈Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘
=0 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-14)
where mi is binary variable indicating whether supplies from a (primary) substation
are used or not, ocij is binary operational (‘open/close’) variable, whilst yi=1 for all
existing nodes and yi=yn (i.e. binary variable) for all new nodes – locations. Active
Pij(·) and reactive power flows Qij(·) over branches ij are specified via terminal
voltages and angles in the standard way [121].
Branch thermal constraints are expressed via current flows Iij(·); different inequality
constraints are set for existing and new branches:
𝑒𝑥 𝑟𝑒𝑖𝑛 2
𝑜𝑐𝑖𝑗 ∙ 𝐼𝑖𝑗2 (∙) ≤ (𝑐𝑒𝑖𝑗 ∙ 𝐼̅𝑖𝑗 + 𝑐𝑟𝑖𝑗𝑟𝑒𝑖𝑛 ∙ 𝐼̅𝑖𝑗 ) ∀𝑖𝑗 ∈ Ω𝑏𝑟 (6-15)
𝑛𝑒𝑤 2
𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
∙ 𝑜𝑐𝑛𝑘𝑖 ∙ 𝐼𝑛2𝑘𝑖 (∙) ≤ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
∙ (𝐼̅̅̅̅̅
𝑛𝑘 𝑖 ) ∀𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘 ∀𝑛𝑘
∈ Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 ∀𝑘 ∈ Ω𝐷𝐺 (6-16)
Note the existence of binary variable crninew on both sides of relation (6-16), because
no additional constraint was set on the link between crninew and ocni (i.e. where
crninew=0 => ocni=0). Voltage magnitude constraints are specified for both existing
and new nodes:
𝑦𝑖 ∙ 𝑉𝑖𝑚𝑖𝑛 ≤ (𝑦𝑖 ∙ 𝑉𝑖 ) ≤ 𝑦𝑖 ∙ 𝑉𝑖𝑚𝑎𝑥 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-17)
where yi has the same meaning as in equations (6-13) and (6-14). It is present in all
inequalities because the constraints should not be set at the potential generation
connection nodes where yi=1.
Another formulation of the branch thermal limits could be that the loading constraints
140
___________________________________________________________________________________
on aged cable sections be limited to some defined maximum limits in order to control
the current flowing through critical sections.
f) Other Constraints
Integration of DG units into a radially operated network creates a possibility that the
optimal solution contains isolated islands in which all loads are supplied by DG units.
This problem was overcome in [98, 121] by introducing the concept of fictitious
power flows. Fictitious loads are connected at nodes with DG units and they create
fictitious power flows over the interconnected branches prohibiting the network
islanding. Fictitious power balance equations are defined for each node:
∑ 𝑜𝑐𝑖𝑗 ∙ 𝑓𝑖𝑗 + ∑ 𝑐𝑟𝑛𝑛𝑒𝑤

𝑘𝑖
∙ 𝑜𝑐𝑛𝑘𝑖 ∙ 𝑓𝑛𝑘 𝑖 = 𝑦𝑛𝑘 ∀𝑛𝑘
𝑖𝑗∈Ω𝑏𝑟 𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑘
∈ Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 ∀𝑘 ∈ Ω𝐷𝐺 (6-18)
∑ 𝑜𝑐𝑖𝑗 ∙ 𝑓𝑖𝑗 + ∑ 𝑐𝑟𝑛𝑛𝑒𝑤

𝑘𝑖
∙ 𝑜𝑐𝑛𝑘 𝑖 ∙ 𝑓𝑛𝑘𝑖 = 0 (6-19)
𝑖𝑗∈Ω𝑏𝑟 𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑘
where fictitious load 𝐾𝑛 = 0 for nodes without DG and 𝐾𝑛 = 1 for nodes with DG.
Fictitious power flows in branches also need to be limited:
|𝑓𝑛𝑘 𝑖 | ≤ 𝑁𝐷𝐺 ∙ 𝑐𝑟𝑛𝑛𝑒𝑤

𝑘𝑖
∙ 𝑜𝑐𝑛𝑘 𝑖 ∀𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘 ∀𝑛𝑘 ∈ Ω𝐷𝐺−𝑛𝑜𝑑,𝑘
∀𝑘 ∈ Ω𝐷𝐺 (6-20)
|𝑓𝑖𝑗 | ≤ 𝑁𝐷𝐺 ∙ 𝑜𝑐𝑖𝑗 ∀𝑖𝑗 ∈ Ω𝑏𝑟 (6-21)
where NDG is the total number of DG plants. Fictitious power flows are forced to
zero for open or non-constructed branches, as shown by right-hand side of inequalities
(6-20) and (6-21).
Radial operation of MV networks can be expressed mathematically as ‘the total
number of in-service branches is equal to the total number of buses minus one’ [121],
which is valid only when a single supply substation is used. In the general case when
several substations can supply the network under consideration, this condition
becomes:
141
___________________________________________________________________________________
∑ 𝑜𝑐𝑖𝑗 + ∑ ∑ ∑ 𝑐𝑟𝑛𝑛𝑒𝑤
𝑘𝑖
∙ 𝑜𝑐𝑛𝑘𝑖
𝑖𝑗∈Ω𝑏𝑟 𝑘∈Ω𝐷𝐺 𝑛𝑘 ∈Ω𝐷𝐺−𝑛𝑜𝑑,𝑘 𝑛𝑘 𝑖 ∈ Ω𝑛𝑒𝑤−𝑏𝑟,𝑛𝑘
(6-22)
= 𝑁𝑏 + 𝑁𝐷𝐺 − ∑ 𝑚𝑖
Ω𝑠𝑢𝑏
Finally, a very small load (=0.001 pu) is connected to all transfer nodes (i.e. nodes
with zero injection) to provide connectivity of all nodes in the optimal solution [98,
121].
6.3 Smart operations (network reconfiguration)
Distribution systems operation schemes include system load balancing, switching -

(re)configuration, generation dispatch, voltage control etc. Switching operations are
one of the most important activities used in distribution network operations
management [4, 110, 113, 146]. Network switching changes the network topology
and reallocates loads from one feeder to another in order to isolate faults, balance
loads across the feeders, or decrease real power losses etc. However, some
reconfiguration options could cause voltage and thermal issues with severe
consequences on the system assets especially to underground cables (UGC) and
transformers [89, 98, 121, 146, 147]. These potential threats can be prevented by
studying the power flows before realizing any proposed network modification
(especially switching and topology change) [89, 98, 121, 146-148].
Developed model for optimal network reconfiguration with the objective to reduce
asset ageing and postpone investments is presented below.
6.3.1 Optimal network reconfiguration
System operators are always interested in optimal network configuration that will
achieve desired performance. Generally, in order to facilitate and maintain simplicity
of operation and effective network protection, 11(6.6) kV distribution networks are
radially operated. However, the vast majority of distribution networks are constructed
as interconnected networks with normally open switches at certain ‘convenient’
points, so that the individual feeders have radial topology in system operation. The
normally open switches can be closed and other switches opened, which changes the
configuration of the distribution network.
142
___________________________________________________________________________________
The technological advances made in automation industry in recent years have made
power utility companies interested in deploying more measurement and automation
devices onto power networks. Distribution network automation involves utilizing
sensors, communication channels, local intelligent computer system, centralized
mainframe computer, control algorithms, and so on; it can be very beneficial in the
area of distribution network reconfiguration [4, 24, 25, 94, 146]. Network
reconfiguration involves closing and opening of switches in a power distribution
system in order to alter the network topology, and thus the power flow from the
substation to load points.
Dynamic network reconfiguration (DNR) refers to a process of executing an adaptable

topology optimisation process in ‘near real-time’ in order to achieve certain
objective(s) such as minimization of losses, enhancing reliability, load balancing,
improving voltage profile, etc. [98, 113, 123, 148-153]. By deploying a DNR scheme,
network topology can be frequently changed whilst maintaining the radial operational
configuration – often called radiality, connectivity of all load points, whilst complying
with voltage and line thermal limits. An appropriately developed DNR scheme can
enhance both system reliability and economics.
The aim of deploying a network reconfiguration (NR) scheme in this thesis is to

extend lives of critical assets (cables), enhance system reliability and economics while
maintaining the feeder’s radial topology, connectivity, and operational constraints.
The proposed DNR model is used to determine the alternative network configuration
that would produce minimal thermal consequences on the critical ageing cables. Here,
the network reconfiguration problem generates a set of branches, more precisely one
set from each supply source that has the least line currents flowing through the
critically ageing cable sections. Another formulation is a more general approach that
includes cost of reliability and losses beside the cost of cable loss-of-lives.
Literature review on network reconfiguration
As mentioned earlier, distribution network configuration and reconfiguration are very

important operational functions on distribution networks and have been extensively
studied in [89, 98, 113, 121, 147, 148, 150, 152, 153]. The objective of the network
configuration is to find the feeder topology so that losses are minimized, load is
143
___________________________________________________________________________________
balanced and/or system reliability is improved during system normal operation [89,
113, 121, 154]. Restoration of disconnected customers with possibly minimal
deviation of operational constraints is the primary goal of network reconfiguration in
system emergencies [147, 148, 150, 152, 153].
The distribution systems optimization models were developed for both capital
investment minimization and optimal operational management of the network. They
can be viewed from two angles: optimization objectives and optimization techniques
used to solve the problem (see Figure 6-2). Common optimization objectives in this
wide area are the minimization of copper losses through network reconfiguration [97,
110, 113, 121, 146, 155], minimization of expansion planning investments involving
costs of substations, capacitors, operational reliability, greenhouse gasses emission
[97, 110, 113] and the maximization of DG utilization [97-99, 110, 148].
Network reconfiguration schemes have been successfully modelled and evaluated

using computer simulations [89, 97, 98, 110, 124, 146-148, 153]. There are three
primary reasons to reconfigure network during normal operation - for load balancing,
for power loss minimization and reliability improvement. Depending on current
loading conditions, reconfiguration maybe necessary to eliminate overloads on
specific network components using load balancing techniques [89, 124, 153]. As the
loading conditions on the system change, it may also become economically viable to
reconfigure the network in order to reduce the real power losses in the network,
extend lives of critical ageing components, or improve reliability. Network
reconfiguration during emergencies (i.e. outages of network components) is done with
the goal to restore as many customers (or load) as possible. Some of the operational
constraints in emergency conditions can be relaxed (e.g. voltage magnitudes in rural
networks).
Many researches on distribution network reconfiguration have focused on solving

minimum losses configuration problem [113, 146, 155]. Thus, various methods have
been proposed to find good feasible solutions in acceptable time and each technique
has its merits and demerits. The techniques outlined in Figure 6-2 are can also be
applied to the network reconfiguration problem.
144
___________________________________________________________________________________
Distribution system
optimization
Techniques Objectives
used explored
Metaheuristic q Minimization of
Classical/Analytical Heuristic/Technical
approaches line losses
q Minimization of
Deterministic planning costs
q GA Classic Modified
approaches q Maximization of
q Ant colony heuristics heuristics DG utilization
q Simulated q Maximization of
q Linear annealing network reliability
programming q Etc.
q Non-linear
programming
q Mixed-integer
non-linear
programming
Figure 6-2: Distribution networks optimization - literature review
Solution methods can be classified as metaheuristic, classical/analytical, and

heuristic/technical approaches. The metaheuristic approaches or population-based
stochastic approaches are more computationally intensive than the deterministic
algorithms but suffer from premature convergence issues that are yet to be solved [97,
110, 113]. There are numerous (re)configuration approaches based on metaheuristics
(artificial intelligence) techniques and numerous models applied to solving
reconfiguration problem [113, 121, 149, 153, 156]. The methods include: simulated
annealing [156], ant colony [149], particle swarm algorithm [157], genetic algorithm
[124, 146, 148] etc. Metaheuristic (population based stochastic approaches) are
inspired by natural behaviours of animals and other species and it is more efficient
strategy compared to heuristic methods. Metaheuristic methods are well suited for
solving mixed-binary non-linear optimization problems and are more likely to achieve
solutions that are close to the global optimal. However, these methods require
numerous iterations to obtain the best solution and thus higher performance
computing facilities. Furthermore, premature convergence to local optima instead of
global optimum is still an issue for metaheuristic approaches.
Classical approaches have fewer applications and their implementation is generally

considered as less efficient [97, 98, 113, 121, 146-148]. The classical analytical
methods are simpler and convenient for smaller networks. They can become
145
___________________________________________________________________________________
complicated when applied on complex networks consisting of many nodes and

branches. One advantage of analytical methods is that the optimal solution should be
obtained. Classical approaches consist of deterministic algorithms. The deterministic
algorithms enjoy wider acceptance and they include linear programming, non-linear
programming, mixed-integer non-linear programming (MINLP) and dynamic
programming techniques [89, 97-99, 110, 113, 121, 146-148]. For larger and complex
distribution networks, the network (re)configuration is a complex mixed integer
nonlinear problem and the solution method has to be able of dealing with such
problems [98, 113, 151]. With the development of commercial optimization software,
the MINLP methods are becoming more popular for several reasons, mainly because
that they are not plagued by premature convergence and can be relied upon to deliver
global optimal solution [98, 111, 118, 121, 123, 154, 158, 159]. The recent advances
in high-performance computing facilities also made mixed-integer programming
methods more acceptable for solving network reconfiguration problems [146, 151].
From an optimization viewpoint, the DNR problem has many discrete variables and
highly nonlinear power flow equations in continuous variables and it is an imperative
to employ an efficient global search method like the MINLP algorithm. Furthermore,
the proposed MINLP model must also be able to handle infeasibility issues when
radiality constraints are violated, thermal limits exceeded and/or when the voltage
magnitudes at certain buses are below or above the prescribed limits.
The second largest group of solution techniques are the heuristic/technical methods.
The heuristic/technical approaches involve the application of technical switching
strategies and do not involve the use of mathematical optimization algorithms. These
approaches exist in two main variants: classical heuristic approaches and modified
heuristic approaches – Fig. 5-2 [113, 146]. The classical heuristic approaches include
the simple branch exchange method (the most common method), optimal power flow
pattern and enhanced power flow methods [113, 146]. The determination of global
optimum is very difficult when using classical heuristic approaches [113, 146]. Early
publications can be traced to Merlin and Back [160] in which they proposed a branch
and bound type heuristic technique for distribution network reconfiguration to
minimise branch power losses. According to their branch and bound type method, all
network switches are by default closed to obtain a meshed network. Then,
sectionalizing switches are opened one at a time until a new radial structure is
146
___________________________________________________________________________________
achieved, and the switch selected to open at each time is to ensure that power losses
are minimized. However, the method considers only active loads, neglects other
power flow constraints such as voltage magnitudes etc. The authors in [161] modified
the method developed in [160] by proposing a similar method but with the switch
opening selection based on determining the optimum power flow pattern. They also
included a power flow method capable of handling both radial and weakly meshed
distribution networks. A number of authors, such as [89, 151, 152, 162-164], also
reported different modifications to how the switches are selected and network
reconfigured to minimise line losses. In general, heuristic algorithms are fast and
simple, but optimality of the global solution is usually a potential issue. The modified
heuristic approaches found in the literature are rules-based comprehensive methods,
which consist of a rules database incorporated into the heuristic methodology in order
to determine local minima in a less complex way [97, 110, 113, 146].
Despite the advances in computational mathematics and numerical modelling, the

majority of distribution utility planners still adopt heuristic processes and empirical
rules backed by expert judgements and practical analysis for planning decisions [97,
110, 113]. However, the current state of the art in both research and computational
facilities does not support such arguments.
In comparison to the investigated optimization models in distribution systems, the

model proposed in this thesis considers a different optimization objective, which can
be generally expressed as optimal management of distribution networks with DG to
reduce cable ageing and replacement requirements. Thus, a new concept and
methodology for optimal use of low carbon technologies and ‘smart’ solutions is
proposed. The main objective is to extend cable lives by optimally managing new
wind sources and network (re)configuration in real time. The entire problem is set
within probabilistic framework, which is presented in the next chapter. The proposed
dynamic reconfiguration model is presented below.
6.3.2 Proposed optimal reconfiguration model
The developed optimal reconfiguration model is the final control action within the
considered hourly interval in the developed simulation procedure. It is used to
reconnect customers, alleviate overloads and reduce thermal ageing of cable. The
147
___________________________________________________________________________________
problem is formulated as a MINLP model and is briefly presented below.
a) Objective Function
The proposed objective is to minimize thermal loss-of-life of ageing cable, which can
be expressed through relation (2-33) in which Arrhenius model (2-25) and IEC cable
thermal model [44] are incorporated. The thermal loss-of-life needs to be calculated
over the entire planning period, so that the ‘exact formulation’ would become a multi-
stage optimization model in which the number of stages is equal to the number of
years multiplied by 8760 hours. Given that such a large-scale MINLP problem is
impossible to solve, a single-stage optimization with the following objective function
can be used:
𝐶𝑂𝑆𝑇𝑎𝑔𝑒 (𝑡) = 𝑚𝑖𝑛 ( ∑ 𝐿𝑇 𝑖𝑗 (𝑡) ∙ 𝑜𝑐𝑖𝑗 ∙ 𝐼𝑖𝑗 (∙)2 ) (6-23)

𝑖𝑗∈Ω𝑜𝑙𝑑−𝑏𝑟
where LTij(t) is (relative) thermal loss-of-life up to the studied hour t, ocij is

operational binary variable indicating branch open/close status and Iij(·) is current
flow through branch ij. Objective (6-23) tends to minimize further degradation of
ageing cable in hourly interval t, whilst giving priority to the most ‘worn-out’ cables.
It is deemed appropriate when the replacement budget is pre-specified by the
Regulator. However in the case when there is a single budget for replacement and
reinforcement investments, it may be appropriate to give priority to the most
expensive (i.e. longest) ‘worn out’ cable sections. In that case, reinforcement cost
costijrein should be included in the objective function (6-23).
An alternative to the above approach is consideration of several types of costs in the

studied hourly interval. More specifically, hourly costs of cable ageing, active power
losses and curtailed loads can be combined into the composite objective function. To
this end, the hourly loss-of-life of cable i-j needs to be calculated. It is given by
expression:
𝐵 𝐵 𝐵 𝐵
∆𝐿𝑇 𝑖𝑗 (𝑡) = exp ( − ) = exp ( − ) (6-24)
363 𝜃𝑐 (𝑡) 363 𝜃𝑐 (𝐼(𝑡))
which is equal to unity when conductor temperature θc(t) is equal to nominal value of
900C. The second equality in (6-24) is obtained by substituting cable IEC thermal
model (2-5), where θc(I(t)) can be simplified to:
148
___________________________________________________________________________________
𝜃𝑐 (𝐼(𝑡)) = 𝜑 ∙ 𝐼(𝑡)2 + 𝜔 (6-25)
Expression (6-24) combined with (6-25) is relatively complex and its linear
approximation (i.e. Taylor’s expansion up to the first order) around the nominal point
can be used:
∆𝐿𝑇 𝑖𝑗 (𝐼(𝑡)) ≈ 1 + 𝜉 ∙ (𝐼(𝑡) − 𝐼𝑛 ) (6-26)
𝐿𝑜𝐿
The cost of hourly loss-of-life of cable i-j in the considered interval t, 𝑐𝑜𝑠𝑡𝑖𝑗 (𝐼(𝑡)),
is then equal to:
𝑟𝑒𝑖𝑛 𝑟𝑒𝑖𝑛
𝐿𝑜𝐿
𝑐𝑜𝑠𝑡𝑖𝑗 𝑐𝑜𝑠𝑡𝑖𝑗
𝑐𝑜𝑠𝑡𝑖𝑗 (𝐼(𝑡)) = ∙ ∆𝐿𝑇 𝑖𝑗 (𝑡) ≈ ∙ (1 + 𝜉 ∙ (𝐼(𝑡) − 𝐼𝑛 )) (6-27)
8760 ∙ 𝑇𝑙𝑖𝑓𝑒 8760 ∙ 𝑇𝑙𝑖𝑓𝑒
where 𝑇𝑙𝑖𝑓𝑒 is the cable design life (40 years). The composite objective function
representing the hourly cost of cable ageing and losses, as well the cost of reliability is
then equal to:
𝐶𝑂𝑆𝑇𝑡𝑜𝑡 (𝑡)
𝐿𝑜𝐿
= 𝑚𝑖𝑛 ( ∑ 𝑜𝑐𝑖𝑗 ∙ (𝑐𝑜𝑠𝑡𝑖𝑗 (𝐼(𝑡)) + 𝑐𝑜𝑠𝑡𝐸𝐸∙ 𝑅𝑖𝑗 ∙ 𝐼𝑖𝑗 (∙)2 )
𝑖𝑗∈Ω𝑏𝑟 (6-28)
+ ∑ 𝑐𝑜𝑠𝑡𝐸𝑁𝑆 ∙ 𝐿𝐶𝑖 )
𝑗∈Ω𝑛𝑜𝑑
where Rij is cable resistance in branch i-j, costEE is cost of 1kWh and costENS is cost of
1kWh of electrical energy not supplied. A fixed value of costENS without distinction of
load types (e.g. industry, residential etc.) is used in this study for reasons of
simplification. An advanced approach is to associate different costENS values for
different types of customers in the formulation.
b) Model Constraints
Model constraints are simpler than for the wind integration, because there are neither
locational nor investment binary variables; the only decision variables are operational
variables ocij. Assuming that sets of existing nodes and branches now include the
known DG locations and connecting ‘DG-branches’, and branch thermal limits are
modified according to the solution of model (6-12) to (6-22), the constraints are:
149
___________________________________________________________________________________
𝑚𝑖 ∙ 𝑃𝑆𝑖 + 𝑃𝐷𝐺𝑖 + 𝐿𝐶𝑖 − 𝑃𝐷𝑖 − ∑ 𝑜𝑐𝑖𝑗 ∙ 𝑃𝑖𝑗 (∙) = 0 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-29)

𝑚𝑖 ∙ 𝑄𝑆𝑖 + 𝑄𝐷𝐺𝑖 + 𝐿𝐶𝑖 ∙ tan 𝜑𝑖 − 𝑄𝐷𝑖 − ∑ 𝑜𝑐𝑖𝑗 ∙ 𝑄𝑖𝑗 (∙) = 0 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-30)
𝑒𝑥 2
𝑜𝑐𝑖𝑗 ∙ 𝐼𝑖𝑗2 (∙) ≤ (𝐼̅𝑖𝑗 ) ∀𝑖𝑗 ∈ Ω𝑏𝑟 (6-31)
𝑉𝑖𝑚𝑖𝑛 ≤ 𝑉𝑖 ≤ 𝑉𝑖𝑚𝑎𝑥 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-32)

0 ≤ 𝐿𝐶𝑖 ≤ 𝑃𝐷𝑖 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-33)
∑ 𝑜𝑐𝑖𝑗 ∙ 𝑓𝑖𝑗 = 𝐾𝑖 ∀𝑖 ∈ Ω𝑛𝑜𝑑 (6-34)

|𝑓𝑖𝑗 | ≤ 𝑁𝐷𝐺 ∙ 𝑜𝑐𝑖𝑗 ∀𝑖𝑗 ∈ Ω𝑏𝑟 (6-35)
∑ 𝑜𝑐𝑖𝑗 = 𝑁𝑏 + 𝑁𝐷𝐺 − ∑ 𝑚𝑖 (6-36)

𝑖𝑗∈Ω𝑏𝑟 Ω𝑠𝑢𝑏
Nodal active and reactive power balance equations (6-29) and (6-30) contain load
curtailments, branch thermal constraints are defined by (6-31), nodal voltage limits by
(6-32), whilst load curtailment limits by (6-33). Relations (6-34) and (6-35) specify
constraints on fictitious power flows needed when DG units are connected, whilst
radiality constraint is given by (6-36).
The MINLP model consisting of the objective function (6-23) or (6-28) and the
associated model constraints is incorporated into the developed probabilistic
simulation model for reliability assessment. A probabilistic simulation is required to
examine the impact of adding new wind-DG units, implementing new operational
schemes on existing cables’ ageing and to plan the cables replacement.
6.4 Smart grid planning study
This section provides details of the techniques used to develop the proposed smart
distribution system planning and operation approach. The entire proposed
methodology consists of several main building blocks, which are for convenience
presented below:
 Line/cable thermal models (Section 2.3.1),

 IEC-Arrhenius-Weibull model for ageing cable (Section 3.6),
 Cable ranking criteria (Section 5.5),
 Wind generation model (Section 6.2.2)
150
___________________________________________________________________________________
 Wind-DG integration model (Section 6.2.4),

 Network reconfiguration model (Section 6.3.2) and
 Sequential Monte Carlo simulation (SMCS) model (presented below),
A high level description of the developed SMCS procedure is as follows. System

elements (cables) are classified as either at the ‘useful life’ or ‘wear-out’ stage of their
life cycle. Standard two-state model based on exponential distribution of up and
downtimes is used to model components from the first group. Aging distribution
cables are modelled with the aid of the proposed IEC-Arrhenius-Weibull model for
up-times and exponential model for down-times. Integration of wind-DG units and
dynamic network reconfiguration with the objective to reduce cable ageing are
incorporated into the SMCS procedure. Some of the procedure outputs, such as
customer interruptions (CIs), customer minutes lost (CMLs) and the numbers of
failures are associated with each cable section, because they clearly show each
section’s contribution towards unreliability and can be used for cable section ranking
(equation (5-1)). Another ranking scheme was proposed as the impact of individual
cable failures onto other cable ageing was added (equation (5-2)). The proposed
ranking methods were finally compared to the cable ranking based on the
conventional thermal loss-of-life model (equation (2-33)).
6.5 Details of SMCS approach for smart grid planning studies
Sequential Monte-Carlo simulation (SMCS) technique is used to simulate time

chronological events that characterise power system operation over a stipulated
planning period. The results (in the form of several types of indices, as well as their
probability distributions) of the simulation provide an insight on the expected
performance levels of the investigated system. However, in order to distinguish
between ageing and non-ageing cable and to determine cable ageing parameters (i.e.
Arrhenius constants), it was necessary to simulate both historic and planning period
using ‘similar’ SMCS procedures. All components are considered to be at the ‘useful
cycle’ of their lifetimes in the historic period, so that random sampling is drown from
exponentially distributed uptimes and downtimes [2, 3]. Following the completion of
simulation of the historic period, cable sections are classified as ageing and non-
151
___________________________________________________________________________________
ageing and the proposed IEC-Arrhenius-Weibull model is used in the second SMCS
run to simulate planning period. A summary of the study is as follows:
 Development of a probabilistic method which considers all asset

interactions on the network for replacement planning of ageing distribution
cable.
 Development of the optimal wind integration and network reconfiguration
models that account for costs of cable loss-of-lives in the planning period.
 Development of cable ranking lists for replacement that take into account
financial costs in terms of CIs, CMLs, number of failures and additional
costs of loss-of-lives of other cable sections.
 Development of the SMCS procedure that integrates all blocks mentioned
above.
6.6 Proposed SMCS model with optimal wind-DG integration and network
reconfiguration
A flowchart of the overall methodology is shown in Figure 6-3. The first block is
hourly simulation of the network historic period - SMCSI with aim to find parameters
of the Arrhenius ageing model [36, 65] and classify cable as ageing and non-ageing.
Analysis of the planning period starts with the second block, in which new DG units
are connected with the minimal capital cost for DG connection and network
reinforcement. Results from the first two blocks are used to initialize the SMCS II of
the planning period on an hourly basis.
The last stage of hourly analysis is network reconfiguration, with the objective to
minimize hourly cost of variable losses, cable loss-of-lives and cost of reliability. One
of the simulation stage outputs, the loss-of-lives of all cable sections, is then used to
find the improved connection of new DG units by minimizing the connection and
reinforcement costs, as well as costs of cable loss-of-lives in the planning period.
Convergence of the iterative process is obtained when there is no difference in DG
connections between two consecutive iterations. Reliability and cost indicators,
ranking lists for cable replacement and investment & operation costs are calculated at
the end of the procedure.
152
___________________________________________________________________________________
Analysis of historic period (SMCSI): Classification of cables as non-ageing &

ageing
Optimal integration of wind DG: Objective = minimize (cost of DG connection

investment & reinforcements) – equations (6-6) to (6-10)
Perform SMCSII in planning period: Investigate impact of

network reconfiguration to extend lives of ageing cables –
equations (6-28) to (6-36)
Optimal integration of wind

DG: Objective = minimize (cost
Is this 1st
Yes of DG connection and cost of
iteration?
ageing cable loss-of-life) -
equations (6-11) to (6-22)
No
Any changes in loss-of-lives

Yes
or DG connection nodes?
No
Compute final reliability indices, ranking lists and replacement costs
Figure 6-3: Flowchart of the global methodology for cable replacement in the
presence of wind-DG and network reconfiguration
Initialization of the methodology, integration of wind-DG units, developed SMCS

procedure with optimal network reconfiguration and model outputs are presented
below.
6.6.1 Integration of wind generation and network reconfiguration
Optimal connection of new DG units is done before the simulation of network

operation in the planning period, as shown in Figure 6-3. Different approaches had to
be applied for the first and subsequent ‘main’ SMCS runs, because cable thermal loss-
of-lives in the planning period are calculated in the next step – main SMCS run. A
simplified flow chart of the simulation procedure is shown in Figure 6-4 and the
individual steps are briefly summarized below:
153
___________________________________________________________________________________
Input network data, DG locations, demand and generation profiles, cable

physical & installation properties, failure data
Next hour Set the initial hourly period of the SMCS
when
network Randomly select component up-times and down-times using:
does change 1. IEC-Arrhenius-Weibull model for up-times of aging cables
2. Exponential model for down-times and up-times of non-aging cables
3. ARMA model for wind generation output profile
Next hour when
Random selection of loads from windows around profiles
the network does
Run power flow not change
Scheme to minimize total hourly cost
Record data for increments in loss-of-lives of other ageing cable(s)
and allocate SAIFI, SAIDI, ENS to the faulty cable section
Next hourly
interval is the end
YES
of simulation
period ?
NO
Next hour is with

YES NO
network change ?
The SMCS has

NO
converged?
YES
Compute final reliability indices & ranking lists
Figure 6-4: Flow chart of the SMCS for optimal wind-DG integration and dynamic
network reconfiguration
A simplified flowchart of the ‘main’ SMCS II procedure is shown in Figure 6-4.

Simulation is done over the next regulatory period whose length is between 5 and 8
years; it covers 5(8)*8760 hourly intervals and is repeated until the coefficient of
variation of the expected energy not served (EENS) is less than the prescribed
tolerance (5%).
Following the data input and SMCSII initialization, load-flow analysis of the current
hour is done. To this end, loads are sampled from a window around the profile and
wind generations are determined from the ARMA processes [127]. Where
154
___________________________________________________________________________________
appropriate, statuses of network components are set/changed using the results from
the previous samplings and/or possible resupply from the back-feeding feeders. The
up-and down-times of non-ageing components are sampled from the exponential
model [2], whilst ‘IEC-Arrhenius-Weibull’ model as published in [36] is used to
sample up-times of ageing cable that have just been repaired. This model describes
the wear-out stage with the aid of Weibull probability distribution in which scale
parameter is replaced with L(θc) from the Arrhenius model (1), and conductor
temperature θc is calculated from the IEC cable thermal model [44], giving function
θc(I(t)). In this way, cumulative failure function F(θc(I(t)) is used to sample up-times
of ageing cable as published in [36].
Optimal network reconfiguration is done next with the objective to minimize hourly
cost of losses, cable loss-of-lives and curtailed demand. Reliability indices SAIFI (or
CIs), SAIDI (or CMLs) and ENS (if any), as well as thermal loss-of-lives are updated.
Reliability indices are allocated to the cable section which has caused supply
disruption. Increments in loss-of-lives of other cable sections are found and allocated
to the faulty cable section. They are calculated from the post-fault and pre-fault
currents; the latter are determined in a simple way from the radial network
configuration.
Following the SMCS convergence, final reliability indices, cost indicators and cable
ranking lists are found.
6.7 Results of optimal integration of wind-DG and network reconfiguration to

reduce cable ageing and replacement
This section presents results and analysis of the proposed optimal integration of wind-
DG and dynamic network reconfiguration for the IEEE 33-bus and 69-bus test
systems. The total numbers of connected customers are 1600 and 3400 for the 33-bus
and 69-bus systems, respectively. A 5-year planning period with hourly granularity
was investigated using the third developed SMCS procedure given in Section 6.6 -
Figure 6-3. The objective of the study is to optimally integrate wind-DG resources
and optimally manage the network using the dynamic network reconfiguration in such
a way that old cable(s) are preserved and network reliability improved. The discussion
of results begins with the ARMA model used as input to the SMCS simulations.
155
___________________________________________________________________________________
6.7.1 Results of ARMA time series for wind-DG modelling
ARMA time series model was used to generate the wind turbine generator outputs
using the wind speed data from [93]. Matlab software’s system identification toolbox
together with the specially designed ARMASA suite of programs [142] is used to
compute the ARMA model parameters. The parameters of the ARMA function are
derived based on historical data pattern identification using parameter estimations and
model checking [127]. ARMA(4,3) model provided the best match with 9.2% fitting
error. The reason for this level of error could be attributed to the inherent noise within
the measuring instruments coupled with erroneously recorded measurements. The
ARMA(4,3) model is presented in equation Error! Reference source not found.:
yt = 1.6191yt−1 + 0.3083yt−2 − 1.4924yt−3 + 0.5649yt−4 + αt −

0.7721αt−1 − 0.9269αt−2 + 0.7008αt−3 (6-37)
The validity of the developed ARMA (4,3) model was examined using an 8 hour
forecast. The developed ARMA (4,3) model was used to forecast future wind speeds
and its performance is displayed in Fig. 7-8 - this figure present plots of wind speeds
against time intervals in order to evaluate forecast performance measured in terms of
forecast values and statistical 95% confidence intervals. The measured wind speeds
are from the actual data file, while the 8 hour-ahead forecasts are obtained from the
ARMA (4,3) model – equation (6-37).
156
___________________________________________________________________________________
Figure 6-5: Measured and 8-hour ahead forecast wind speeds using the ARMA (4,3)
model
From the plots in Figure 6-5 it can be seen that the ARMA (4,3) model captures the
highly variable characteristics of the wind speed reasonably well, and can provide a
good estimate of the wind speed. These forecasts are utilized in estimating wind-DG
power outputs within the simulation procedures. Finally, by mapping the time-series
wind speeds to the wind-DG power output curve, the forecast intermittent wind
generations can be obtained.
6.7.2 Optimisation models results and system wide reliability indices
This section presents results and analysis of the proposed methodology for the two
test distribution systems described above. The total numbers of connected customers
are 1600 and 3400 for the 33-bus and 69-bus systems, respectively. A 5-year planning
period with hourly granularity was investigated using the approach from Figure 6-3.
The objective of the study is to find optimal cable replacement programme where
wind DG units and network management contribute to the reduced asset ageing. A
MATLAB software package was developed using Matlab’s MIPS solver. Typical
157
___________________________________________________________________________________
CPU times obtained on a 3.6GHz quad core, 64-bit system are shown in Table II;
SMCS denotes all Monte Carlo actions other than reconfiguration.
Table 6-2: CPU time and simulation duration for 33 and 69 buses test systems
CPU time (secs)

Network reconfiguration
Test System DG connection (NR) SMCS Total
33 bus 3741.90 22428.71 6032.17 32202.8
69 bus 6244.46 33615.64 7534.73 47394.8
Simulation results are shown in Table 6-3 to Table 6-7. Table 6-3 presents results of
the optimal connection of wind-DG units for three network management strategies.
System wide reliability indices along with sensitivity studies on the degree of ageing
are given in Table 6-4 for two proposed strategies. Table 6-5 shows a comparison
between the three types of ‘operational’ costs, which is, cost of cable loss-of-lives,
copper losses, and curtailed energy, within the planning period. Ranking of cable
sections using the thermal loss-of-live criterion is shown in Table 6-6, whilst Table
6-7 provides cable ranking based on the direct financial exposure (5-1) and ‘total
exposure’ (5-2).
Table 6-3: DG connection, reinforcements, and investment costs for three strategies
Optimization programs results

Distance: node to
Reinforced cable
Total investment
£ reinforcement
£ new DG cable
DG connection
Test System
point (bus #)
DG (kM)
Potential Applied DG integration

sections
cables
New DG DG nodes technique and Network

cost
Units (bus #) Reconfiguration

Coventional (eqns (6-6) to 5 1
5, 7, 8, 26 £60,000 £80,000 £140,000
A-33 2, 3, 4, 5, 6 (6-10) ) 26 1
33 bus
New approach (eqns (6-11) 5 1

2, 3, 26 £75,000 £60,000 £135,000
to (6-22)) 25 1.5
B-33 25, 26, 27 New approach with optimal 5 1
2, 26 £75,000 £40,000 £115,000
NR (eqns (6-11) to (6-36)) 25 1.5
Coventional (eqns (6-6) to 52 1
39, 40, 51, 2, 3, 48, 49 £60,000 £80,000 £140,000
A-69 (6-10) ) 50 1
52
69 bus
New approach (eqns (6-11) 50 1

5, 6, 40 £60,000 £60,000 £120,000
to (6-22)) 39 1
B-69 48, 49, 50 New approach with optimal 50 1
5, 6 £60,000 £40,000 £100,000
NR (eqns (6-11) to (6-36)) 40 1
Table 6-3 shows that the optimal point of connection of DG units is dependent on the
applied optimization objective. The conventional minimum cost approach (model
(6-6) to (6-10)) usually connects new DG centers via shortest branches, particularly
158
___________________________________________________________________________________
when voltage and thermal constraints are not an issue during summer minimum
periods. On the other hand, when the cost of cable ageing is considered, the optimal
DG connection plan can include longer (new) cable sections that will relieve loading
of some ageing cable. The required reinforcements are also different for the three
strategies; the minimal amount of reinforcement was in all cases obtained when a
combination of the new approach with optimal NR was applied.
Table 6-4: Reliability indices for two strategies
Impact indices - new approach (equations (6-11) to (6-22))

33 bus test system 69 bus test system
SAIDI (hours / cust. 5yr) SAIDI (hours / cust. 5yr)
1 1.05 1.1 1.2 1 1.05 1.1 1.2
( 0.015 0.361 0.799 2.153 ( 1.163 2.3252 6.634 40.349
(1pu) (24pu) (53pu) (143pu) (1 pu) (2pu) (5.7pu) (34.7pu)
SAIFI (interruptions / cust. 5yr) SAIFI (interruptions / cust. 5yr)
1 1.05 1.1 1.2 1 1.05 1.1 1.2
( 0.003 0.045 0.101 0.220 ( 0.126 0.225 0.665 4.054
(1pu) (15pu) (33pu) (73pu) (1pu) (1.8pu) (5.3pu) (32pu)
Impact indices - new approach with optimal NR (equations (6-11) - (6-36) )
33 bus test system 69 bus test system
SAIDI (hours / cust. 5yr) SAIDI (hours / cust. 5yr)
( 1.05 1.1 1.2 ( 1.05 1.1 1.2
0.311 0.699 1.99 2.052 5.924 39.146
(21pu) (47pu) (132pu) (1.8pu) (5.1pu) (33.7pu)
SAIFI (interruptions / cust. 5yr) SAIFI (interruptions / cust. 5yr)
( 1.05 1.1 1.2 ( 1.05 1.1 1.2
0.04 0.092 0.190 0.20 0.612 3.851
(13pu) (30pu) (63pu) (1.6pu) (4.9pu) (30.5pu)
Table 6-4 shows a comparative analysis of the system reliability indices for the two
proposed approaches. All results are also given in per unit values, where the reference
case in both cases is non-ageing SMCS with =1 for conventional approach (not
repeated in Table 6-4). The results show that the new approach with optimal NR
provides the highest reliability and consequently asset preservation. Table 6-4 also
shows the sensitivity of SAIDI and SAIFI indices for several Weibull scale parameter
 values which reflect intensity of cable ageing; they increase quite rapidly
(particularly SAIDI) when  increases. Practical choice of  should be based on
comparison with real-life data on historic performance.
159
___________________________________________________________________________________
Table 6-5 shows the total ‘operation’ costs (6-28) in the planning period for the two
developed strategies; they are calculated using the cost figures from Table I and
𝑐𝑜𝑠𝑡𝐸𝑁𝑆 = 15£/𝑘𝑊ℎ. 69-node test network has much fewer ageing cable sections
and the proportion of loss-of-life costs is smaller. Appreciable savings in ‘operation’
costs are obtained by applying the optimal NR. It is evident that network
reconfiguration reduces not only reliability costs, but cost of losses and loss-of-lives
as well. The latter indicates the possibility to defer asset replacement.
Table 6-5: Total ‘operation’ costs for two strategies
Applied DG integration Costs (£)

technique and Network Test System Loss of life Copper losses load curtailment
33 bus £10,275.90 £13,837.20 £28,232.13
New approach
69 bus £3,023.01 £9,416.62 £20,680.27
33 bus £8,258.06 £10,975.23 £22,251.25
New approach with optimal NR
69 bus £2,193.68 £7,710.51 £13,232.08
Ranking of cable sections shown in Table 6-6 is based on the historic cable thermal
loss-of-lives. Additional loss-of-lives in the planning period are also presented for the
three network management strategies. The conventional DG connection gives
additional loss-of-lives in descending order which is in line with the historic ranking.
In case of the other two strategies, a perturbation in the ranking order is experienced;
however, the same sections were in the top part of the list.
Table 6-6: Cable ranking based on thermal loss-of-life

Applied SG integration technique
New approach with optimal
Coventional New approach NR
total total total
System
Historic incremental spent incremental spent incremental spent

Line_id period life spent life life spent life life spent life
Line 5 0.932 Line 5 0.087 1.019 Line 2 0.068 1.000 Line 5 0.048 0.980
33 bus
69 bus
160
___________________________________________________________________________________
Table 6-6 also shows that the total loss-of-life (‘historic’ plus ‘conventional’ column)
of cable sections Line 5, Line 1, Line 2, and Line 4 in the 33-bus network is greater
than unity, which means they need to be replaced in the planning period. However, if
either the second or the third strategy is applied, their replacement is deferred for the
next regulatory period. It can also be noticed that lower loss-of-lives in the planning
period are obtained when the new approach with optimal NR is used for restoration of
supplies during emergencies. The third strategy delivers the best results in terms of
system reliability performance (Table 6-4) and cable loss-of-lives (Table 6-6).
Table 6-7 presents results of cable ranking based on the new approach with optimal
NR. Two case studies are investigated: the first makes use of the cost criterion (5-1),
while the second is based on the total cost which includes the cost of cable loss-of-
lives (5-2). Differences in costs for these two ranking schemes are not that big,
because the last term in (5-2) represents the cost of additional ageing with respect to
the pre-fault condition. We have obtained almost the same ranking orders of cable
sections using the criteria (5-1) and (5-2); they are however different from the ranking
based on cable loss-of-lives Table 6-6. This is an important conclusion particularly for
companies that do not have a fixed replacement budget.
Table 6-7: Cable ranking based on: a) ‘Direct’ Cost, b) ‘Total’ Cost
System COST1 COST2

Line 3 £42,787.97 Line 3 £44,167.07
Line 5 £22,412.08 Line 5 £23,926.79
Line 2 £17,330.69 Line 2 £18,841.28
Line 4 £16,650.86 Line 4 £18,024.77
33 bus Line 7 £14,557.19 Line 7 £15,889.56
Line 26 £14,442.10 Line 26 £15,743.78
Line 1 £11,698.13 Line 1 £13,013.30
Line 8 £9,363.90 Line 8 £10,686.85
£149,242.92 £160,293.41
Line 6 £15,219.81 Line 6 £16,524.44
Line 5 £8,295.45 Line 5 £9,600.71
69 bus Line 40 £3,827.58 Line 40 £5,128.59
Line 48 £1,854.31 Line 48 £3,153.79
£29,197.15 £34,407.53
The results presented in Table 6-3 – Table 6-7 show that the choice of the new DG
connection points, network operation strategy, as well as budget constraints need to be
161
___________________________________________________________________________________
carefully considered when making investment decisions. Furthermore, the findings of

these studies suggest that the effective connection of wind-DG units and prioritization
of distribution cables for replacement is best done when system impacts are
considered in conjunction with asset stress conditions (i.e. asset ageing). These
conclusions indicate that the existing real-life practice of cable replacement based on
individual assets only will need to be replaced with planning techniques that consider
the overall system.
6.8 Chapter summary
This chapter presents main concepts related to ‘smart’ distribution networks. Smart
distribution networks are future networks with integrated low carbon technologies and
applied advanced technological solutions (i.e. ‘smart’ solutions) to operate and control
the network. The first part of the chapter addressed optimal integration of wind
generation onto distribution networks. A novel optimization model is developed; its
objective is to minimize connection cost as well as the cost of spent loss-of-lives of
ageing cable. In the second part of the chapter, a smart solution, dynamic network
reconfiguration, is applied as a real-time control action in contingent situations. The
proposed objective is again to reduce cable ageing and postpone replacement. This
objective was further extended to model the total hourly costs expressed in terms of
loss-of-life, losses, and reliability costs. In order to quantify the system benefits of the
new approach; both developed models are incorporated into the probabilistic
simulation procedure. The next chapter provides further details of the probabilistic
technique used for the simulation of the distribution network operation.
This chapter provided details of the proposed simulation methodology for distribution
network replacement planning in the presence of wind-DG and network
reconfiguration. Importance of the proposed models lies in the need to manage ageing
network infrastructure and the possibility to apply modern solutions to extend the
asset (cable) lives. The proposed research work can provide basis for the planning of
cable replacements, as well as the impact of new low carbon technologies and smart
solutions on the planning outcomes. The next section presents the input test systems,
results, and discussion of findings from the computer simulation of the methodologies
presented in this chapter.
162
___________________________________________________________________________________
Using these data and the simulation procedures described in chapter 6.5 several sets of
results were generated. The two main study cases are presented: the first involves
investigation of the impact of ageing distribution cable on distribution system
reliability and cable ranking for replacement, whilst the second deals with the optimal
management of future distribution systems with integrated wind-DG and dynamic
network reconfiguration and the impact on cable replacement planning.
This chapter presented the results and various discussions to affirm the successful
application of the developed cable replacement models carried out on two medium
voltage test systems. The results presented include the calculated Arrhenius model
parameters, system wide indices, as well as sensitivity studies on varying Weibull
stress parameter. Several reliability assessments were carried out for different network
management schemes, and the presented discussions are related to system wide
indices for the schemes, sensitivity studies on varying Weibull stress parameter,
comparison of the conventional and proposed ranking schemes, and the ARMA model
results used for predicting the DG-wind output.
From the analysis of the simulation results, it can be generally concluded that use of
relevant asset data, asset condition assessment, and optimal network management can
reduce replacement costs, extend asset lifetimes, and improve system reliability.
163
CHAPTER SEVEN
CONCLUSIONS AND FUTURE WORK
7 Conclusions
This thesis studied the influence of end-of-life failures of underground cables on

distribution system reliability in order to make optimal decisions on their replacement.
The research was motivated by the rising number of increasingly ageing cable assets
in power distribution networks in addition to current price control schemes imposed
by electricity regulators. Distribution business regulating regime constrains the assets
replacement expenditure budget, whereas ageing cables reduce the expected levels of
reliability and quality of supply in distribution systems. This implies that the balance
between investment costs and reliability needs to be managed carefully. Underground
cables are selected for the analysis because they are very expensive to install and
disruptive to replace. The analyses conducted are particularly beneficial for planners,
asset managers and operators. The studies assist in identifying the most critical cables
for system reliability and in formulating cost-effective replacement plans.
The thesis focused on end-of-life failure of distribution systems underground cables,

provided explanation of its concept and highlighted differences with the conventional
repairable failure of underground cables. Based on this, literature review on the
methods used in evaluating end-of-life was utilised in the development of a novel
method. By distinguishing among types of failure, the thesis enhanced the
understanding of components’ age related failure and stimulated researchers to focus
on methods and models that are related to the impact of components ageing on
replacement planning. This is the first contribution of this thesis.
In the reliability assessments of power system, time varying renewal process

represents a component failure and repair cycle. It is a common assumption that the
repair phase is perfect, meaning the component is ꞌnormallyꞌ restored to its original
state after repairs. However, for most network components, the renewal process is
associated with an ageing trend influenced by the system dynamic loading conditions
and its physical characteristics. This thesis describes a new method for modelling the
components aging and assessing the impacts on (medium voltage) distribution
164
Chapter seven: Conclusion and future work
___________________________________________________________________________________
network reliability that is expressed through ‘appropriate’ reliability indices. The

thesis proposed a new reliability model and a prioritization method for ageing
distribution cables. The reliability model integrates the cable thermal model and
Arrhenius ageing model into the Weibull distribution. A novel estimation method for
the calculation of Arrhenius parameters was proposed. The reliability model is then
incorporated into the developed sequential Monte Carlo simulation (SMCS)
procedure.
The proposed IEC-Arrhenius-Weibull model incorporated into the SMCS technique

was applied on two medium voltage distribution networks. It was shown that ageing
models have significant (negative) impact on network reliability. This also means that
ignoring components physical properties and ageing in the reliability assessments will
result in gross underestimation of system risks during planning of future energy
networks. This developed framework is the second main contribution of this thesis.
The problem of ageing assets in a system creates a concern and makes the system
vulnerable to multiple failure events. Thus, the proposed IEC-Arrhenius-Weibull
model incorporated into the SMCS technique thesis is adopted in order to investigate
the impact of individual cable sections on system reliability and to develop a
financially based cable ranking scheme for replacement planning. Prioritization of
cable using the financial scheme that considers system-wide effects can give different
results compared to the schemes based on isolated assets (e.g. thermal loss-of-life).
This developed financially based cable ranking scheme was utilised to prioritize
individual cable(s) due for replacement and it is the third contribution of this thesis.
This thesis further investigated a new concept of DG integration and distribution

network management whose goal is to reduce distribution cable ageing and postpone
its replacement. The proposed methodology is based on modelling the cable historic
loss-of-life, physical characteristics of network components, envisaged network
operating conditions, and different network management strategies. Two new
optimization models are developed, one for optimal DG connection and one for
optimal network reconfiguration, and they are incorporated within the probabilistic
methodology developed to analyse regulatory planning periods. The results show that
simultaneous consideration of individual asset characteristics and network aspects
will give the best financial rewards in terms of extended assets lives, increased
165
___________________________________________________________________________________
network reliability and reduced operation costs. To this end, closer co-operation
between replacement and planning teams in utilities will need to be established. The
developed optimization models constitute the fourth contribution of this thesis.
7.1 Future work
Future studies could focus on assessing the impact of other low carbon technologies
(such as solar, storage, electric vehicle, etc.) and new control concepts, such as co-
ordinated voltage control, demand side management - DSM etc. to extend asset lives
and improve the quality of supply – network reliability. The final goal is to develop a
methodology that will provide a single prioritization list of assets for both
replacement and reinforcement purposes.
Transient-state power flow studies could be investigated on the effect of potential

switching transients on other equipment such as cables and transformers as well as the
impact of temporary overloading (thermal effects) on network elements when
implementing the proposed dynamic reconfiguration approach.
For a more realistic assessment and appreciation of the presented technique, the
proposed methodology could be tested on larger real life distribution networks;
analysis involving all three hierarchical levels (HLI, HLII; and HLIII) and; different
mix of cable types (such as OHL, UGC, installation methods etc.) could be analysed.
SMCS analysis incorporating different ageing models for the different network
elements could also be explored for improved accuracy of network performance
evaluation.
An advanced approach could be to analyse the whole component lifetime as described

by the bathtub curve and consideration given to all the three types of repair processes.
Furthermore, customer average interruption duration index - CAIDI and other similar
indices could be used to evaluate the network performance. This would add depth to
the discussion of average restoration time and cost compensation (due on the DNO)
for affected customers and for increased accuracy of cable ranking. In addition,
consideration of customer (load) distinction can be introduced in the problem
formulation in order to add depth to the analysis of ‘important’ network branches.
166
___________________________________________________________________________________
Non-parametric approaches (such as multiple adaptive regression splines, neural

networks, and so on) could be used to develop components failure functions in order
to capture the finer details (increase accuracy) of component ageing behaviours.
Finally, parallel computational techniques could be utilised to increase the
computational efficiency of the reliability assessment tool for larger power networks.
167
APPENDIX
A 1: Lines and load demand data for IEEE 33-bus radial distribution system
Lines Bus inter- Resistance Reactance Active Power Reactive Power

ID connection () () (kW) (kVAR)
L1 1&2 0.0922 0.047 100 60
L2 2&3 0.493 0.2511 90 40
L3 3&4 0.366 0.1864 120 80
L4 4&5 0.3811 0.1941 60 30
L5 5&6 0.819 0.707 60 20
L6 6&7 0.1872 0.6188 200 100
L7 7&8 0.7114 0.2351 200 100
L8 8&9 1.03 0.74 60 20
L9 9 & 10 1.044 0.74 60 20
L10 10 & 11 0.1966 0.065 45 30
L11 11 & 12 0.3744 0.1238 60 35
L12 12 & 13 1.468 1.155 60 35
L13 13 & 14 0.5416 0.7129 120 80
L14 14 & 15 0.591 0.526 60 10
L15 15 & 16 0.7463 0.545 60 20
L16 16 & 17 1.289 1.721 60 20
L17 17 & 18 0.732 0.574 90 40
L18 2 & 19 0.164 0.1565 90 40
L19 19 & 20 1.5042 1.3554 90 40
L20 20 & 21 0.4095 0.4784 90 40
L21 21 & 22 0.7089 0.9373 90 40
L22 3 & 23 0.4512 0.3083 90 50
L23 23 & 24 0.898 0.7091 420 200
L24 24 & 25 0.896 0.7011 420 200
L25 6 & 26 0.203 0.1034 60 25
L26 26 & 27 0.2842 0.1447 60 25
L27 27 & 28 1.059 0.9337 60 20
L28 28 & 29 0.8042 0.7006 120 70
L29 29 & 30 0.5075 0.2585 200 600
L30 30 & 31 0.9744 0.963 150 70
L31 31 & 32 0.3105 0.3619 210 100
L32 32 & 33 0.341 0.5302 60 40
L33 7 & 20 2 0.4132 no-load no-load
L34 8 & 14 2 0.4132 no-load no-load
L35 11 & 21 2 0.4132 no-load no-load
168
Appendix & references
___________________________________________________________________________________
Lines Bus inter- Resistance Reactance Active Power Reactive Power

ID connection () () (kW) (kVAR)
L36 17 & 32 0.5 0.1033 no-load no-load
L37 24 & 28 0.5 0.1033 no-load no-load
A 2: Lines and load demand data for IEEE 69-bus radial distribution system
Reactive
Resistance Reactance Active Power
Lines ID From To Power
() () (kW )
(kVAR )
L1 1 2 0.0003 0.0007 0 0
L2 2 3 0.0003 0.0007 0 0
L3 3 4 0.0009 0.0022 0 0
L4 4 5 0.0157 0.0183 0 0
L5 5 6 0.2284 0.1162 0 0
L6 6 7 0.2377 0.1211 2.6 2.2
L7 7 8 0.0575 0.0293 40.4 30
L8 8 9 0.0308 0.0157 75 54
L9 9 10 0.511 0.1689 30 22
L10 10 11 0.1168 0.0386 28 19
L11 11 12 0.4439 0.1467 145 104
L12 12 13 0.6426 0.2121 145 104
L13 13 14 0.6514 0.2153 8 5
L14 14 15 0.6601 0.2181 8 5
L15 15 16 0.1227 0.0406 0 0
L16 16 17 0.2336 0.0772 45.5 30
L17 17 18 0.0029 0.001 60 35
L18 18 19 0.2044 0.0676 60 35
L19 19 20 0.1314 0.0431 0 0
L20 20 21 0.2131 0.0704 1 0.6
L21 21 22 0.0087 0.0029 114 8.1
L22 22 23 0.0993 0.0328 5 3.5
L23 23 24 0.2161 0.0714 0 0
L24 24 25 0.4672 0.1544 28 20
L25 25 26 0.1927 0.0637 0 0
L26 26 27 0.1081 0.0357 14 10
L27 3 28 0.0027 0.0067 14 10
L28 28 29 0.0399 0.0976 26 18.6
L29 29 30 0.2482 0.082 26 18.6
L30 30 31 0.0438 0.0145 0 0
L31 31 32 0.219 0.0724 0 0
L32 32 33 0.5235 0.1757 0 0
169
___________________________________________________________________________________
Reactive
() () (kW )
(kVAR )
L33 33 34 1.0657 0.1651 14 10
L34 34 35 0.9197 0.304 19.5 14
L35 3 36 0.0027 0.0067 6 4
L36 36 37 0.0399 0.0976 26 18.55
L37 37 38 0.0657 0.0767 26 18.55
L38 38 39 0.019 0.0221 0 0
L39 39 40 0.0011 0.0013 24 17
L40 40 41 0.4544 0.5309 24 17
L41 41 42 0.1934 0.226 1.2 1
L42 42 43 0.0256 0.0298 0 0
L43 43 44 0.0057 0.0072 6 4.3
L44 44 45 0.0679 0.0857 0 0
L45 45 46 0.0006 0.0007 39.22 26.3
L46 4 47 0.0021 0.0052 39.22 26.3
L47 47 48 0.0531 0.013 0 0
L48 48 49 0.1808 0.4424 79 56.4
L49 49 50 0.0513 0.1255 384.7 274.5
L50 8 51 0.0579 0.0295 384.7 274.5
L51 51 52 0.2071 0.0695 40.5 28.3
L52 9 53 0.1086 0.0553 3.6 2.7
L53 53 54 0.1267 0.0645 4.35 3.5
L54 54 55 0.1773 0.0903 26.4 19
L55 55 56 0.1755 0.0894 24 17.2
L56 56 57 0.992 0.333 0 0
L57 57 58 0.489 0.1641 0 0
L58 58 59 0.1898 0.0628 0 0
L59 59 60 0.2409 0.0731 100 72
L60 60 61 0.3166 0.1613 0 0
L61 61 62 0.0608 0.0309 1244 888
L62 62 63 0.0905 0.046 32 23
L63 63 64 0.4433 0.2258 0 0
L64 64 65 0.6495 0.3308 227 162
L65 11 66 0.1255 0.0381 59 42
L66 66 67 0.0029 0.0009 18 13
L67 12 68 0.4613 0.1525 18 13
L68 68 69 0.0029 0.001 28 20
L69 11 43 0.312 0.312 28 20
L70 13 21 0.312 0.312 0 0
170
___________________________________________________________________________________
Reactive
() () (kW )
(kVAR )
L71 15 46 0.6239 0.312 0 0
L72 50 59 1.2479 0.6239 0 0
L73 27 65 0.6239 0.312 0 0
A 3: Hourly peak loads as percent of daily peak
Hourly% Winter Summer Spring/Fall

Hour Week day Week end Week day Week end Week day Week end
1am 67 78 64 74 63 75
2am 63 72 60 70 62 73
3am 60 68 58 66 60 69
4am 59 66 56 65 58 66
5am 59 64 56 64 59 65
6am 60 65 58 62 65 65
7am 74 66 64 62 72 68
8am 86 70 76 66 85 74
9am 95 80 87 81 95 83
10am 96 88 95 86 99 89
11am 96 90 99 91 100 92
12am 95 91 100 93 99 94
1pm 95 90 99 93 93 91
2pm 95 88 100 92 92 90
3pm 93 87 100 91 90 90
4pm 94 87 97 91 88 86
5pm 99 91 95 92 90 85
6pm 100 100 95 94 92 88
7pm 100 99 93 95 96 92
8pm 96 97 92 95 98 100
9pm 91 94 92 100 96 97
10pm 83 92 93 93 90 95
11pm 73 87 87 88 80 90
12pm 63 81 72 80 70 85
A 4: Daily peak load as percent of weekly peak
Week day Monday Tuesday Wednesday Thursday Friday Saturday Sunday

% 93 100 98 96 94 77 75
A 5: Weekly peak load as percent of annual peak
Week % Week % Week % Week %

1 86.2 14 75 27 75.5 40 72.4
2 90 15 72.1 28 81.6 41 74.3
171
___________________________________________________________________________________
Week % Week % Week % Week %

3 87.8 16 80 29 80.1 42 74.4
4 83.4 17 75.4 30 88 43 80
5 88 18 83.7 31 72.2 44 88.1
6 84.1 19 87 32 77.6 45 88.5
7 83.2 20 88 33 80 46 90.9
8 80.6 21 85.6 34 72.9 47 94
9 74 22 81.1 35 72.6 48 89
10 73.7 23 90 36 70.5 49 94.2
11 71.5 24 88.7 37 78 50 97
12 72.7 25 89.6 38 69.5 51 100
13 70.4 26 86.1 39 72.4 52 95.2
172
173
___________________________________________________________________________________
REFERENCES
[1] C. o. E. E. Regulators, "5th Benchmarking Report on Quality of Electricity

Supply," Brussels2011.
[2] R. Billinton and W. Li, Reliability Assessment of Electrical Power Systems

Using Monte Carlo Methods: Springer, 1994.
[3] R. N. Allan, Billinton, R., Reliability Evaluation of Power Systems: Springer,

1996.
[4] R. E. Brown, Electric Power Distribution Reliability, Second Edition: Taylor

& Francis, 2008.
[5] A. Chowdhury and D. Koval, Power Distribution System Reliability: Practical

Methods and Applications: Wiley, 2011.
[6] CIGRE, "WG 37.27: Aging of the System – Impact on Planning," in Statistics:
Overhead Lines and Underground Cables vol. 176, ed. Paris: CIGRE, 2000.
[7] R. N. Allan, R. Billinton, A. M. Breipohl, and C. H. Grigg, "Bibliography on

the Application of Probability Methods in Power System Reliability
Evaluation: 1987-1991," IEEE Transactions on Power Systems, vol. 9, pp. 41-
49, 1994.
[8] R. N. Allan, R. Billinton, and S. H. Lee, "Bibliography of the Application of

Probability Methods in Power System Reliability Evaluation 1977-1982,"
IEEE Power Engineering Review, vol. PER-4, pp. 24-25, 1984.
[9] R. Billinton, M. Fotuhi-Firuzabad, and L. Bertling, "Bibliography on the

Application of Probability Methods in Power System Reliability Evaluation
1996-1999," IEEE Transactions on Power Systems, vol. 16, pp. 595-602,
2001.
[10] IEEE, "Bibliography on the Application of Probability Methods in Power

System Reliability Evaluation 1971-1977," IEEE Transactions on Power
Apparatus and Systems, vol. PAS-97, pp. 2235-2242, 1978.
[11] W. Li, Risk Assessment Of Power Systems: Models, Methods, and

Applications, illustrated ed.: John Wiley & Sons, 2005.
[12] R. Biillinton and R. Allan, "Basic Power System Reliability Concepts,"

Reliability Engineering & System Safety, vol. 27, pp. 365-387, 1990.
[13] R. Billinton, Allan, R.N., Reliability Assessment of Large Electric Power

Systems, illustrated ed.: Springer, 1988.
[14] T. K. Vrana and E. Johansson, "Overview of Power System Reliability

Assessment Techniques," Cigre, Paris2011.
174
___________________________________________________________________________________
[15] G. T. Heydt, "The Next Generation of Power Distribution Systems," IEEE

Transactions on Smart Grid, vol. 1, pp. 225-235, 2010.
[16] IEEE, "IEEE Std 1366, IEEE Guide for Electric Power Distribution Reliability
Indices," in IEEE Guide for Electric Power Distribution Reliability Indices,
ed. New York, USA: IEEE Transmission and Distribution Committee, 2012.
[17] R. Allan and R. Billinton, "Power System Reliability and its Assessment. III.
Distribution Systems and Economic Considerations," Power Engineering
Journal, vol. 7, pp. 185-192, 1993.
[18] R. Y. Rubinstein, Dirk, P. K., Simulation and The Monte Carlo Method:
Wiley, 2007.
[19] L. Wenyuan, "Incorporating Aging Failures in Power System Reliability

Evaluation," IEEE Transactions on Power Systems, vol. 17, pp. 918-923,
2002.
[20] R. Billinton and R. N. Allan, Reliability Evaluation of Engineering Systems:

Concepts and Techniques: Springer, 1992.
[21] A. M. Leite da Silva, A. M. Cassula, L. C. Nascimento, J. C. Freire, C. E.

Sacramento, and A. C. R. Guimaraes, "Chronological Monte Carlo-Based
Assessment of Distribution System Reliability," in International Conference
on Probabilistic Methods Applied to Power Systems (PMAPS), 2006, pp. 1-7.
[22] R. Billinton and W. Li, "Composite System Reliability Assessment Using a

Monte Carlo Approach," in Third International Conference on Probabilistic
Methods Applied to Electric Power Systems (PMAPS), 1991, pp. 53-57.
[23] G. J. Anders, J. Endrenyi, and C. Yung, "Risk-based Planner for Asset

management [of electric utilities]," IEEE Computer Applications in Power,
vol. 14, pp. 20-26, 2001.
[24] H. L. Willis and R. R. Schrieber, Aging Power Delivery Infrastructures,

Second Edition: CRC Press, 2013.
[25] H. L. Willis and L. Philipson, Understanding Electric Utilities and De-

Regulation: CRC Press, 2005.
[26] R. E. Brown and L. A. A. Freeman, "Analyzing the Reliability Impact of

Distributed Generation," in Power Engineering Society Summer Meeting,
2001, pp. 1013-1018 vol.2.
[27] L. Goel and R. Billinton, "Determination of Reliability Worth for Distribution

System Planning," IEEE Transactions on Power Delivery, vol. 9, pp. 1577-
1583, 1994.
[28] A. Keane, L. F. Ochoa, C. L. T. Borges, G. W. Ault, A. D. Alarcon-

Rodriguez, R. A. F. Currie, et al., "State-of-the-Art Techniques and
175
___________________________________________________________________________________
Challenges Ahead for Distributed Generation Planning and Optimization,"

IEEE Transactions on Power Systems, vol. 28, pp. 1493-1502, May 2013.
[29] T.-N. Preda, K. Uhlen, and D. E. Nordgard, "An overview of the present grid
codes for integration of distributed generation," in Workshop on Integration of
Renewables into the Distribution Grid (CIRED), 2012, pp. 1-4.
[30] E. L. Brancato, "Insulation Aging a Historical and Critical Review," IEEE

Transactions on Electrical Insulation, vol. EI-13, pp. 308-317, 1978.
[31] G. Mazzanti, "The Combination of Electro-thermal Stress, Load Cycling and

Thermal Transients and its Effects on the Life of High Voltage ac Cables,"
IEEE Transactions on Dielectrics and Electrical Insulation, vol. 16, pp. 1168-
1179, Aug 2009.
[32] S. B. Dalal, R. S. Gorur, and M. L. Dyer, "Aging of Distribution Cables in

Service and its Simulation in the Laboratory," IEEE Transactions on
Dielectrics and Electrical Insulation, vol. 12, pp. 139-146, 2005.
[33] C. Holyk and G. J. Anders, "Power Cable Rating Calculations - A Historical

Perspective," IEEE Industry Applications Magazine, vol. 21, pp. 6-64, 2015.
[34] IEC, "Electric Cables - Calculations for Current Ratings - Finite Element
Method: IEC TR 62095:2003 " in IEC TR 62095:2003 ed. Switzerland: IEC,
2003, p. 69.
[35] S. K. E. Awadallah, J. V. Milanovic, and P. N. Jarman, "The Influence of

Modeling Transformer Age Related Failures on System Reliability," IEEE
Transactions on Power Systems, vol. 30, pp. 970-979, 2015.
[36] M. Buhari, V. Levi, and S. K. E. Awadallah, "Modelling of Ageing

Distribution Cable for Replacement Planning," IEEE Transactions on Power
Systems, vol. DOI: 10.1109/TPWRS.2015.2499269, pp. 1-9, 2015.
[37] R. Olsen, G. J. Anders, J. Holboell, and U. S. Gudmundsdottir, "Modelling of

Dynamic Transmission Cable Temperature Considering Soil-Specific Heat,
Thermal Resistivity, and Precipitation," IEEE Transactions on Power
Delivery, vol. 28, pp. 1909-1917, 2013.
[38] A. Kapetanaki, K. Kopsidas, C. Tumelo-Chakonta, and M. Buhari, "Network

planning evaluation implementing time varying thermal ratings," in
International Conference on Probabilistic Methods Applied to Power Systems
(PMAPS), 2014, pp. 1-6.
[39] R. Mackinlay, "Managing an Aging Underground High-Voltage Cable

Network," Power Engineering Journal, vol. 4, pp. 271-277, 1990.
[40] F. M. George, Electric Cables Handbook, illustrated ed.: Wiley, 1997.
176
___________________________________________________________________________________
[41] G. Anders, J. M. Braun, M. Vainberg, S. Rizzetto, and H. Brakelmann,

"Rating of Cables in a Non-uniform Thermal Environment," in IEEE
Transmission and Distribution Conference, 1999, pp. 83-88 vol.1.
[42] J. B. Pegram, H. E. Orton, and R. Samm, "Worldwide Underground

Distribution Cable Practices," in IEEE Transmission and Distribution
Conference, 1999, pp. 62-67 vol.1.
[43] M. Stotzel, M. Zdrallek, and W. H. Wellssow, "Reliability Calculation of MV-

Distribution Networks with Regard to Ageing in XLPE-Insulated Cables," IEE
Proceedings of Generation, Transmission and Distribution, vol. 148, pp. 597-
602, 2001.
[44] B. I. British Standards, "Electric Cables - Calculation of the current rating, BS

IEC 60287-1-1:2006," in Calculation of current rating, ed. London: BSI,
2006.
[45] M. B. Schiffer, Power Struggles: Scientific Authority and the Creation of

Practical Electricity Before Edison: MIT Press, 2008.
[46] A. E. Kennelly, "On the Carrying Capacity of Electric Cables, Submerged,

Buried or Suspended in Air," in The 9th Annual Meeting, Association Edison
Illuminating Companies, 1893, p. 14.
[47] G. Mie, "Über die Wärmeleitung in Einem Verteilten Kabel,"

Elektrotechnische Zeitschrift, vol. 26, p. 7, 1905.
[48] S. Melsom and H. Booth, "The Heating of Cables with Current," Institute of
Electrical Engineering, vol. 47, p. 25, 1911.
[49] R. W. Atkinson and H. Fisher, "Current Rating of Electric Cables,"

Transactions of American Institute of Electrical Engineers, vol. 33, p. 7, 1913.
[50] J. H. Neher and M. H. McGrath, "The Calculation of the Temperature Rise

and Load Capability of Cable Systems," Transactions of the American
Institute of Electrical Engineers, Power Apparatus and Systems, Part III, vol.
76, pp. 752-764, 1957.
[51] G. J. Anders, M. A. El-Kady, R. W. D. Ganton, D. J. Horrocks, and J. Motlis,

"Calculations of Power Cable Load Capability on a Desktop Computer: A
Review," IEE Proceedings Generation, Transmission and Distribution, vol.
133, pp. 431-436, 1986.
[52] G. J. L. Poidevin, "Manual for CRATER Current Rating Software for

Dynamic Ratings (v1.0a)," UK Report No: 6110, 2008.
[53] CEI/IEC, "Calculation of the Cyclic and Emergency Current Rating of Cables;
Part 2: Cyclic Rating of Cables greater than 18/30 (36) kV and Emergency
Ratings for Cables of all Voltages," vol. 60853-2:1989, ed: IEC, 1989.
177
___________________________________________________________________________________
[54] IEEE, "IEEE Standard Power Cable Ampacity Tables," IEEE Std 835-1994,
pp. 1-3151, 1994.
[55] IEEE, "IEEE Standard Power Cable Ampacity Tables Amendment 1: Revision
to Introduction," IEEE Std 835a-2012 (Amendment to IEEE Std 835-1994), pp.
1-16, 2013.
[56] R. S. Olsen, J. Holboll, and U. S. Gudmundsdottir, "Dynamic Temperature

Estimation and Real Time Emergency Rating of Transmission Cables," in
IEEE Power and Energy Society General Meeting, 2012, pp. 1-8.
[57] F. de Leon and G. J. Anders, "Effects of Backfilling on Cable Ampacity

Analyzed With the Finite Element Method," IEEE Transactions on Power
Delivery, vol. 23, pp. 537-543, 2008.
[58] J. A. Pilgrim, D. J. Swaffield, P. L. Lewin, and D. Payne, "Rating Methods for

Cables Installed in Unventilated and Partially Ventilated Surface Troughs," in
Electrical Insulation Conference (EIC), 2011, pp. 74-78.
[59] J. Kolczynski, W. Tylman, and G. J. Anders, "Application of Continuous

Probabilistic Networks to Leak Detection in Pipe-type Cable Installations," in
IEEE 11th International Conference on Probabilistic Methods Applied to
Power Systems (PMAPS), 2010, pp. 448-453.
[60] M. Buhari, K. Kopsidas, C. Tumelo-Chakonta, and A. Kapetanaki, "Risk

Assessment of Smart Energy Transfer in Distribution Networks," in IEEE PES
Innovative Smart Grid Technologies (ISGT), 2014, pp. 1-6.
[61] M. Buhari and K. Kopsidas, "Probabilistic Assessment of Impacts of Real-

Time Line Ratings on Distribution Networks," in International Conference on
Probabilistic Methods Applied to Power Systems (PMAPS), 2014, pp. 1-6.
[62] L. Song-Bo, "Calculation of the Steady-State and Transient Temperature Rises

of Round Cable Bundles," IEEE Transactions on Power Delivery, vol. 25, pp.
1229-1235, 2010.
[63] G. Mazzanti, "Analysis of the Combined Effects of Load Cycling, Thermal

Transients, and Electrothermal Stress on Life Expectancy of High-Voltage AC
Cables," IEEE Transactions on Power Delivery, vol. 22, pp. 2000-2009, 2007.
[64] H. Endicott, B. Hatch, and R. Sohmer, "Application of the Eyring Model to

Capacitor Aging Data," IEEE Transactions on Component Parts, vol. 12, pp.
34-41, 1965.
[65] P. H. G. Alleb and A. Tustin, "The Aging Process in Electrical Insulation: A

Tutorial Summary," IEEE Transactions on Electrical Insulation, vol. EI-7, pp.
153-157, 1972.
[66] M. A. Miner, "Cumulative Damage in Fatigue," Journal of Applied

Mechanics-Transactions of the Asme, vol. 12, pp. A159-A164, 1945.
178
___________________________________________________________________________________
[67] M. R. Yenchek and P. G. Kovalchik, "The Impact of Current Load on Mine-

Trailing Cable Thermal Life," IEEE Transactions on Industry Applications,
vol. 29, pp. 762-767, Jul-Aug 1993.
[68] K. Hagkwen and C. Singh, "Reliability Modeling and Simulation in Power

Systems With Aging Characteristics," IEEE Transactions on Power Systems,
vol. 25, pp. 21-28, 2010.
[69] S. Asgarpoor and M. J. Mathine, "Reliability Evaluation of Distribution

Systems with Non-Exponential Down Times," IEEE Transactions on Power
Systems, vol. 12, pp. 579-584, 1997.
[70] C. Singh and R. Billinton, System Reliability, Modelling and Evaluation:

Hutchinson, 1977.
[71] L. Wenyuan, E. Vaahedi, and P. Choudhury, "Power system Equipment

Aging," IEEE Power and Energy Magazine, vol. 4, pp. 52-58, 2006.
[72] L. Wenyuan, Z. Jiaqi, X. Kaigui, and X. Xiaofu, "Power System Risk

Assessment Using a Hybrid Method of Fuzzy Set and Monte Carlo
Simulation," IEEE Transactions on Power Systems, vol. 23, pp. 336-343,
2008.
[73] K. Xie and W. Li, "Analytical Model for Unavailability due to Aging Failures
in Power Systems," International Journal of Electrical Power & Energy
Systems, vol. 31, pp. 345-350, 9// 2009.
[74] J. Qiu, H. F. Wang, D. Y. Lin, B. T. He, W. F. Zhao, and W. Xu,

"Nonparametric Regression-Based Failure Rate Model for Electric Power
Equipment Using Lifecycle Data," IEEE Transactions on Smart Grid, vol. 6,
pp. 955-964, Mar 2015.
[75] M. T. Schilling, J. C. G. Praca, J. F. De Queiroz, C. Singh, and H. Ascher,

"Detection of Ageing in the Reliability Analysis of Thermal Generators,"
IEEE Transactions on Power Systems, vol. 3, pp. 490-499, 1988.
[76] M. Jong-Fil, K. Jea-Chul, L. Hee-Tae, L. Sang-Seung, Y. Yong-Tae, and S.

Kyung-Bin, "Time-Varying Failure Rate Extraction in Electric Power
Distribution Equipment," in International Conference on Probabilistic
Methods Applied to Power Systems (PMAPS), 2006, pp. 1-6.
[77] M. T. Schilling, M. B. Do Coutto Filho, A. M. Leite da Silva, R. Billinton, and

R. N. Allan, "An Integrated Approach to Power System Reliability
Assessment," International Journal of Electrical Power & Energy Systems,
vol. 17, pp. 381-390, 12// 1995.
[78] W. Peng and D. W. Coit, "Repairable Systems Reliability Trend Tests and
Evaluation," in Proceedings of Annual Reliability and Maintainability
Symposium, 2005, pp. 416-421.
179
___________________________________________________________________________________
[79] M. Jong-Fil, K. Jae-Chul, L. Hee-Tae, P. Chang-Ho, Y. Sang-Yun, and L.

Sang-Seung, "Reliability Evaluation of Distribution System Through the
Analysis of Time-Varying Failure Rate," in IEEE Power Engineering Society
General Meeting, 2004, pp. 668-673 Vol.1.
[80] A. Lisnianski and I. Frenkel, Recent Advances in System Reliability:

Signatures, Multi-state Systems and Statistical Inference: Springer-Verlag
London Limited, 2011.
[81] S. K. E. Awadallah, "Probabilistic Methodology for Prioritising Replacement

of Ageing Power Transformers Based on Reliability Assessment of
Transmission System," Doctorate, School of Electrical and Electronic
Engineering, The University of Manchester, UK, Manchester, 2014.
[82] A. Hoyland and M. Rausand, System Reliability Theory: Models and

Statistical Methods: Wiley, 2009.
[83] R. M. Bucci, R. V. Rebbapragada, A. J. Mcelroy, E. A. Chebli, and S. Driller,

"Failure Prediction of Underground Distribution Feeder Cables," IEEE
Transactions on Power Delivery, vol. 9, pp. 1943-1955, Oct 1994.
[84] R. E. Brown, G. Frimpong, and H. L. Willis, "Failure Rate Modeling Using

Equipment Inspection Data," IEEE Transactions on Power Systems, vol. 19,
pp. 782-787, 2004.
[85] H. Rinne, The Weibull Distribution: A Handbook: CRC Press, 2008.
[86] L. H. Crow, "Evaluating the Reliability of Repairable Systems," in

Proceedings of Annual Reliability and Maintainability Symposium, 1990, pp.
275-279.
[87] M. Kijima and U. Sumita, "A Useful Generalization of Renewal Theory:

Counting Processes Governed by Non-Negative Markovian Increments,"
Journal of Applied Probability, vol. 23, p. 71, 1986.
[88] C. Grigg, P. Wong, P. Albrecht, R. Allan, M. Bhavaraju, R. Billinton, et al.,

"The IEEE Reliability Test System-1996. A Report Prepared by the Reliability
Test System Task Force of the Application of Probability Methods
Subcommittee," IEEE Transactions on Power Systems, vol. 14, pp. 1010-
1020, 1999.
[89] M. E. Baran and F. F. Wu, "Network Reconfiguration in Distribution-Systems

for Loss Reduction and Load Balancing," IEEE Transactions on Power
Delivery, vol. 4, pp. 1401-1407, Apr 1989.
[90] ENA, "National Fault and Interruption Reporting Scheme - National System
and Equipment Performance," in NaFIRS database, ed. UK: Energy networks
Association, 2010.
180
___________________________________________________________________________________
[91] M. E. Baran and F. F. Wu, "Optimal Capacitor Placement on Radial-

Distribution Systems," IEEE Transactions on Power Delivery, vol. 4, pp. 725-
734, Jan 1989.
[92] L. Lindstrom, "Evaluating Impact on Ampacity According to IEC-60287

Regarding Thermally Unfavourable placement of Power cables," Masters,
KTH Electrical Engineering, KTH, Sweden, Stockholm, 2011.
[93] N. B. A. D. Centre, "Met Office Integrated Data Archive System (MIDAS)

Land and Marine Surface Stations Data (1853-current)," in UK Soil
Temperature Data, ed. UK, 2015.
[94] E. P. DeGarmo, Engineering Economy, 8th ed. ed. London: London: Collier
Macmillan, 1988.
[95] OFGEM. (2016). RIIO-ED1 price control. Available:

https://www.ofgem.gov.uk/network-regulation-riio-model/riio-ed1-price-
control
[96] J. T. Licenses, "Network Output Measures Methodology," ed. UK: National

Grid UK, SP Transmission PLC, Scottish Hydro Electric Transmission
Limited, 2015, pp. 1-60.
[97] P. S. Georgilakis and N. D. Hatziargyriou, "Optimal Distributed Generation

Placement in Power Distribution Networks: Models, Methods, and Future
Research," IEEE Transactions on Power Systems, vol. 28, pp. 3420-3428,
2013.
[98] F. Capitanescu, L. F. Ochoa, H. Margossian, and N. D. Hatziargyriou,

"Assessing the Potential of Network Reconfiguration to Improve Distributed
Generation Hosting Capacity in Active Distribution Systems," IEEE
[99] S. N. Gopiya Naik, D. K. Khatod, and M. P. Sharma, "Analytical Approach

for Optimal siting and sizing of distributed generation in radial distribution
networks," IET Generation, Transmission & Distribution, vol. 9, pp. 209-220,
2015.
[100] I. Hernando-Gil, I. S. Ilie, and S. Z. Djokic, "Reliability Planning of Active

Distribution Systems Incorporating Regulator Requirements and Network-
Reliability Equivalents," IET Generation, Transmission & Distribution, vol.
10, pp. 93-106, 2016.
[101] C. L. Nick Jenkins, Jianzhong Wu, "An Overview of the Smart Grid in Great
Britain," Engineering, vol. 1, pp. 413-421, 2015-12-30 2015.
[102] B. Fox, Wind Power Integration: Connection and System Operational

Aspects: Institution of Engineering and Technology, 2007.
181
___________________________________________________________________________________
[103] W. El-khattam, K. Bhattacharya, Y. Hegazy, and M. M. A. Salama, "Optimal

Investment Planning for Distributed Generation in a Competitive Electricity
Market," IEEE Transactions on Power Systems, vol. 19, pp. 1674-1684, 2004.
[104] C. L. T. Borges and D. M. Falcão, "Optimal distributed generation allocation

for reliability, losses, and voltage improvement," International Journal of
Electrical Power & Energy Systems, vol. 28, pp. 413-420, 7// 2006.
[105] W. S. Andrade, C. L. T. Borges, and D. M. Falcao, "Modeling Reliability

Aspects of Distributed Generation Connected to Distribution Systems," in
IEEE Power Engineering Society General Meeting, 2006, p. 6 pp.
[106] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, and R. Seethapathy,

"Optimal Renewable Resources Mix for Distribution System Energy Loss
Minimization," IEEE Transactions on Power Systems, vol. 25, pp. 360-370,
2010.
[107] S. Sheng-Yi, L. Chan-Nan, C. Rung-Fang, Gutie, x, and G. rrez-Alcaraz,

"Distributed Generation Interconnection Planning: A Wind Power Case
Study," IEEE Transactions on Smart Grid, vol. 2, pp. 181-189, 2011.
[108] I. Hernando-Gil, I. S. Ilie, A. J. Collin, J. L. Acosta, and S. Z. Djokic, "Impact

of DG and Energy Storage on Distribution Network Reliability: A
Comparative Analysis," in IEEE InternationalEnergy Conference and
Exhibition (EnergyCon), 2012, pp. 605-611.
[109] P. J. Ramirez, D. Papadaskalopoulos, and G. Strbac, "Co-Optimization of

Generation Expansion Planning and Electric Vehicles Flexibility," IEEE
Transactions on Smart Grid, vol. PP, pp. 1-11, 2015.
[110] P. S. Georgilakis and N. D. Hatziargyriou, "A review of Power Distribution

Planning in the Modern Power Systems Era: Models, Methods and Future
Research," Electric Power Systems Research, vol. 121, pp. 89-100, 2015.
[111] N. N. Mansor and V. Levi, "Development of Distribution Networks with Low

Carbon Technologies," in IEEE Innovative Smart Grid Technologies - Asia
(ISGT ASIA), 2015, pp. 1-6.
[112] F. Ding and K. A. Loparo, "Feeder Reconfiguration for Unbalanced

Distribution Systems With Distributed Generation: A Hierarchical
Decentralized Approach," IEEE Transactions on Power Systems, vol. 31, pp.
1633-1642, 2016.
[113] A. R. Jordehi, "Optimisation of electric distribution systems: A review,"

Renewable and Sustainable Energy Reviews, vol. 51, pp. 1088-1100, 2015.
[114] G. Hui, V. Levi, and M. Buhari, "Reliability Assessment of Smart Distribution

Networks," in IEEE Innovative Smart Grid Technologies - Asia (ISGT ASIA),
2015, pp. 1-6.
182
___________________________________________________________________________________
[115] IEEE, "IEEE Standard for Interconnecting Distributed Resources with Electric
Power Systems: IEEE Std 1547.2-2008," 2009.
[116] X. Yixing and C. Singh, "Adequacy and Economy Analysis of Distribution

Systems Integrated With Electric Energy Storage and Renewable Energy
Resources," IEEE Transactions on Power Systems, vol. 27, pp. 2332-2341,
2012.
[117] L. F. Ochoa and G. P. Harrison, "Minimizing Energy Losses: Optimal

Accommodation and Smart Operation of Renewable Distributed Generation,"
[118] M. F. Shaaban, Y. M. Atwa, and E. El-Saadany, "DG Allocation for Benefit

Maximization in Distribution Networks," IEEE Transactions on Power
Systems, vol. 28, pp. 639-649, 2013.
[119] W. Chai, Y. Wen, and W. Liu, "Optimal Allocation of Switches in DG

Enhanced Distribution Systems," in Asia-Pacific Power and Energy
Engineering Conference (APPEEC), 2011, pp. 1-5.
[120] M. S. Thomas and P. P. Terang, "Interconnection Issues for Distributed

Resources in a Smart Distribution System," in India International Conference
on Power Electronics (IICPE), 2012, pp. 1-6.
[121] M. Lavorato, J. F. Franco, M. J. Rider, and R. Romero, "Imposing Radiality

Constraints in Distribution System Optimization Problems," IEEE
[122] C. H. Lo and M. D. Anderson, "Economic Dispatch and Optimal Sizing of

Battery Energy Storage Systems in Utility Load-Leveling Operations," IEEE
Transactions on Energy Conversion, vol. 14, pp. 824-829, 1999.
[123] R. S. Abbasi and T. Mehmood, "Feeder Reconfiguration Techniques: A

Review," in International Conference on Energy Systems and Policies
(ICESP), 2014, pp. 1-5.
[124] A. M. Eldurssi and R. M. O. Connell, "A Fast Nondominated Sorting Guided

Genetic Algorithm for Multi-Objective Power Distribution System
Reconfiguration Problem," IEEE Transactions on Power Systems, vol. 30, pp.
593-601, 2015.
[125] H. Bagheri Tolabi, M. H. Ali, and M. Rizwan, "Simultaneous Reconfiguration,

Optimal Placement of DSTATCOM, and Photovoltaic Array in a Distribution
System Based on Fuzzy-ACO Approach," IEEE Transactions on Sustainable
Energy, vol. 6, pp. 210-218, 2015.
[126] H. Xiaoqing, Q. Ying, W. Peng, and Y. Junhu, "Four-Dimensional Wind

Speed Model for Adequacy Assessment of Power Systems With Wind Farms,"
183
___________________________________________________________________________________
[127] R. Billinton and H. Dange, "Incorporating Wind Power in Generating

Capacity Reliability Evaluation Using Different Models," IEEE Transactions
on Power Systems, vol. 26, pp. 2509-2517, 2011.
[128] M. Blanchard and G. Desrochers, "Generation of Autocorrelated Wind Speeds

for Wind Energy Conversion System Studies," Solar Energy, vol. 33, pp. 571-
579, 1984.
[129] P. Giorsetto and K. F. Utsurogi, "Development of a New Procedure for

Reliability Modeling of Wind Turbine Generators," IEEE Transactions on
Power Apparatus and Systems, vol. PAS-102, pp. 134-143, 1983.
[130] R. Billinton, C. Hua, and R. Ghajar, "A Sequential Simulation Technique for
Adequacy Evaluation of Generating Systems Including Wind Energy," IEEE
Transactions on Energy Conversion, vol. 11, pp. 728-734, 1996.
[131] G. E. P. Box and G. M. Jenkins, Time Series Analysis: Forecasting and

Control: Holden-Day, 1976.
[132] R. Billinton, H. Chen, and R. Ghajar, "Time-Series Models for Reliability

Evaluation of Power Systems Including Wind Energy," Microelectronics
Reliability, vol. 36, pp. 1253-1261, 1996.
[133] W. Wangdee and R. Billinton, "Considering Load-Carrying Capability and

Wind Speed Correlation of WECS in Generation Adequacy Assessment,"
IEEE Transactions on Energy Conversion, vol. 21, pp. 734-741, 2006.
[134] M. S. Miranda and R. W. Dunn, "Spatially Correlated Wind Speed Modelling

for Generation Adequacy Studies in the UK," in IEEE Power Engineering
Society General Meeting, 2007, pp. 1-6.
[135] L. de Magalhaes Carvalho, M. A. da Rosa, A. Martins Leite da Silva, and V.

Miranda, "Probabilistic Analysis for Maximizing the Grid Integration of Wind
Power Generation," IEEE Transactions on Power Systems vol. 27, pp. 2323-
2331, 2012.
[136] M. Lei, Shiyan, L., Chuanwen, J., HongLing, L., Yan, Z., "A review on the
forecasting of windspeed and generated power," Renewable and Sustainable
Energy Reviews, vol. 13, p. 6, 2009.
[137] G. Li and J. Shi, "On Comparing Three Artificial Neural Networks for Wind
Speed Forecasting," Applied Energy, vol. 87, pp. 2313-2320, July 2010 2010.
[138] K. R. Voorspools and W. D. D’haeseleer, "Critical Evaluation of Methods for

Wind-Power Appraisal," Renewable and Sustainable Energy Reviews, vol. 11,
pp. 78-97, 1// 2007.
[139] S. M. Wu and S. M. Pandit, Time Series and System Analysis Modeling and
Applications: University of Michigan Engineering Summer Conferences,
1979.
184
___________________________________________________________________________________
[140] Y. Gao, R. Billinton, and R. Karki, "Large Wind Integrated Bulk System
Reinforcement Planning Using the Joint Deterministic Probabilistic Method,"
IET Conference Publications, vol. 2011, pp. 3B1-3B1, 2011.
[141] R. K. Billinton, R. Gao Yi.,Huang, D. and P. W. Hu, "Adequacy Assessment

Considerations in Wind Integrated Power Systems," IEEE Transaction on
Power Systems,, vol. 27, pp. 2297-2305, 2012.
[142] P. M. T. Broersen, Automatic Autocorrelation and Spectral Analysis: Springer

London, 2006.
[143] A. Keane and M. O'Malley, "Optimal Distributed Generation Plant Mix with
Novel Loss Adjustment Factors," in IEEE Power Engineering Society General
Meeting, 2006, p. 6 pp.
[144] C. J. Mozina, "Impact of Smart Grid and Green Power Generation on

Distribution Systems," in IEEE PES Innovative Smart Grid Technologies
(ISGT), 2012, pp. 1-13.
[145] National Grid, "National Grid’s Stand-Alone DCLF ICRP TNUoS Great
Britain Transport & Tariff Model," National Grid GB, UK2011.
[146] J. Zhu, Optimization of Power System Operation: Wiley, 2015.
[147] B. Venkatesh, R. Ranjan, and H. B. Gooi, "Optimal Reconfiguration of Radial

Distribution Systems to Maximize Loadability," IEEE Transactions on Power
Systems, vol. 19, pp. 260-266, 2004.
[148] R. S. Rao, K. Ravindra, K. Satish, and S. V. L. Narasimham, "Power Loss

Minimization in Distribution System Using Network Reconfiguration in the
Presence of Distributed Generation," IEEE Transactions on Power Systems,
vol. 28, pp. 317-325, 2013.
[149] C. Chung-Fu, "Reconfiguration and Capacitor Placement for Loss Reduction

of Distribution Systems by Ant Colony Search Algorithm," IEEE
[150] A. Kavousi-Fard and T. Niknam, "Optimal Distribution Feeder

Reconfiguration for Reliability Improvement Considering Uncertainty," IEEE
Transactions on Power Delivery, vol. 29, pp. 1344-1353, 2014.
[151] Q. Peng, Y. Tang, and S. H. Low, "Feeder Reconfiguration in Distribution

Networks Based on Convex Relaxation of OPF," IEEE Transactions on Power
Systems, vol. 30, pp. 1793-1804, 2015.
[152] M. W. Siti, D. V. Nicolae, A. A. Jimoh, and A. Ukil, "Reconfiguration and

Load Balancing in the LV and MV Distribution Networks for Optimal
Performance," IEEE Transactions on Power Delivery, vol. 22, pp. 2534-2540,
2007.
185
___________________________________________________________________________________
[153] Y. K. Wu, C. Y. Lee, L. C. Liu, and S. H. Tsai, "Study of Reconfiguration for

the Distribution System With Distributed Generators," IEEE Transactions on
Power Delivery, vol. 25, pp. 1678-1685, 2010.
[154] N. G. Paterakis, A. Mazza, S. F. Santos, O. Erdin, G. Chicco, A. G. Bakirtzis,

et al., "Multi-Objective Reconfiguration of Radial Distribution Systems Using
Reliability Indices," IEEE Transactions on Power Systems, vol. 31, pp. 1048-
1062, 2016.
[155] L. Changhyeok, L. Cong, S. Mehrotra, and B. Zhaohong, "Robust Distribution

Network Reconfiguration," IEEE Transactions on Smart Grid, vol. 6, pp. 836-
842, 2015.
[156] L. Fangxing, "Distributed processing of reliability index assessment and

reliability-based network reconfiguration in power distribution systems," IEEE
[157] T. M. Khalil, A. V. Gorpinich, and G. M. Elbanna, "Combination of Capacitor

Placement and Reconfiguration for Loss Reduction in Distribution Systems
Using Selective PSO," in The 22nd International Conference and Exhibition
on Electricity Distribution (CIRED 2013) 2013, pp. 1-4.
[158] R. C. Lotero and J. Contreras, "Distribution System Planning With

Reliability," IEEE Transactions on Power Delivery, vol. 26, pp. 2552-2562,
2011.
[159] N. N. B. Mansor and V. Levi, "Expansion Planning of Medium Voltage

Distribution Networks," in 50th International Universities Power Engineering
Conference (UPEC), 2015, pp. 1-6.
[160] A. Merlin and H. Back, "Search For a Minimal-Loss Operating Spanning Tree
Configuration in an Urban Power Distribution System," in 5th Power System
Computation Conference (PSCC), 1975.
[161] D. Shirmohammadi and H. W. Hong, "Reconfiguration of Electric

Distribution Networks for Resistive Line Losses Reduction," IEEE
Transactions on Power Delivery, vol. 4, pp. 1492-1498, 1989.
[162] T. E. DcDermott, I. Drezga, and R. P. Broadwater, "A Heuristic Nonlinear

Constructive Method For Distribution System Reconfiguration," IEEE
[163] S. K. Goswami and S. K. Basu, "A New Algorithm For The Reconfiguration
of Distribution Feeders For Loss Minimization," IEEE Transactions on Power
Delivery, vol. 7, pp. 1484-1491, 1992.
[164] F. V. Gomes, S. Carneiro, J. L. R. Pereira, M. P. Vinagre, P. A. N. Garcia, and

L. R. D. Araujo, "A New Distribution System Reconfiguration Approach
Using Optimum Power Flow and Sensitivity Analysis for Loss Reduction,"
186
___________________________________________________________________________________
187
LIST OF PUBLICATIONS
International peer reviewed journals
JP1 M. Buhari, V. Levi, and S. K. E. Awadallah, "Modelling of Ageing

Distribution Cable for Replacement Planning," IEEE Transactions on Power Systems,
vol. DOI: 10.1109/TPWRS.2015.2499269, pp. 1-9, 2015. Accepted and awaiting
publication.
JP2 M. Buhari, V. Levi, and A. Kapetanaki, "Optimal Management of Distribution

Networks with DG to Reduce Cable Ageing and Replacement," IEEE Transactions on
Smart Grids, 2016. Under review.
Conference Proceedings
CP1 M. Buhari and K. Kopsidas, "Probabilistic assessment of impacts of real-time

line ratings on distribution networks," in International Conference on Probabilistic
Methods Applied to Power Systems (PMAPS), 2014, pp. 1-6. Accepted for oral
presentation and published.
CP2 M. Buhari, K. Kopsidas, C. Tumelo-Chakonta, and A. Kapetanaki, "Risk

assessment of smart energy transfer in distribution networks," in IEEE PES Innovative
Smart Grid Technologies (ISGT), 2014, pp. 1-6. Accepted for poster presentation
and published.
CP3 G. Hui, V. Levi, and M. Buhari, "Reliability assessment of smart distribution

networks," in IEEE Innovative Smart Grid Technologies - Asia (ISGT ASIA), 2015,
pp. 1-6. Accepted for oral presentation and published.
CP4 A. Kapetanaki, K. Kopsidas, C. Tumelo-Chakonta, and M. Buhari, "Network

planning evaluation implementing time varying thermal ratings," in International
Conference on Probabilistic Methods Applied to Power Systems (PMAPS), 2014, pp.
1-6. Accepted for oral presentation and published.
188

PHD Thesis

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

PHD Thesis

Transféré par

Droits d'auteur :

Formats disponibles

Reliability Assessment of Ageing Distribution

Cable for Replacement in Smart Distribution

A Thesis submitted to The University of Manchester for the Degree of

in the Faculty of Science and Engineering

School of Electrical and Electronic Engineering

Electrical Energy and Power Division

List of Tables ................................................................................................................ 7

List of Abbreviations ................................................................................................... 9

List of Symbols ........................................................................................................... 12

Chapter one ................................................................................................................ 22

1 Power system reliability studies ...................................................................... 22

1.1 Distribution network reliability ................................................................. 24

1.2 Reliability evaluation methods .................................................................. 26

1.3 Monte Carlo simulation (MCS) ................................................................. 28

1.4 Research motivation .................................................................................. 33

1.5 Research question ...................................................................................... 35

1.6 Research aim and objectives ..................................................................... 35

1.7 Contributions ............................................................................................. 37

1.8 Thesis structure .......................................................................................... 38

Chapter two ................................................................................................................ 42

Thermal modelling of ageing cables ......................................................................... 42

2 Modelling of ageing cables .............................................................................. 42

2.2 Brief historical overview of studies on conductor thermal ratings ............ 43

2.3 Review of cable thermal rating standards .................................................. 45

2.4 Arrhenius model of ageing cables ............................................................. 55

2.5 Proposed method for the determination of Arrhenius parameters ............. 56

2.6 Cable thermal loss of life model ................................................................ 59

2.7 Chapter summary ....................................................................................... 61

Reliability assessment of ageing cables .................................................................... 62

3 Reliability assessment considering end-of-life failure ..................................... 62

3.1 Literature review ........................................................................................ 62

3.2 Basic features of reliability distribution functions ..................................... 68

3.3 Exponential probability distribution function ............................................ 70

3.4 Weibull probability distribution function .................................................. 71

3.5 Weibull renewal process ............................................................................ 73

3.6 Developed IEC-Arrhenius-Weibull reliability model for ageing cable ..... 74

3.7 Chapter summary ....................................................................................... 78

Chapter four ............................................................................................................... 80

Cable ageing study ..................................................................................................... 80

4 Details of SMCS model ................................................................................... 80

4.2 Coefficient of variation (CoV) .................................................................. 83

4.3 Simulation methodology ........................................................................... 84

4.4 Test systems............................................................................................... 91

4.5 Study cases and presentation of results ................................................... 100

4.6 Chapter summary..................................................................................... 105

Chapter five .............................................................................................................. 108

Investment planning ................................................................................................ 108

5 Distribution system investment planning ...................................................... 108

5.1 Reinforcement planning (load related planning) ..................................... 109

5.2 Replacement planning ............................................................................. 110

5.3 Regulatory investment constraints .......................................................... 111

5.4 Ranking for cable replacement ................................................................ 112

5.5 Proposed cable ranking index .................................................................. 116

5.7 Chapter summary..................................................................................... 121

Chapter six................................................................................................................ 122

Smart distribution systems ..................................................................................... 122

6 Smart power networks ................................................................................... 122

6.1 Smart measures to reduce replacement cost ............................................ 124

6.3 Smart operations (network reconfiguration) ............................................ 142

6.4 Smart grid planning study ........................................................................ 150

6.7 Results of optimal integration of wind-DG and network reconfiguration to

6.8 Chapter summary ..................................................................................... 162

Chapter seven ........................................................................................................... 164