Vous êtes sur la page 1sur 81

  EEE 6906

  Reliability of Power System


  Generation Capacity Reliability
Assessment
  October 2017

A H Chowdhury
Professor, EEE, BUET
Table of Contents

 Generation system model


 Loss of load indices
 Loss of energy indices
 Frequency and duration method

2
2
Generation System Model

• Planning and operation need to determine amount of system generating


capacity to ensure adequate supply

• Problem divided into two areas

– Static capacity requirements: long-term evaluation of overall system requirement

– Operating capacity requirements: short-term evaluation of actual capacity


required to meet a given load level

• Fundamental difference between static and operating capacity evaluation is


in time period considered

3
Generation System Model

Static capacity requirement


• Installed capacity that must be planned and constructed in advance of
system requirements

• Static reserve must be sufficient to provide for

– Overhaul of generating equipment

– Unscheduled outages

– Load growth in excess of estimates

4
Generation System Model

 Basic approach to evaluating generation


capacity adequacy consists of three parts
– generation model
– load model
– risk model

 Calculated indices do not normally


include transmission constraints or
transmission reliabilities
 These are overall system adequacy
Conceptual tasks in generating
indices capacity reliability evaluation

5
Generation System Model

Conventional system model

 Calculated indices in this case do not reflect generation deficiencies at any


particular customer load point

 Measure overall adequacy of generation system

6
Generating Unit Unavailability

• Downtime or outage duration  period of time that a component fails to provide its
primary function

• Unavailability  Probability of finding an equipment on outage at some future time

– Known as unit forced outage rate (FOR) in power system applications

• It is ratio of two time values

• Basic generating unit parameter used in static capacity evaluation  probability of


finding unit on forced outage at some distant time in future

7
Generating Unit Unavailability

8
Generating Unit Unavailability

Two-state model for a base load unit

• Concepts of availability and unavailability associated with simple two-state model


• This model is directly applicable to a base load generating unit which is either
operating or forced out of service
• Scheduled outages must be considered separately

9
Generating Unit Unavailability

• FOR is adequate estimate in case of Four-state model proposed by IEEE


generating equipment with relatively long Subcommittee on Application of
operating cycles Probability Methods

 inadequate estimate when demand


cycle relatively short (as in peaking unit)
• Start-up period is most critical period in
operation of a unit
• Peaking unit have fewer operating hours
and more start-ups and shut-downs than
base load unit
• These aspects must be included in
estimating unit unavailability

10
Generating Unit Unavailability

• Difference between 2-state model and 4-state model  inclusion of 'reserve


shutdown' and 'forced out but not needed' states

• In 4-state model, 2-state model represented by States 2 and 3

• Two additional states included to model effect of relatively short duty cycle

• Failure to start condition represented by transition rate from State 0 to State 3

• Can be represented as a Markov process and equations developed for probabilities


of residing in each state in terms of state transition rates

11
Generating Unit Unavailability

• Equations for 4-state model

i.e. 'reserve shutdown‘ state eliminated

12
Generating Unit Unavailability

• In case of an intermittently operated unit, conditional probability that the unit will
not be available given that a demand occurs is P

• P can be found from generic data shown in 4-state model


• Over a relatively long period of time,

Factor f serves to weight forced


outage time FOT to reflect time
unit was actually on forced outage
when in demand by system

13
Generating Unit Unavailability

Example
Average unit data Calculated times

• P dependent on demand placed upon unit


• Demand placed upon it in past may not be
same as demand which may exist in future
• Demand should be determined from load
model as capacity table created
sequentially
• P then determined prior to adding unit to
capacity model
14
Generation System Model

Capacity Outage Probability Table


• Generation model required in loss of load approach known as a capacity
outage probability table
– A simple array of capacity levels and associated probabilities of existence

– If all units in system identical, capacity outage probability table obtained using
binomial distribution

– Units combined using basic probability concepts

• This approach can be extended to powerful recursive technique in which


units are added sequentially to produce final model

15
Generation System Model

• A system consists of two 3 MW units and one 5 MW unit with forced outage
rates of 0.02

• Two identical units can be combined to give capacity outage probability


table shown

(0.98 x 0.98)
(0.98 x 0.02)+(0.02x0.98)
(0.02 x 0.02)

16
Generation System Model

• The 5 MW generating unit can be added to this table by considering that it can
exist in two states
– unit can be in service with probability 1 —0.02 = 0.98
– unit can be out of service with probability 0.02

• Two resulting tables are therefore conditional upon assumed states of unit
• This approach can be extended to any number of unit states

5 MW unit in service 5 MW unit out of service

17
Generation System Model

• Two tables combined and re-ordered

• Probability value in table gives the


Capacity outage probability table for
the three-unit system probability of exactly the indicated
amount of capacity being out of service

• An additional column gives cumulative


probability

Cumulative probability
• Probability of finding a quantity of
capacity on outage equal to or greater
than indicated amount

18
Generation System Model

• Cumulative probability values decrease as


capacity on outage increases
• Same general trend is followed with some
exceptions
• For instance,
– Probability of losing 8 MW >
probability of losing 6 MW
– Because in 8 MW case, 3 MW loss
contribution can occur in two ways

• In practical system probability of large quantity


of capacity forced out of service quite small as
this requires outage of several units

19
Generation System Model

• Theoretically capacity outage probability table incorporates all system


capacity

• Table can be truncated by omitting all capacity outages for which


cumulative probability less than a specified amount, e.g. 10-8

• Table truncated progressively with each unit addition

20
Generation System Model

• Large number of units of different capacities in practical system


• Table contain several hundred possible discrete capacity outage levels

• Number reduced by
– Grouping units into identical capacity groups prior to combining 
approximations necessary
– Rounding table to discrete levels after combining  no approximations necessary

• Capacity rounding increment used depends upon accuracy desired


• Final rounded table contains capacity outage magnitudes that are
multiples of rounding increment
• Number of capacity levels decreases as rounding increment increases 
corresponding decrease in accuracy

21
Generation System Model

Procedure for rounding a table


• Two 3 MW units and one 5 MW unit with
forced outage rates of 0.02 combined to
form generation model
• Rounded at 5 MW increments  table
contains only capacity outage magnitudes
of 0, 5, 10 and 15 MW
• General expression for rounding process
 for all states i falling between required
rounding states j and k

Ck  Ci Ci  C j
P(C j )  P(Ci ) P(Ck )  P(Ci )
Ck  C j Ck  C j

22
Generation System Model

C j  0 Ck  5 Ci  3
53 2
P(0)  P(3)  (0.038416)
50 5
30 3
P(5)  P(3)  (0.038416)
50 5

C j  5 Ck  10 Ci  6,8
Ci  6
10  6 4
P (5)  P ( 6)  (0.0 00392)
10  5 5
65 1
P (10)  P (6)  (0.000392)
10  5 5
Ci  8
10  8 2
P (5)  P ( 6)  (0.0 00784) C j  10 Ck  15 Ci  11
10  5 5
15  11 4
85 3 P(10)  P(11)  (0.000008)
P (10)  P (8)  (0.000784) 15  10 5
10  5 5 11  10 1
P(15)  P(11)  (0.000008)
15  10 5

23
Generation System Model

• Use of a rounded table in combination with load model to calculate risk


level introduces certain inaccuracies
• Error depends upon rounding increment used and on slope of load
characteristic

• Error decreases with increasing slope of load characteristic

• For a given load characteristic error increases with increased rounding


increment

• Rounding increment used should be related to system size and composition

• Also first non-zero capacity-on-outage state should not be less than


capacity of smallest unit

24
Generation System Model

Comparison of Deterministic and Probabilistic Criteria


• Deterministic risk criteria such as 'percentage reserve' and 'loss of largest
unit' do not define consistently true risk
• Consider following four systems:
System 1 24 x 10 MW units each having a FOR of 0.01
System 2 12 x 20 MW units each having a FOR of 0.01
System 3 12 x 20 MW units each having a FOR of 0.03
System 4 22 x 10 MW units each having a FOR of 0.01

• In each system, units are identical


• Therefore, capacity outage probability table can be easily constructed using
binomial distribution

25
Generation System Model

• Capacity outage probability tables for systems 1-4


• Truncated to a cumulative probability of 10-6
• Considerable number of capacity outage states deleted using truncation technique

26
Generation System Model

• Load level or demand assumed constant


• Risk in system = probability of not meeting load

• True risk in system given by value of cumulative probability corresponding

to outage state one increment below that which satisfies load on system

• The two deterministic risk criteria ‘percentage reserve’ and ‘loss of largest

unit’ can now be compared with this probabilistic risk

27
Generation System Model

(a) Percentage reserve margin

• Expected load demands in systems 1, 2, 3 and 4 : 200, 200, 200 and 183 MW

• Installed capacity in all four cases such that there is a 20% reserve margin

• Probabilistic or true risks in each system found from Table on previous slide:

Risk in system 1 = 0.000004 Risk in system 3 = 0.004847


Risk in system 2 = 0.000206 Risk in system 4 = 0.000063

28
Generation System Model

Risk in system 1 = 0.000004 Risk in system 3 = 0.004847


Risk in system 2 = 0.000206 Risk in system 4 = 0.000063

System 1 24 x 10 MW units each having a FOR of 0.01 • True risk in system 3 is x1000 times
than that in system I
System 2 12 x 20 MW units each having a FOR of 0.01
• Risk merit order 3-2-4-1
System 3 12 x 20 MW units each having a FOR of 0.03
• Variation in true risk depends upon
System 4 22 x 10 MW units each having a FOR of 0.01
forced outage rate, number of units,
load demand
• Percentage reserve method cannot
account for these factors  using a
'constant' risk criterion, but does not
give a consistent risk assessment of
system

29
Generation System Model

(b) Largest unit reserve


• Expected load demands in systems 1, 2, 3 and 4 : 230, 220, 220 and 210 MW
• Installed capacity in all cases such that reserve equal to largest unit
• The probabilistic risks are:
risk in system 1 = 0.023855
risk in system 2 = 0.006175
risk in system 3 = 0.048650
risk in system 4 = 0.020229
• Variation in risk is much smaller
• Ratio between smallest and greatest risk levels 8:1
• Risk merit order changed to 3-1-4-2

30
Generation System Model

• Comparisons show that use of deterministic or 'rule-of-thumb' criteria can


lead to very divergent probabilistic risks; even for systems that are very
similar

• They are therefore inconsistent, unreliable and subjective methods for


reserve margin planning

31
Recursive Algorithm for Capacity Model
Building

• Capacity model can be created using simple algorithm which can also be used to
remove a unit from model

• Can also be used for a multi-state unit  a unit which can exist in one or more
derated or partial output states as well as in fully up and fully down states

• Two cases
– Case 1 No derated states (2-state unit)
– Case 2 Derated states included (multi-state unit)

32
Recursive Algorithm for Capacity Model
Building

Case 1 No derated states System data (each unit has an unavailability of 0.02)
Cumulative probability of a particular
capacity outage state of X MW after
a unit of capacity C MW and FOR U
added:
P(X) = (1 - U)P’(X) + (U)P’(X- C)
P’(X), P(X)  cumulative probabilities
of capacity outage state of X MW
before and after unit is added
• Initialized by setting P’(X) = 1.0 for
X  0 and P’(X) = 0 otherwise
• System capacity outage probability
created sequentially as shown

33
Table of Contents

 Generation system model

 Loss of load indices


 Loss of energy indices
 Frequency and duration method

34
34
Loss of Load Indices

Loss of Load Expectation

 Applicable system capacity outage probability table combined with system load
characteristic give an expected risk of loss of load

 Capacity outage

‘Capacity outage’ indicates a loss of generation which may or may not result in a
loss of load

 Loss of load

A ‘loss of load’ will occur only when capability of generating capacity remaining in
service is exceeded by system load level

35
Loss of Load Indices

System data n
LOLE   Pi (Ci  Li )
i 1
where Ci  available capacity on day i
Li  forecast peak load on day i
Pi (Ci  Li )  probabilit y of loss of load on day i
Load data for a period of 365 days This value is obtained directly
from the capacity outage cumulative
probability table

LOLE = 12P (100-57) + 83P(100-52) + I07P(100-46)


+ 116P(100-41) + 47P(100-34)
= 12(0.020392) + 83(0.020392) + 107(0.000792)
+ 116(0.000792) + 47(0.000792)
= 2.15108 days/year.

36
Loss of Load Indices

Installed capacity (MW)


n
Ok Reserve LOLE   pk tk time units
Daily peak load (MW)

k 1
n
LOLE   (tk  tk 1 ) Pk
tk k 1
Pk  cumulative outage probabilit y for capacity state Ok

0 Time load exceeds the indicated value 365

Relationship between load, capacity and reserve


Ok = Magnitude of kth outage in system capacity
outage probability table
tk = Number of time units in study interval that an
outage magnitude of Ok would result in a loss
of load

37
Loss of Load Indices

 Consider a system containing five 40 MW units each with a forced outage rate of 0.01
 System installed capacity = 200 MW

Daily peak load variation curve


Generation model for the five-unit system
Probability values less than lO-6 neglected

100
Daily peak load (%)
40
0 Percentage of days daily peak load 100
Exceeded indicated value

38
Loss of Load Indices

LOLE using individual probabilities

Time periods during which


loss of load occurs

39
Loss of Load Indices

LOLE using cumulative probabilities

40
Loss of Load Indices

Scheduled Outages
Installed capacity (MW)
 System capacity evaluation examples
previously considered assumed Ok Reserve

Daily peak load (MW)


– Load model applied to entire period
– System capacity model applicable for entire
tk
period

 Not the case if units removed from service


for periodic inspection and maintenance in
accordance with a planned program 0 Time load exceeds the indicated value 365

 During this period, capacity available for Relationship between load,


capacity and reserve
service not constant
– single capacity outage probability table is
not applicable

41
Loss of Load Indices

 Annual LOLEa can be obtained by


– dividing the year into periods

– calculating period LOLEp values


using modified capacity model and
appropriate period load model

 Annual risk index is then given by


n
LOLEa   LOLEp A hypothetical example of a maintenance
schedule for a winter peaking system
p 1

42
Loss of Load Indices

• Modified capacity model can be obtained by creating new capacity outage


probability table for each capacity condition
– Unit removal algorithm can be used in this case

• Total installed capacity may increase due to commissioning of a new unit


– Can be added to capacity model in appropriate periods

• If actual in-service date of new unit is uncertain, it can be represented by


a probability distribution and incorporated on a period basis using following
equation
LOLEp = period LOLE value
LOLEpa = period LOLE value including the unit
LOLEpu = period LOLE value without the unit
a = probability of unit coming into service
u = probability of unit not coming into service

43
Loss of Load Indices

 The unit still has opportunity to fail given that it comes into service

 This is included in LOLEpa value


n
 Annual risk index is then obtained using, LOLEa   LOLEp
p 1

44
Loss of Load Indices

 Maintenance may be considered by

 (Fig. A) adding capacity on maintenance to


load and using a single capacity outage
probability table
Fig. A
 (Fig. B) using original capacity outage
probability table, and reducing total
available capacity by quantity on outage

 Both methods are approximations


 state probabilities in generation model are
Fig. B
unaltered and therefore do not really relate to
system during maintenance

45
Loss of Load Indices

 Most realistic approach  combine units actually available to system into a


capacity outage probability table applicable for period considered

 Practical system studies using both approximate and realistic method indicate that
– Adding capacity on maintenance to load or subtracting it from installed capacity without
altering outage probabilities results in higher calculated risk levels
– Error increases with increased maintenance capacity
– Error negligible in large system with an extremely small percentage of total installed
capacity on maintenance

• Removing units on maintenance from capacity outage probability table results in


negligible error for normal magnitudes of capacity on maintenance for those cases
when units removed are not exact multiples of rounding increment used in table

46
Loss of Load Indices

 If maintenance is scheduled to
minimize risk or in accordance with
a constant risk criterion then
reserve may be quite variable

 Constant reserve not same as


constant risk
– System reliability with a low FOR unit
removed not same as that with a high Annual load and capacity model
FOR unit removed

47
Loss of Load Indices

 Approximate techniques for scheduling maintenance


1. (i) Reduce total installed capacity by expected capacity loss (i.e. unit capacity x
its availability) rather than by actual unit capacity

(ii) Schedule maintenance on a constant reserve basis

2. A better and quite accurate approach if only a few units on maintenance at any
given time

(i) Determine decrease in PLCC at appropriate risk level for each individual unit
on maintenance

(ii) Use these values in scheduling maintenance on a constant reserve basis

 Applicable approach depends on capacity composition, required


maintenance level and system load profile

48
Table of Contents

 Generation system model

 Loss of load indices

 Loss of energy indices


 Frequency and duration method

49
49
Loss of Energy Indices

Loss of Energy Expectation (LOEE)


 LOLE calculated using load duration curve or individual hourly values
 Area under load duration curve represents energy utilized
 Can be used to calculate expected energy not supplied due to insufficient installed
capacity and expressed as a ratio
Load energy curtailed due to deficienci es in generating capacity available
Total load energy required to serve requiremen ts of system

 Area under load duration curve represents energy utilized


 For a given load duration curve this ratio is independent of time period considered
and gives 'energy index of unreliability‘
 Or, (1 - 'energy index of unreliability' ) = ('energy index of reliability‘)

50
Loss of Energy Indices

 Any outage of generating capacity exceeding Probabilities of having varying


amounts of capacity unavailable
reserve result in a curtailment of system load combined with system load
energy
 Let:
Ok = magnitude of capacity outage
Pk = probability of a capacity outage equal to Ok
Ek = energy curtailed by a capacity outage equal
to Ok

 Probable energy curtailed: EkPk

 Sum of these products gives total expected


energy curtailment or loss of energy
expectation (LOEE) n Energy curtailment due to a given
LOEE   Ek Pk capacity outage condition
k 1

51
Loss of Energy Indices

 Per unit LOEE


1
Ek Pk
LOEE pu   normalized total energy, E, under load duration curve
k 1 E

Probable load energy curtailed due to deficienci es in available generating capacity


LOEE pu 
Total load energy required to serve systemdemand

 Energy index of reliability, EIR

EIR  1  LOEE pu

52
Loss of Energy Indices

 System load-duration curve Risk as a function of system peak load


represented by a straight line from EIR for 5 x 40 MW unit system using
100% to 40% load level LOLE approach

 ‘Loss of energy’ approach has more


physical significance than ‘loss of
load’ approach

 Not as flexible in overall application


and has not been used as extensively

 For energy limited system energy


based indices is more suitable than
that focused on power or capacity

53
Loss of Energy Indices

Expected Energy Not Supplied (EENS)


 Expected energy curtailed concept can be
used to determine expected energy
produced by each unit in system
– Provides a relatively simple approach to
Load duration curve (LDC) for a
production cost modelling period of 100 hours

 Assumed economic loading order: units Generating unit capacity data


1,2, 3
 Total required energy = area under LDC =
4575 MWh
 With no units in system
– EENS = 4575 MW (=EENS0)

54
Loss of Energy Indices

(4575MWh - 25MWx100hr)

(2075x0.65)

EENS with only Unit 1 (EENS1)

 Contribution from Unit 2 obtained by adding Unit 2 to capacity model of EENS1 and
calculating EENS for Units 1 and 2 combined

Expected energy produced


by Unit 2 = EENSI-EENS2
= 2500.0 - 401.7
= 2098.3 MWh

EENS with Units 1 and 2 (EENS2)

55
Loss of Energy Indices

EENS with units 1, 2 and 3 (EENS 3)

 Expected contribution from Unit 3: (401 .7 - 64.08) = 337.6 MWh

56
Loss of Energy Indices

Summary of individual unit expected energy outputs

 Expected energy not supplied is 64.08 MWh


 In terms of EIR n
E P
LOEE pu   k k EIR  1  LOEE pu
k 1 E
64.08
EIR  1   0.985993
4575.0

57
Table of Contents

 Generation system model


 Loss of load indices
 Loss of energy indices

 Frequency and duration method

58
58
Frequency and Duration Method

Loss of Load Expectation (LOLE)


– Expected number of days (or hours) in a given period load exceeded available
capacity

Loss of Energy Expectation (LOEE)


– Expected energy not supplied in a given period due to insufficient installed capacity

 Both can be used to compare adequacy of alternative configurations and expansions

 No indication of frequency or duration of occurrence of an insufficient capacity

59
Frequency and Duration Method

Basic approach

Basic system model

60
Frequency and Duration Method

λ
State 0 State 1
Two-state model for a base component component
load generating unit operable μ failed

 LOLE or LOEE methods utilize steady-state availability (A) and unavailability (U)
for this model

 F&D technique utilizes A&U, and transition rate parameters λ&μ

 Frequency of encountering state 0 = (probability of being in state) x (rate of


departure from state)

61
Frequency and Duration Method

 Frequency of a particular condition can be expressed as mathematical expectation


of encountering boundary wall surrounding that condition

 Frequency of entry equal to frequency of leaving

62
Frequency and Duration Method

Generation Model
System data Failure modes and effects

 n = number of units; each unit can exist in two states  2n states in total system
(23 = 8 in this case)

 Total number of states in system enumerated in second table

63
Frequency and Duration Method

 States represented as a state transition diagram


– enumerates all possible system states and shows
transition modes from one state to another

Failure modes and effects

64
Frequency and Duration Method

Example
 Given that system in State 2 can transit to States 1,
5 or 6 in following ways:
 State 2 to 1 if unit 1 repaired
 State 2 to 5 if unit 2 fails
 State 2 to 6 if unit 3 fails
 Total rate of departure from State 2  sum of
individual rates of departure (μ1+λ2+λ3)

 Probabilities associated with each state can be


calculated assuming event independence

 Frequencies of encountering each state when rate of


departure or entry sum of appropriate rates 

65
Frequency and Duration Method

Generation model of the three unit system identical capacity states


can be combined

i refers to identical states


k refers to new merged state

0.01+0.49+0.01=0.51

66
Frequency and Duration Method

Reduced generation model

 This model gives probability and frequency of having a given level of capacity
forced out of service

 Can be modified to give cumulative probabilities and frequencies rather than


values corresponding to a specific capacity level

67
Frequency and Duration Method

Cumulative state values


 At any given capacity level cumulative values give probability and frequency of
having that capacity or more forced out of service
 Individual state probabilities and frequencies combined to form cumulative state
values

pk  probability of state k
n  cumulative state with known probability and frequency
k  state which is being combined to form cumulative state n – 1
λ+k and λ-k  transition rates to higher and lower available capacity levels

 Process starts from last state, in which individual and cumulative values are same

68
Frequency and Duration Method

Combining individual state probabilities and


frequencies to form cumulative state values

69
Frequency and Duration Method

Basic equations for combining individual state probabilities and frequencies to form
cumulative state values 

Reduced generation model


Process starts with
last state

Individual and cumulative


values same in the last state

70
Frequency and Duration Method
Reduced generation model

0.000784(0.01)=0.0000784

0.000784(0.49+0.49)=0.00076832
71
Frequency and Duration Method

Generation model

failure rate

repair rate

72
Frequency and Duration Method

73
Frequency and Duration Method

Generation system model


 Table contains both cumulative
probability and frequency values at
each capacity level  complete
system capacity model
 This model can be used directly as an
indication of system generating
capacity adequacy
- the approach is known as ‘loss of
capacity method’ Conventional capacity outage
probability table as used in LOLE
approach does not include any
frequency parameters

74
Frequency and Duration Method

System Risk Indices

 Generating capacity model combined with load to obtain system risk indices

 Number of possible load models which can be used, e.g.


– individual state load model

– cumulative state load model

75
Frequency and Duration Method

Individual State Load Model


 Load cycle for a period is a random sequence of N load levels
 Daily load model contains a peak load level of mean duration e day and a fixed
low load of 1 - e day
 e  exposure factor
 As X decreases, e increases  a more severe load model

No precise method
for selecting value
X at which to
determine
magnitude of e

Daily load model

76
Frequency and Duration Method

Period load model • Each peak load returns to low load


level each day before transferring to
another peak next day
• Each peak load has constant ‘mean’
duration e day separated by low load
of mean duration 1 - e day

 When e = 1, load represented by daily peak value as in conventional LOLE approach

 X of 85% of daily peak load sometimes used

 Load state transitions occur independently of capacity state transitions

 F&D analysis normally done on a period basis  since constant low load level and
random peak load sequence do not apply over long period of time

 Maintenance of generating capacity results in modified capacity model

77
Frequency and Duration Method

Parameters required to completely define period load model

78
Frequency and Duration Method

 Combination of discrete levels of available capacity and discrete levels of load


creates a set of discrete capacity margins mk
– A margin  difference between available capacity and system load

– Negative margin represents a state in which system load exceeds available capacity 
system failure condition

– Cumulative margin state contains all states with a margin less than or equal to specified
margin

 A margin state mk is combination of load state Li and capacity state Cm

 Rates of departure associated with mk

79
Frequency and Duration Method

 Probability of occurrence of two or more events in a single small increment of time


assumed negligible
 Transition from one margin state to another made by change in load or change in
capacity but not by both simultaneously
 Upward margin transition rate λ+m summation of upward capacity transition rate
and downward load transition rate
 Opposite true for downward margin transition rate λ-m
 Probability of margin state = (Capacity state) x (load state probabilities)

 Frequency of encountering margin state mk = (Steady-state probability of margin


state) x (summation of rates of departure from state)

80
Frequency and Duration Method

Combining identical margin states

 Different combinations of capacity and load states can result in identical


margin states

 Identical states independent of each other


 can only transit from one to another by going through low load state Lo and by a
change in capacity

 For a given margin state mk made up of s identical margin states

81

Vous aimerez peut-être aussi