Académique Documents
Professionnel Documents
Culture Documents
Research Article
ISSN 1751-8687
Effectiveness of power electronic voltage Received on 2nd March 2016
Revised on 27th June 2016
regulators in the distribution network Accepted on 27th July 2016
doi: 10.1049/iet-gtd.2016.0300
www.ietdl.org
Abstract: Steady-state voltage levels will be a significant problem in the future distribution network due to a high
penetration of new technologies, particularly photovoltaics. A smart transformer which incorporates a fixed tap
transformer with a power electronic voltage regulator has the potential to mitigate this problem by varying the line
voltage automatically in response to changes in loading. This study considers various control strategies for the voltage
regulator and looks at the potential impact that a smart transformer could have on reducing system voltage violations.
These impact studies are performed on a computer model of an entire utilitys distribution system in New Zealand.
Previous research in the area of low voltage regulators has focused The effectiveness of the voltage regulator depends on how its output
more on the technology used to regulate the voltage, or has been voltage is set based on line and load conditions, and the regulators
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
3816 This is an open access article published by the IET under the Creative Commons
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
Table 1 Summary of previous studies to evaluate the impact of the voltage regulator accurately. It
therefore becomes important to know how to set the regulator
Ref. Electricity network Contribution/comments
based on the available data (such as current or voltage), and to
[3] IEEE 34 bus system The contribution was to develop
know whether additional sensors are worthwhile. Once the
a NewtonRaphson based appropriate control strategies have been established, it is possible
three-phase power-flow for to evaluate the impact of the voltage regulator accurately.
modelling unbalanced Two possible control modes are available. The rst is to supply a
distribution networks.
Demonstrated the possible
voltage set-point (analogue quantity) to the voltage regulator. This is
impact of DG on the distribution referred to as full voltage range. The second, already implemented in
system. the voltage regulator, is a logical input that allows the regulator to
[4] A UK Generic distribution The coordination of three switch between two pre-set programmed voltage set-points.
system, which consists of 95 devices: OLTC, voltage regulator,
11 kV busbars and the PV inverter reactive
In the case of full voltage range the set-point is a linear (or non-linear)
power was presented. The function of; current, voltage, power, or a combination of them. The ideal
number of tap operations is is full output voltage range and full measurement data (voltages at all
minimised to extend the the nodes). Other options have been explored such as using one
operating life of the tap control
mechanism.
voltage sensor along the feeder or no voltage sensors and using the
[5] The IEEE 123 bus unbalanced Proposed a power-flow current through the regulator to linearly determine the set-point
test feeder formulation which incorporated voltage (effectively acting like line-compensation). Another option
series voltage regulators in explored is to use the proportion of maximum power being supplied
unbalanced distribution feeders.
The importance of modelling
(either in a linear or non-linear manner).
voltage regulators is highlighted Four ways of controlling the switching between the two set-point
by comparing results with and voltages were compared. They were (i) current compared with an
without the voltage regulator. The offset-current level, (ii) end-of-feeder voltage (compared with 1.06
voltage regulator is an
autotransformer and an on-load
p.u.), (iii) end-of-feeder voltage (compared with 1.00 p.u.) and (iv)
tap-changing mechanism. the direction of the current.
[6] 19 busbar test system Proposed a method for
determining optimal allocation
and tap setting of voltage
regulators in an unbalanced 3 Modelling
radial distribution system using
particle swarm optimisation. A
tapped voltage regulator is used.
3.1 Distribution system
[7] IEEE 34 bus test Different configurations of a
tapped autotransformer voltage The computer simulations use the complete Orion low voltage
regulator were incorporated into distribution system, which supplies the city of Christchurch, New
a power-flow algorithm and their Zealand, and its surrounding rural environs. Electrically, this
behaviour compared. (Wye,
open-delta and closed delta) comprises 10,558,11 kV/415 V transformers, their associated feeders,
[8] Singapores high voltage Stepped (solid-state) voltage and nearly 200,000 customer loads. A GIS spreadsheet containing
transmission/distribution regulator, using a full transformer the lengths, conductor types, number of loads on any conductor, and
system (400 kV, 220 kV, 66 kV, rather than autotransformer. peak load values was obtained from Orion. Parent, branch and asset
10.516 kV busbars). Three designs are presented and
applied to Singapores system. IDs allowed the construction of each network. The line impedances
[9] one distribution feeder from Looked at the coordination of DG and transformer ratings were also provided. A program in MATLAB
Aurora Energys network (48 and OLTC for voltage was written to construct each low voltage distribution feeder, assign
km long with 1.0136 MVA load management. loads to network nodes, look up conductor impedances, current
distributed along the line)
[10] not applied to any test system Proposed a methodology for limits and transformer ratings [12]. This was used as input to another
controlling traditional voltage program which processes the power-ow of each low voltage
regulators assuming smart-meter transformer and associated feeders, returning the voltage at each
data is available. node, all branch currents, and other statistics.
[11] 22 kV distribution system from The effect of various DG
one substation in Thailand scenarios on voltage magnitude Time-varying load information was not available. Each load was
and losses was considered. able to be matched with the correct LV conductor, however the
exact location of loads along the conductor is not known. These
were assumed to be distributed equally along the conductor, with
the last customer placed at the end of a feeder. Exact load values
output voltage should compensate for the line voltage drop (or rise), for each installation control point (ICP) (these are customer
ensuring that the voltage at all customers is within the statutory connection points) were not available, and the transformer load
requirements. Ideally full network voltage data would be available was distributed equally amongst ICPs. In practice the load
to the regulator. However, in practice full network data is usually distribution is less reliable in those cases where the feeder supplies
not available to the regulator. Smart-metering data may present an a signicant number of non-residential loads. The single-phase
opportunity to supply the regulator with this eventually, but at loads were distributed across the phases, with each successive load
present this is not feasible commercially. The regulator needs to being assigned to the next phase. This results in the system being
have a sufciently quick response to load changes, and this either balanced or only slightly unbalanced. The local utility has
voltage regulator offers sub-cycle response. However, this would little difculty in mitigating unbalance by changing the connection
be limited by the sampling rate of the measurement data received of any particular ICP.
by the regulator. At present most smart-meters only save
15-minute or half-hourly voltage readings. The meters do have the
capability to record at a higher sample rate, however, the amount 3.2 Voltage regulator and PV systems
of data transfer needed becomes immense if each smart-meter in a
network sends several readings a second. This would undoubtedly The regulator is modelled as a device which clamps its output voltage to
overload the communication channels for most advanced metering the desired setpoint. This matches the testing in the laboratory.
infrastructure in New Zealand. Laboratory measurements of a 5 kW PV inverter were used as the
It therefore becomes important to know how to set the regulator basis of the PV model, which was modelled at an output power of
based on the available data (such as current or a voltage), and to 3.7 kW. A stochastic modelling approach was taken for the uptake of
know whether additional sensors are worthwhile. Once the PV units due to the uncertainty regarding which customers will adopt
appropriate control strategies have been established, it is possible PV. The approach is similar to that taken for electric vehicle studies
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
This is an open access article published by the IET under the Creative Commons 3817
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
Fig. 1 Illustration of the benet of using a smart transformer with inbuilt voltage regulator
[1619]. The solar PV installations were distributed randomly steady-state node voltages in each scenario which fall outside the
throughout the network. The PV systems are modelled as xed 6% threshold, multiplied by a weighting and summed over all the
current injections; however, the current spectra were shifted between cases. Hence a low value corresponds to less steady-state voltage
iterations in accordance with the voltage angle (using the previous problems, which in turn corresponds to effective regulator control
iteration estimate). By doing this the correct phase relationship settings. The ultimate goal is therefore to minimise the measure of
between the terminal voltage and current spectrum is maintained. performance by nding optimal control settings. The scenarios and
weightings chosen are given in Table 2.
The weighted measure (M) used in Sections 4 and 5 to evaluate
3.3 Power-flow control strategies is therefore
A network admittance matrix based unbalanced three-phase
power-ow developed in MATLAB was used to solve for the M = P(1) 0.1 + P(2) 0.2 + P(3) 0.3 + P(4) 0.4 (1)
voltages in each LV network. The loads were represented as
constant power loads, although modelling them as constant Where P(s) is the proportion of steady-state voltage violations for
impedance loads was also performed for comparison. At peak scenario s (s = 1, 24). Hence four scenarios are considered resulting
load, the difference between the two methods did not exceed a in P(1), P(2), P(3)&P(4), which are proportion of node voltages
volt. The program converts the load power into equivalent outside the 6% threshold. These are then weighted and summed to
impedances, solves the linear system of equations for voltages, give the overall measure of performance. A lower weighted measure
then uses the estimated voltages to recalculate the load (M) is preferable, as it indicates less nodes with voltage violations.
impedances. The algorithm iterates until the power mismatch at
every node is less than 0.1 W [12].
4 Results
3.4 Measurement of performance
4.1 Control optimisation
The voltage regulator must be controlled in a way which allows it to
cope with both network extremes: reverse power-ow and peak load. Many different control algorithms were trialled, including choosing
Therefore an evaluation measure is needed to take account of the voltage setpoint based on:
different network conditions, assess the regulators performance in
each case, and come up with a unied measure of performance. Full network information
The measure of performance is assessed by the number of Current magnitude
One or more voltage sensors
Current magnitude and network impedance estimate (line droop
compensation)
Power per ICP
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
3818 This is an open access article published by the IET under the Creative Commons
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
To understand how the ideal regulator output setpoint varies with based on current direction was the best option, and Fig. 6 shows
network conditions, a very large sample (many cases with different how the two setpoints were tuned. The ideal setpoints are therefore
system conditions) was run. As an example, Fig. 3 shows the ideal 1.05 p.u. (241.5 V) for forward current, and 0.97 p.u. (223.1 V) for
setpoint against normalised load, where the x coordinate represents reverse power-ow. A low voltage setpoint for forward current or
the normalised load, which is the load as a proportion of the peak a high setpoint for reverse current perform poorly, as expected.
loading (1 = peak loading condition). The y coordinate is the Fig. 5a displays the results for using a full voltage range for the
voltage that the voltage regulator needs to produce in order to setpoint alongside the basecase as well as strengthening the supply
eliminate all steady-state voltage violations. Note that each point (see scenario ix). The scenarios displayed are:
in the gure corresponds to a network where the voltage regulator
successfully corrected the voltage violations. Cases where it was
not possible for the regulator to prevent any steady-state voltage (i) Full voltage range based on full voltage data (i.e. all node
violations are not shown (i.e. no suitable setpoint could be found). voltages). With the voltage at all nodes monitored the voltage
Some additional correlations are given in Appendix. regulators setpoint is adjusted to minimise the number of nodes
These correlations allow a possible control setting to be with voltage violations.
formulated, such as the line shown in Fig. 4, which will give a (ii) Full voltage range with setpoint depending linearly on current
satisfactory outcome for most cases (even though not optimum for magnitude (at the transformer).
a given scenario). Finding the exact optimum is difcult, (iii) Full voltage range with setpoint depending linearly on one
especially in view of the fact that in many networks there is a voltage sensor (at the end of the line).
range of possible regulator output voltages which ensure that all (iv) Full voltage range with setpoint depending on one voltage
steady-state voltage levels are within the limit. This analysis gives sensor (at the end of the line), however, the voltage regulator
guidance as to what inputs to the controller are well-correlated to setpoint is only adjusted if a voltage violation is detected (to
the ideal setpoint (and are therefore important to have). It also remove it).
enables an informed design of regulator control curve. (v) Full voltage range setpoint depending on current at the
transformer (using modelled curve). The modelled curve comes
from the previous results used to identify the setpoints that
removed all voltage violations (section 4.1). This is illustrated in
4.2 Overview the appendix where Fig. 10a shows the results for ideal setpoint
against transformer current as well as a tted curve for control
Fig. 5 shows the effectiveness of some of the various control purposes.
strategies simulated. All control strategies (except the basecase and (vi) Full voltage range setpoint depending on setpoint matrix. This
network reinforcement options) involve either using; full output is formed by scaling the transformer current by the number of ICP
voltage range (output voltage is innitely adjustable between 150 and using the modelled curve.
and 315 V) for the setpoint (with the setpoint depending linearly (vii) Full voltage range setpoint depending linearly on proportion
on some input parameter), which are displayed in Fig. 5a, or two of maximum load at the transformer. This means a single straight
voltage setpoints, shown in Fig. 5b. These are further line tted to the scatter plot of Fig. 3.
differentiated by how the setpoints are determined. For example, (viii) viii.Full voltage range setpoint based on modelled curve and
algorithm 1 in Fig. 5b switches between the two setpoints based proportion of maximum load. This modelled curve is the
on whether the current is above or below a predetermined value piecewise linear characteristic displayed in Fig. 4.
(offset). Similarly, for the full voltage range algorithms, an (ix) Effect of strengthened supply only. This means reducing the
equation involving current and/or voltage is used to calculate the supply impedance as seen from the LV side of the transformer.
setpoint. The best control algorithm not requiring any further This was modelled by reducing the transformer impedance and
sensors to be installed is algorithm 8 (Fig. 5a). maintaining the LV voltage at 415 V via the voltage regulator.
The particular voltage regulator tested has an inbuilt feature where (x) Basecase result.
it can switch between two pre-determined setpoints based on a signal
from an external relay. There are denite nancial advantages from
using such a simple control scheme, and research was performed to For Fig. 5b the algorithms displayed are:
nd the optimal values for the setpoints, as well as what criteria the
setpoints should be switched on. Fig. 5b showed that switching
(i) Switching between two setpoints based on current (compared
with an offset).
Fig. 3 Ideal setpoint vs proportion of peak load Fig. 4 Illustration of a simple control strategy for the regulator
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
This is an open access article published by the IET under the Creative Commons 3819
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
Fig. 6 Tuning one voltage setpoint for
a Usual load conditions
b Reverse power conditions
The optimised control strategy has been determined in the case of:
Fig. 8 shows the effectiveness of the voltage regulator under two
conditions; removing undervoltages with loading (Fig. 8a) and
Full output voltage range and full measurement data (ideal
removing overvoltages due to PV installations (Fig. 8b). In the PV
regulator in terms of control).
case the load is set to 20% of peak, as in [12, 13].
A full output voltage range but no additional sensors (no voltage
data).
Output conned to two setpoints and no additional sensors (no
voltage data). 4.3 Voltage regulator placement
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
3820 This is an open access article published by the IET under the Creative Commons
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
Fig. 7 Illustration of the benet of a smart transformer with inbuilt voltage
regulator (a) View 1, (b) View 2 Fig. 8 Illustration of the benet of the voltage regulator
a With inbuilt voltage regulator (view 1) a With load
b With inbuilt voltage regulator (view 2) b With PV
feeder, Fig. 9 shows that this is not as effective as having the voltage The network has traditionally been designed for a transformer
regulator at the transformer. This is because urban networks in New secondary voltage above 1.0 p.u.
Zealand the target market for the voltage regulator usually have On average the network will be drawing power rather than
two or more feeders per transformer and although the voltage supplying it.
regulator partway down the longest feeder xes voltage violations
for this feeder the other feeders may well be experiencing voltage
problems. A single voltage regulator at the transformer is One key nding is that the method used to design the regulator
sufcient, as clearly demonstrated in Fig. 9. controller (e.g. shown in Fig. 4) simplies the design signicantly.
In practice it could be possible to implement this with a learning
algorithm with the regulator, where additional voltage sensors are
5 Discussion and conclusions placed on the network for a training period. These sensors would
communicate with the regulator software, in order to build up a set
There were two major goals of this work: the development of an of data points as in Fig. 3. After the training period is complete, a
optimal control strategy for the voltage regulator; and an best-t line to the data points could be used to control the regulator.
evaluation of the effectiveness of the voltage regulator and hence Another key nding was that using the transformer current as a
smart transformer in the New Zealand power distribution system. proportion of maximum current as the input was superior to other
Almost all the regulator impact study results presented can also be system inputs. These included using a single voltage sensor at the
extended to any voltage regulator which allows its output voltage end of the longest feeder, or using the current measurement
to be controlled in real-time. without normalisation. In terms of practical implementation, the
maximum current of the network transformer needs to be known.
This could be provided by linking the maximum demand indicator
5.1 Control strategy of the transformer to the voltage regulator to provide this
information, or the regulator microcontroller could update this
The optimal control strategies have been found and are summarised value over time as it already measures the current. Hence this
in Table 3. maybe a practical and cost-effective algorithm.
Note that the full output voltage range controller is not symmetric This controller relies on the distribution network designs being
around a 1.00 p.u./zero current point. This is a result of two facts: consistently suited to its typical and maximum loading. Then the
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
This is an open access article published by the IET under the Creative Commons 3821
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
Table 3 Controller descriptions
Normalised Voltage (full range), Voltage (switched setpoint),
current p.u. p.u.
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
3822 This is an open access article published by the IET under the Creative Commons
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)
5.2 Effectiveness and Energy Technologies (APCET), Mylavaram, Andhra Pradesh, 24 August
2012, pp. 16.
7 Gallego, L., Padilha-Feltrin, A.: Voltage regulator modeling for the three-phase
The regulator is effective in maintaining an appropriate LV voltage, power ow in distribution networks. IEEE/PES Transmission and Distribution
solving (in simulation) 9697% of previously-existing voltage level Conf. and Exposition, Latin America, Bogota, 1315 August 2008, pp. 16
violations depending on which of the proposed algorithms is chosen. 8 Toh, C.W., Lock, C.S., Chen, S., et al.: A voltage quality improvement solution
It should also be noted that by clamping the output voltage, the using step dynamic voltage regulator. 2000 Int. Conf. of Power Systems
Technology (PowerCon 2000), Perth Australia, December 2000, vol. 2,
voltage regulator effectively removes the transformer impedance pp. 983988
from the circuit. This strengthens the supply to the loads 9 Le, A.D.T., Muttaqi, K.M., Negnevitsky, M., et al.: Response coordination of
connected to the regulator. The benet of this alone is quite distributed generation and tap changers for voltage support. Australasian
signicant in the distribution network, as can be seen in Fig. 5. Universities Power Engineering Conf., 2007, 912 December 2007, pp. 17
10 Periera, P., Canha, L., Malbradt, R., et al.: Optimization of voltage regulators settings
In terms of placement, placing the regulator right at the and transformer tap zones in distribution systems with great load variation using
transformer (i.e. right after the LV winding) makes commercial distribution automation and the smart grids initiatives. Eighth Int. Conf. on the
sense. The results clearly show that this is sufcient to keep the European Energy Market (EEM), Zagreb, Croatia, 2527 May 2011
voltage within range for the vast majority of LV networks. 11 Kongtripop, S., Premrudeepreechacharn, S., Kasirawat, T., et al.: Effect of
distributed generation on very long distribution line with automatic voltage
In summary, distribution network voltage regulators are a regulator. Sixth Int. Conf. on Electrical Engineering/Electronics, Computer,
promising option for dealing with the challenges associated with Telecommunications and Information Technology (ECTI-CON 2009), Chonburi,
DG. In the short-term it may be worth the investment for urban Thailand, 69 May 2009, pp. 190193
networks prone to overvoltage or undervoltage and not limited by 12 Watson, J.D., Watson, N.R., Santos-Martin, D., et al.: Low voltage network
modelling. EEA Conf. and Exhibition 2014, Auckland, NZ, June 2014, pp. 115
conductor or transformer ratings. 13 Watson, J.D., Watson, N.R., Santos-Martin, D., et al.: Impact of solar
photovoltaics on the low-voltage distribution network in New Zealand, IET
Gener. Transm. Distrib., 2015, 10, (1), pp. 19
6 Acknowledgments 14 Nan, J., Hou-jun, T., Jian-yu, L., et al.: Dynamic voltage regulator based on PWM
AC chopper converter: topology and control, WSEAS Trans. Syst., 2010, 9, (5),
pp. 539548
The authors wish to record their appreciation of Mr Peter Leece, 15 Khan, M., Rana, A., Dong, F.: Improved ac/ac choppers-based voltage regulator
CEO (ETEL Ltd) and Mr Gopalan Ramesh, Group-Manager designs, IET Power Electron., 2014, 7, (8), pp. 19892000
Engineering (ETEL Ltd) for their support of this work. The 16 Papadopoulos, P., Skarvelis-Kazakos, S., Grau, I., et al.: Impact of residential
charging of electric vehicles on distribution networks, a probabilistic approach.
authors thank Orion for the use of their system data. 2010 45th Int. Universities Power Engineering Conf. (UPEC), 31 August3
September 2010, pp. 15
17 Papadopoulos, P., Skarvelis-Kazakos, S., Grau, I., et al.: Predicting electric
vehicle impacts on residential distribution networks with distributed generation.
7 References IEEE Vehicle Power and Propulsion Conf., 13 September 2010, pp. 15
18 Frame, D., Ault, G.W., Huang, S.: The uncertainties of probabilistic LV network
1 Miller, A., Williams, J., Wood, A., et al.: Photovoltaic solar power uptake in New analysis. IEEE Power and Energy Society General Meeting, 2226 July 2012,
Zealand. Electricity Engineers Association (EEA) Conf. and Trade Exhibition, pp. 18
2014 19 Frame, D., Ault, G.W.: Exploring the uncertainties of probabilistic LV network
2 New Zealand Government: Electricity (safety) regulations 2010. Available at http:// analysis. CIRED 2012 Workshop Integration of Renewables into the
www.legislation.co.nz/regulation/public/2010/0036/latest/DLM2763501.html [accessed Distribution Grid, 2930 May 2012, pp. 14
14 October 2015]
3 Jan-E-Alam, M., Muttaqi, K.M., Sutanto, D.: Assessment of distributed generation
impacts on distribution networks using unbalanced three-phase power ow
analysis. PES General Meeting, 2011 IEEE, 2429 July 2011, pp. 18 8 Appendix
4 Agalgaonkar, Y.P., Pal, B.C., Jabr, R.A.: Distribution voltage control considering
the impact of PV generation on tap changers and autonomous regulators, IEEE
Trans. Power Syst., 2014, 29, (1), pp. 182192
For these runs the voltage regulator was given full information (all
5 Farag, H.E., Abdelaziz, M., El-Saadany, E.F.: Incorporating voltage regulator and the system voltages) so it can determine the ideal setpoint to
load models in unbalanced power ow studies of active distribution systems. IEEE prevent any steady-state voltage violations. Fig. 10a shows the
Power and Energy Society General Meeting, San Diego, CA, 2226 July 2012, ideal setpoint against transformer current. As expected, when there
pp. 17
6 Reddy, P.U., Reddy, M.L., Sivanagaraju, S., et al.: Optimal location of voltage
is reverse power ow (negative current) the ideal setpoint has to
regulators in unbalanced radial distribution system for loss minimization using drop to ensure the end of the feeder does not experience an
particle swarm optimization. 2012 Int. Conf. on Advances in Power Conversion over-voltage.
IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 15, pp. 38163823
This is an open access article published by the IET under the Creative Commons 3823
Attribution-NoDerivs License (http://creativecommons.org/licenses/by-nd/3.0/)