International Journal of Electrical Energy, Vol. 3, No. 1, March 2015
Voltage Reduction Field Test on a Distribution
Substation
E. Vega-Fuentes, J. M. Cerezo-Sánchez, S. L. Rosario, and A. Vega-Martínez
Department of Electronic Engineering and Automatics, Institute for Applied Microelectronics, University of Las Palmas
de Gran Canaria, Las Palmas G.C., Spain
Email: {eduardo.vega, juan.cerezo, sonia.leon, aurelio.vega}@ulpgc.es
the nature of the loads (type and mix) and the
topology of the network. It is hard to predict the
precise response of the network, summarized in its
CVR factor. For this issue, distribution loads
modeling [5] and estimation of its response to
voltage changes [6] are emerging research topics.
 The practice extent of voltage reduction is quite
limited to avoid exceeding statutory limits. The
advent of smart grids and the advanced metering
infrastructure generalization will enable increasing
this voltage reduction to allow bigger energy
savings [7]. Smart meters will provide real-time
voltage readings to guide voltage trimmings. The
consumer terminals with greater voltage drops and
the lowests service voltages in the network will be
monitored ensuring that limits are not exceeded.
Closed-loop control algorithms will adjust the
voltage according to remote metering.
This paper presents a voltage reduction field test on a
distribution substation. The experience pretends to
highlight the relationship between power demand and
voltage encouraged by results obtained in other networks
[8], [9] and aspires to answer how much energy saving is
achievable on the Canary Islands (Spain) distribution
networks with a voltage optimization strategy. It is in the
scope of the research of integrated Volt/VAr control in
isolated power systems [10]. The rest of this paper is
structured as follows. Firstly it is described the
experiment design, including substation selection process,
network features, voltage reduction extent determination
and savings expected. Then the results obtained are
presented and the CVR performance is evaluated. Next
section exploits the results of the analysis and estimates
the savings achievable by extrapolation of these results to
other feasible networks in the region. Finally, the results
are summarized and subjects for further research are
suggested.
Abstract—This paper highlights the relationship between
power demand and voltage level on the distribution network.
It shows how a small change in voltage with existing assets
delivers a very meaningful response in power demand and
in energy consumption. A 24 hour field trial at a utility
without any experience with Conservation Voltage
Reduction (CVR) is described. CVR effects are assessed by a
combined comparison-regression method resulting in at
least 9.42MWh energy saved during the test, equivalent to
the consumption of 848 households in the region. The
energy savings achievable within the network under this
study have been characterized and the attainable yearly
savings on the Canary Islands distribution networks, are on
the whole estimated. 
Index Terms—conservation voltage reduction (CVR),
demand reduction, distribution efficiency, energy saving,
field trial
I.
INTRODUCTION
The power demand of certain loads changes with
voltage [1]. This is the working principle of Conservation
Voltage Reduction (CVR) a mode of operation to reduce
energy consumption and peak demand. The voltage is
lowered on the distribution system in a controlled manner
so that service voltage at customer terminals is set within
the lower half of its statutory limits, causing no damage
to consumer appliances. CVR effects are evaluated by the
conservation voltage reduction factor (CVRf). It is
calculated from the percent energy savings divided by the
percent voltage reduction. The CVRf obtained in different
published assessments vary between 0.40 and 1.00 [2]
and the maximum achievable has been estimated in 2.00
[3].
Although this is a cost-effective way to save energy, it
is not adopted as widely as it could be. There are three
main arguments to justify why:
 There is a lack of incentives to reduce load
consumption, specially since costs are born by
distribution networks operators and benefits
accrue mainly for end users.
 Scepticism exists on CVR performance and its
potential negative effects on power quality and
system reliability [4]. CVR effects are heavily
dependent on network characteristics, mostly on
II.
As this was the first CVR field test carried out by the
utility, the approach was conservative. The main premise
was to avoid any claim from consumers and guarantee
power quality. The criteria to select the target for the trial
were: limited phase imbalance, good power factor, with
not very long feeders, high load density and low
distributed generation penetration. These characteristics
Manuscript received September 30, 2014; revised April 29, 2015.
©2015 International Journal of Electrical Energy
doi: 10.12720/ijoee.3.1.1-5
TEST DESIGN DESCRIPTION INTRODUCTION
1
International Journal of Electrical Energy, Vol. 3, No. 1, March 2015
factor by application of (1) with the load share of the
network under study. It is obtained a range
[0.1171~1.2538] with an average value of 0.7845, which
will be the CVRf for energy savings estimation.
are fuzzy but with the assistance of an experimented
operator at control centre the selection of a proper
substation was possible.
The studied distribution network starts with a
(66kV/22kV) power transformer equipped with on-load
tap changer (OLTC) to regulate the secondary bus
voltage. The reactive power compensation is
accomplished with 4.2MVAr non-stepped shunt capacitor
banks. There are 5 feeders connected to the busbar which
supply 104 transformer stations and through them 20,112
customers. The operator is unable to regulate each feeder
with different voltages. Voltage line drop is compensated
by off-load tap changer at transformer stations. The
voltage control devices in the network are solely OLTC
and the capacitor switch. There are neither regulating
transformers nor capacitors along the feeders.
There is no information about load composition though
a customer class breakdown may be derived from supply
contracts. Table I shows this breakdown detailed for
feeder and totalized to network level.
TABLE I.
Feeder 1
Feeder 2
Feeder 3
Feeder 4
Feeder 5
Total
TABLE II. CVR FACTOR BY SYNTHESIS FROM CUSTOMER CLASSES
AND MEAN POWER DEMAND REDUCTION ESTIMATIONS
Min.
Avg.
Max.
Commercial
[kW]
[%]
711.2
9.5
3,230.30
12.8
3,037.0
15.8
9,635.8
35.2
12,940.8
47.1
29,555.1
27.7
0.06
0.79
1.30
0.10
0.40
0.83
0.11
0.78
1.25
0.26
0.82
1.20
ΔPm
ΔPm
ΔPm
(YEAR)
(MONT
(DAY)
[kW]
H) [kW]
[kW]
55.30 52.15 59.95
370.55 349.41 401.71
592.22 558.43 642.02
(2)
Power demand mean values on yearly, ongoing month
and the same day of the week before the test basis were
13.32MW, 12.56MW and 14.44MW respectively. The
estimation of mean power demand reduction achievable
by this network is presented in the right section of Table
II on the three bases, using minimum, average and
maximum CVRf and the proposed percent voltage
reduction. The calculation for average CVRf and mean
daily power demand is shown in (3). As the most likely
behaviour of the network corresponds with that shown
the same day of the week before the test, the estimated
mean power demand reduction is in the range [59.95 ~
642.02 kW] with an expected value of 401.71 kW.
Finally, the expected energy saving (ΔE m) during one day
operating with voltage reduction strategy is obtained by
(4).
The reference value for the secondary busbar voltage
during regular operation is 21.0kV and the threshold
values are defined at 20.3kV (-3.33%) and 21.6kV
(+2.86%), in Spain statutory limits for service voltage
provided at consumer terminals are defined between ±7%
of the nominal declared value. As the test will be driven
by a Volt/VAr controller which had demonstrated a good
voltage performance in this network, characterized by an
average resulting voltage equal to the objective
considering its measurement resolution (100V) and a
mean deviation of just 79.2V [10], the voltage will be
trimmed to 20.4kV. This way, it is ensured that lower
operation limit is not overpassed and the service voltage
at consumer terminals is kept within regulatory terms.
The average voltage on this busbar during 2013 was
21.2025kV. During ongoing month (January), it was
21.1793kV and on the last registered day it was
21.1457kV. Expecting an average voltage of 21.15kV,
the trim will be 750V and the percent voltage reduction
will be 3.5461%. The prediction for energy savings will
be estimated on this basis.
The expected circuit-level CVRf may be estimated by
synthesis from customer classes as a linear combination
of CVR factors and load shares of each one [11]:
ΔPm DAILY = ΔPm(%) ·Pm DAILY / 100 = 0.4017MW
ΔEm= ΔPm ·t = 9,640.95 kWh
(3)
(4)
According to the customer service protection policy,
the incidence management software will be monitored
during the test and any complaint or event in circuits
affected by the trial will be studied. Additionally,
personnel at the zone and maintenance crews will be
aware of the voltage reduction test to detect and alert
early about any matter related.
III.
TEST RESULTS
The voltage reduction field test started on Thursday
January 30th at 10:00 am, and it lasted 24 hours. It was
scheduled at that time to suit the personnel timetable and
allow the maximum attention during the first hours of the
trial. The selection of the day was random, it was not a
holiday and there were no special conditions that could
affect the normal service of the network such as weather
alerts or any work programmed along the feeders.
Fig. 1 shows the voltage profile during the test. The
red line is the reference value. There was absolutely no
incidence in any circuit dependent on the network. Over
voltage is appreciable from 22:41:36h till 06:57:56h. At
that time it was tap position 1, there were no more taps to
reduce voltage, and the capacitor shunt was already
(1)
where R, C and I represent the load share and CVRR,
CVRC and CVRI represent the CVR factor for residential,
commercial and industrial customers respectively. Left
section of Table II summarizes published CVRf values of
different customer classes [2] and the resulting CVR
©2015 International Journal of Electrical Energy
CVRf
ΔPm(%) = CVRf ·ΔU(%) = 2.7819%
Industrial
[kW]
[%]
26.4
0.3
1,982.4
7.9
281.5
1.4
1,518.5
5.5
366.2
1.3
4,175.0
3.9
CVRf = R·
CVRR + C·
CVRC + I·
CVRI
CVRI
With voltage percent reduction (ΔU%) and CVRf
known, the mean power demand percent reduction is
calculated by (2) resulting in 2.7819%.
CUSTOMER CLASS BREAKDOWN
Residential
[kW]
[%]
6,749.9 90.1
19,869.5 79.2
15,797.9 82.6
16,163.8 59.1
14,166.0 51.5
72,747.1 68.3
CVRR CVRC
2
International Journal of Electrical Energy, Vol. 3, No. 1, March 2015
switched off (capacitors were disconnected after first onpeak hour, at 13:54:35h when demand fell and voltage
started rising up), therefore the controller had no way to
regulate voltage so its profile reflects high voltage
evolution. Fig. 2 displays the voltage progress at the
transmission boundary. At that time the voltage was
already 67kV and it still rose to 68.2kV during the night.
These considerations regarding voltage optimization
performance will be taken into account when assessing
the CVR potential of the network.
Figure 4. Estimated daily voltage profile. Thursday to Friday, January
2014.
The estimated demand during the day of the test
without CVR in the pilot network is calculated applying
(6) to every record in the time series where subindex (Est)
refers to estimated values.
P1 Est = [(P1 / P2) / (U1 / U2)]Est ·(P2/U2) ·U1Est
The resulting demand curve is shown in Fig. 5. It is
compared with the power demand resulting with CVR.
The estimation error for the inference method was
quantified in ±0.86% [12]. The energy saved this day rose
at least 9.42MWh.
To estimate the CVR factor of the network, the
average percent voltage reduction is calculated. It is
presented in Fig. 6. Table III summarizes the results from
the comparison, these calculations are repeated discarding
records between 22:41:36 h and 06:57:56h to find out the
network CVR potential. The resulting CVRf estimation
for this network is 0.8211.
Figure 1. Voltage profile during the test.
Figure 2. Voltage profile at transmission boundary.
IV.
CVR ASSESS
The evaluation is made by the combined comparisonregression method presented in [12]. The relationship
between the two distribution networks connected to the
same substation through different power transformers is
obtained. It is called relative demand and when the
voltage effect is filtered, presents a daily pattern. It is
calculated by (5).
PR = (P1 / P2) / (U1 / U2)
(5)
Figure 5. Power demand profile comparison.
Fig. 3 shows the estimated relative demand during the
test if CVR had not been applied. It is the average from
the other curves the same week day as the test. The next
step to infere the demand in the pilot network without
CVR is to estimate the voltage profile in the busbar the
day of the trial. It is obtained from voltage behaviour
during the month where daily patterns are detected. The
voltage during the test if CVR had not been applied is
estimated as the average from the other curves the same
week day than the test, as illustrated by Fig. 4.
Figure 6. Voltage profile comparison.
TABLE III. CVR FACTOR CALCULATION
P [MW]
ΔU[%]
ΔP [%]
CVRf
14.036
-2.9711
-2.7216
0.9160
14.5493
21.1771
-0.86%
Discarding records between 22:41:36 h and 06:57:56 h
CVR 20.3688
16.7007
-3.6740
-3.0169
0.8211
w/o
17.3696
21.1457
CVR
-0.86%
CVR
w/o
CVR
Figure 3. Estimated daily relative demand profile. Thursday to Friday,
January 2014.
©2015 International Journal of Electrical Energy
(6)
3
Ū [kV]
20.5479
International Journal of Electrical Energy, Vol. 3, No. 1, March 2015
V.
848 households. This strategy may be implemented using
existing assets and though benefits accrue mainly for end
users, it may be used to defer network reinforcement
investments and to reduce power demand at the time of
system peaks. Anyway, to extend its deployment,
authorities and regulators should promote conservation
programs. A way could be rewarding utilities for the
energy saved.
The energy savings achievable in the network under
study have been characterized by a CVR factor of 0.8211.
A value well into the range predicted by synthesis from
customer classes, and only 4.66% above the average
published results. Further research in the CVR field may
include field validated load models for accurate
calculation of the energy conservation gains and new
voltage reduction field trials in other substations in order
to assess networks suitability for the strategy.
CVR ADVANTAGES
The daily average power demand reduction obtained
was 392.57kW, -2.72% demand reduction. This may be a
valuable strategy to defer network reinforcement
investments and to reduce power demand at the time of
system peaks. The daily energy savings are estimated at
9.42MWh. The average monthly household consumption
in the Canary Islands is 333.08kWh [13], so the voltage
reduction strategy equals the supply to 848 households.
These conclusions result from a conservative voltage
reduction strategy in just half a substation whereas in the
Canary Islands there are 54 substations. Assuming that at
least in 10% of them (5.5 substations) the strategy is
applicable (that means limited phase imbalance, good
power factor, not very long feeders, high load density and
low distributed generation penetration) the amount of
energy that can be saved easily is huge. It is shown in
Table IV in an annual basis.
With present OLTC performance, increasing voltage
reduction extent is not possible. Most of time the tap was
on its lower position. But with other taps distribution,
such as (+14 and -7) instead of present (+6 and -15) over
nominal 22/66kV ratio tap, or with a new agreement for
voltage regulation at transmission boundaries, the amount
of energy saved would only depend on the network’s
CVR factor and on the voltage reduction extent. That
would allow increasing the reduction extent, if not to 7%
for caution, to 5%. Table IV presents the yearly savings
achievable with this voltage reduction in the pilot
network and in 10% regional distribution networks where
(E) is the yearly energy saving. The calculations are on
the basis of the average power demand in 2013,
13.32MW.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
TABLE IV. YEARLY SAVINGS WITH CVR STRATEGY
ΔU [%]
Pilot Network
10% Distribution Networks
Pilot Network
10 % Distribution Networks
3
5
ΔPm [MW]
0.324
3.564
0.546
6.015
E [GWh]
2.838
31.220
4.790
52.694
[7]
[8]
When the utility completes its ongoing conventional to
smart meter changing process, customer terminals with
greater voltage drops and the lowests service voltages in
the network might be monitored and the voltage
reduction might be extended to regulatory or technically
achievable limits improving the energy efficiency of the
network.
VI.
[9]
[10]
[11]
CONCLUSIONS, DISCUSSION AND FURTHER
RESEARCH
[12]
A 24 hour voltage reduction field trial has been
performed in a distribution substation under real
conditions. Customers within the test area did not notice
any impact on their appliances. The pilot test shows that a
small change in voltage can deliver a very meaningful
response in power demand and in energy consumption.
The energy saved during the test has been estimated in at
least 9.42MWh, equivalent to the daily consumption of
©2015 International Journal of Electrical Energy
[13]
4
M. Chen, R. Shoults, J. Fitzer, and H. Songster, “The effects of
reduced voltages on the efficiency of electric loads,” IEEE Trans.
on Power App. Syst., vol. 101, no. 7, pp. 2158-2166, 1982.
Z. Wang and J. Wang, “Review on implementation and
assessment of conservation voltage reduction,” IEEE Trans. on
Power Syst., vol. 29, no. 3, pp. 1306-1315, 2014.
W. Ellens, A. Berry, and S. West, “A quantification of the energy
savings by conservation voltage reduction,” in Proc. IEEE Int.
Conference on Power Syst. Technology, 2012, pp. 1-6.
T. L. Wilson, “Measurement and verification of distribution
voltage optimization results,” in Proc. IEEE Power and Energy
Society General Meeting, 2010, pp. 1-9.
M. Díaz-Aguiló, J. Sandraz, R. Macwan, F. de León, D.
Czarkowski, C. Comack, and D. Wang, “Field validated load
model for the analysis of CVR in distribution secondary networks:
Energy conservation,” IEEE Trans. on Power Delivery, vol. 28, no.
4, pp. 2428-2436, 2013.
F. Lamberti, C. Dong, V. Calderaro, and L. F. Ochoa, “Estimating
the load response to voltage changes at UK primary substations,”
in Proc. IEEE Power and Energy Society Innovative Smart Grid
Technologies Europe, 2013, pp. 1-5.
M. A. Peskin, P. Powell, and E. J. Hall, “Conservation voltage
reduction with feedback from advanced metering infrastructure,”
in Proc. IEEE Transmission and Distribution Conference and
Exposition, 2012, pp. 1-8.
T. A. Short and R. W. Mee, “Voltage reduction field trials on
distributions circuits,” in Proc. IEEE Transmission and
Distribution Conference and Exposition, 2012, pp. 1-6.
K. Fagen and C. Bernier, “Efficiencies in distribution design and
operating practices mid-study analysis,” in Proc. IEEE Rural
Electric Power Conference, 2007, pp. B6-B6-14.
E. Vega-Fuentes, J. M. Cerezo-Sánchez, S. L. Rosario, and A.
Vega-Martí
nez, “Fuzzy inference system for integrated VVC in
isolated power systems,” International Journal of Artificial
Intelligence and Applications, vol. 5, no. 1, pp. 91-106, 2014.
D. Kirshner and P. Giorsetto, “Statistical tests of energy savings
due to voltage reduction,” IEEE Trans. on Power App. Syst., vol.
103, no. 6, pp. 1205-1210, 1984.
E. Vega-Fuentes, S. L. Rosario, J. M. Cerezo-Sánchez, and A.
Vega-Martí
nez, “Combined comparison-regression method for
assessment of CVR effects,” in Proc. Int. Sci. Conference on
Power and Electrical Engineering of Riga Technical University
2014, pp. 172-176.
Government of the Canary Islands and La Laguna University
General Found. (2008). Proyecto piloto sobre la caracterización de
los usos finales de la energía en diferentes tipos de consumidores
en Canarias. El consumo energético en los hogares canaries.
[Online].
Available:
http://www.gobiernodecanarias.org/energia/doc/
eficienciaenergetica/pure/caractusosfinales.pdf.
International Journal of Electrical Energy, Vol. 3, No. 1, March 2015
Eduardo Vega-Fuentes received his M.Sc. and Ph.D. degrees in
electrical engineering from the University of Las Palmas de Gran
Canaria (ULPGC), Spain, in 1998 and 2015 respectively. Since 2004 he
has been with the Department of Electronics Engineering and
Automatics in ULPGC as an Associate Professor in Systems
Engineering and Automatics. His main research interests include
distribution efficiency, optimal power system operation and distribution
automation.
Juan Manuel Cerezo-Sánchez received his M.Sc. degree in
Telecommunications Engineering from the University of Las Palmas de
Gran Canaria (ULPGC), Spain, in 1993. He is currently working
towards the Ph.D. degree in Telecommunications Engineering in the
ULPGC. He is in the Department of Electronics Engineering and
Automatics working as Professor at the same University. His research
interests are SCADA systems and Industrial Communications
Sonia León-Del Rosario received her M.Sc. degree in Electrical
Engineering from the University of Las Palmas de Gran Canaria
(ULPGC), Spain, in 1998. She is currently working towards the Ph.D.
degree in Electrical Engineering in the ULPGC. She is in the
Department of Electronics Engineering and Automatics at the same
university working as Associate Professor. Her research interests
include intelligent techniques applied to Forecasting and Power Systems.
Aurelio Vega-Martínez received his M.Sc. and Ph.D. degrees in
Electrical Engineering from the University of Las Palmas de Gran
Canaria (ULPGC), Spain, in 1987 and 1992 respectively. He is in the
Department of Electronics Engineering and Automatics of the ULPGC.
His main research interests include distribution automation, power
system operation and control systems design.
©2015 International Journal of Electrical Energy
5