146 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 4, NO.

2, JUNE 2018

A Virtual Energy Storage System for Voltage

Control of Distribution Networks
Saif Sabah Sami, Student member, IEEE, Meng Cheng, Jianzhong Wu, Member, IEEE
and Nick Jenkins, Fellow, IEEE

Abstract—Increasing amounts of distributed generation (DG) Several studies have been conducted to investigate the
connected to distribution networks may lead to the violation use of energy storage systems, switched capacitors and DG
of voltage and thermal limits. This paper proposes a virtual active and reactive powers for the distribution network voltage
energy storage system (VESS) to provide voltage control in
distribution networks in order to accommodate more DG. A control [3]. A single 6 MW/10 MWh battery energy storage
VESS control scheme coordinating the demand response and system was installed in a distribution network to defer the
the energy storage system was developed. The demand response substation upgrade of adding a third 38 MVA transformer and
control measures the voltage of the connected bus and changes the to provide voltage support [4]. In [5], a centralised coordinated
power consumption of the demand to eliminate voltage violations. voltage controller of multiple battery energy storage system
The response of energy storage systems was used to compensate
for the uncertainty of demand response. The voltage control of and transformers with an on-load tap changer (OLTC) were
the energy storage system is a droop control with droop gain proposed to solve over-voltage and voltage unbalance prob-
values determined by applying voltage sensitivity factors. The lems caused by DG in the distribution network. The proposed
control strategy of the VESS was applied to a medium-voltage method does not involve a coordination among battery energy
network and the results show that the control of the VESS not storage system units. Instead, a coordinated control of multiple
only facilitates the accommodation of higher DG capacity in
the distribution network without voltage violations or network battery energy storage systems for voltage control of lowvolt-
reinforcements but also prolongs the lifetime of the transformer’s age networks is presented in [6]. The coordinated centralised
on-load tap changer. controller determines which battery energy storage system will
be used to solve voltage problems based on all the units’ state
Index Terms—Demand response, distributed generation, of charge and voltage sensitivity factors.
distribution network, energy storage system, voltage control,
virtual energy storage system. However, costs remain the main barrier to the large-scale
deployment of energy storage systems, in addition to the costs
of the information and communication technologies (ICT)
infrastructure for the central controller implementation.
I. I NTRODUCTION It is estimated in [7] that demand response has the potential
to reduce the energy storage system market size by 50% in
M ODERN distribution networks are witnessing signifi-
cant challenges to control the network voltage due to
changes in generation mix and demand. Decarbonisation of
2030. In [8], the peak reduction from flexible commercial
and industrial loads is forecasted to be approximately 10% of
heat and transport sectors supported by the growing number GB’s power system peak load in 2030. In recent years, several
of electric heat pumps and electric vehicles may cause under- studies were undertaken [9] investigating demand response
voltage problems. In contrast, the connection of Distributed abilities to provide ancillary services to the power system.
Generation (DG) may create over-voltage problems. In Great These services include voltage control of the distribution
Britain’s (GB) power system at present, only 18% of DG network. In [10], a centralised control scheme was proposed.
are fully visible to the system operator [1]. Around 40% The control monitors the bus voltage through remote terminal
of the renewable energy generation in GB is connected to units (RTU) and determines the required load curtailment of
distribution networks [2]. This can pose a serious threat to the customers participating in a distribution network program
distribution network voltage management, which results in based on voltage sensitivity factors. However, the challenges
a slower progress of integrating DG into the distribution facing distribution networks are the uncertainty of the response
network. and the consequent reduction in the diversity among these
flexible loads after the response provision.
Manuscript received October 6, 2016; revised February 24, 2017, accepted
April 17, 2017. Date of publication June 30, 2018; date of current version Several studies proposed a centralised coordinated control
December 20, 2017. The work is supported in part by the Higher Committee of demand response and battery energy storage system units
for Education Development in Iraq (HCED), by the RESTORES project under
the grant (No. EP/L 014351/1) of the UK-EPSRC, by the JUICE project under to minimise electricity costs in buildings [11] and to reduce
the grant (No. EP/P003605/1) and by the P2P-SmarTest project under the grant operational costs and embedded diesel generation emissions
of EU commission. in microgrids [12]. In [13], a centralised coordinated control
All authors are with the Institute of Energy, School of Engineering, Cardiff
University, Cardiff CF24 3AA, UK. algorithm of demand response and battery energy storage
DOI: 10.17775/CSEEJPES.2016.01330 system was presented for tie-line smoothing of a microgrid
2096-0042 © 2016 CSEE
SAMI et al.: A VIRTUAL ENERGY STORAGE SYSTEM FOR VOLTAGE CONTROL OF DISTRIBUTION NETWORKS 147

with an integrated DG. The results illustrate that coordinating is the heater state (Sfinal = 1 if heater is ON and Sfinal = 0 if
with demand response can significantly reduce the required heater is OFF), U (Wm−2 ·◦ C−1 ) is the overall heat transfer
size of the energy storage syste [13]. coefficient, A (m2 ) is the area of the tank surface, T (◦ C)
A virtual energy storage system (VESS) consisting of a is the internal temperature of the tank and Tamb (◦ C) is the
demand response and an energy storage system was developed ambient temperature.
to support the distribution network voltage and hence allows The heat transfer inside a tank leads to a temperature change
more DG integration in the distribution network. The VESS dT (◦ C). The internal temperature change can be associated
concept and its potential applications are first introduced. with the net heat transfer as:
Then, modelling and control of the VESS components and dT
VESS control scheme are presented. A population of industrial Pnet = cv × m × (4)
dt
bitumen tanks (BT) and batteries energy storage systems were
where cv (Jkg−1 ·◦ C−1 ) is the specific heat capacity of the tank
used to demonstrate the performance of the proposed voltage
and m (kg) is the mass. Combining (1)–(4), the tank internal
control scheme of the VESS. Two types of DG, solar and
temperature variations are depicted by a first-order differential
wind generation, and the VESS are connected to a medium-
equation (5)
voltage (MV) network of the United Kingdom generic distri-
bution system (UKGDS). The VESS control scheme operates dT P × Sfinal U × A × (T − Tamb )
= − (5)
cooperatively with on-load tap changers to ensure that no dt cv × m cv × m
voltage hunting will take place. The proposed VESS control The general solution of (5), which is the variation of tem-
was evaluated by time series analysis through different seasons perature over time, can be depicted by exponential functions
of a year. depending on the state of Sfinal . It is required that the tank
internal temperature be maintained all the time between its low
II. VESS C ONCEPT AND P OTENTIAL A PPLICATIONS set-point (Tlow , typically 150◦ ) and its high set-point (Thigh ,
A. VESS Concept typically 180◦ ). Therefore, the BT thermodynamic model was
developed using (6)–(7):
A VESS aggregates miscellaneous controllable components
t
of an energy system to form a single entity, which can −τ
Sfinal = 1 : T(t) = 184.68 − 34.68 × e ON (6)
behave similarly to a large capacity energy storage system with (t−t
− τ ON
)

reduced capital costs. Examples of such components include Sfinal = 0 : T(t) = 145.32 + 34.68 × e OFF (7)
flexible loads with thermal storage such as electric heaters, where τON and τOFF are the time constants which were
and DGs such as combined heat and power units (CHP) or obtained through field measurements to be half of the ON
conventional energy storage systems. The VESS allows those and OFF periods (tON and tOFF ) of each BT [16]. According
components to access the electricity and ancillary markets in to the field tests in [15], for a population of BTs, the ON-
order to provide transmission and/or distribution level services. period and OFF-period were randomly distributed within the
B. Potential Applications range of 30 min to 360 min and of 60 min to 1140 min.
By aggregating different types of energy resources, the B. Model of Energy Storage System
VESS can be characterized as a high-power and high-energy A simplified model of a battery energy storage system
density energy storage system. Hence, its potential applica- model was developed in [17], which consists of a generic
tions extend over widespread multi-disciplines of the power battery model and a simplified power electronics model. The
system [14]. These include but are not limited to, providing generic battery model (“Module (Battery cells)”) in Fig. 1
energy arbitrage, facilitating renewable integration in distri- is composed of a controllable voltage source, a controllable
bution network, deferring the transmission and distribution current source and a resistance connected in series. The
systems reinforcements and providing ancillary services such charging and discharging characteristics are assumed similar.
as frequency response, voltage support and power quality The simplified power electronic converters model is a first-
improvements. order lag to represent the delays in the converters control loop.

III. M ODELLING OF VESS C OMPONENTS Converter PBESS

Pref 1 1 Module(s)
A. Model of Demand Response Units + Tdelay s
− (Battery cells)
A thermodynamic model depicting variations of the temper- Qref QBESS
+ 1 1
ature of industrial BT with time was developed based on [15]. − Tdelay s
BT heat supply power Psupply (W), its heat loss power Ploss
(W) and its net heat transfer power Pnet (W) are:
Fig. 1. Simplified battery energy storage system model.
Psupply = P × Sfinal (1)
Ploss = U × A × (T − Tamb ) (2) In this study, lithium-ion batteries were modelled. A
(DC/AC) voltage source converter (VSC) was assumed to
Pnet = Psupply − Ploss (3)
connect the batteries to the distribution network. Hence, the
where P (W) is the power consumption of the heater, Sfinal active and reactive powers are controlled independently and
148 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 4, NO. 2, JUNE 2018

only the active power passes to the batteries and therefore On the contrary, BTs will be switched on in response to a
affects the stored energy in the battery energy storage system. voltage rise starting from the BT with the lowest temperature.
The reactive power is supplied by the converter. Furthermore, Therefore, the number of BTs committed to respond to voltage
the power converter internal losses were neglected and a 90% deviations increases linearly with the increase in voltage de-
roundtrip efficiency of the battery energy storage system was viations. Hence, all the demand response units are committed
considered. if the voltage accessed the limits. It was assumed that the
distribution network voltage limits follow the British Standard
IV. VOLTAGE C ONTROL OF VESS EN 50160 [18], a distribution network with voltage limits of
A. Voltage Control of Demand Response Units ± 6% of nominal value (i.e. 0.94 p.u.–1.06 p.u.), and voltage
control dead-band of ± 3% (i.e. 0.97 p.u.–1.03 p.u.) were
A distributed voltage controller was added to each bitumen used. BTs have low and high temperature limits of 150◦ C and
tank inherent temperature control. The voltage controller alters 180◦ C.
the demand response units’ power consumption based on local
voltage measurements as shown in Fig. 2. The temperature B. Voltage Control of Energy Storage System
control measures the temperature T of a tank and generates
Energy storage system (ESS) control methodology consists
state signals ST . The voltage control measures the bus voltage
of a main and a supplementary controllers. The main controller
V and generates state signal SHV and SLV . The final switching
drives the ESS’s active and reactive powers output in response
signal Sfinal to the heater is then determined by using logic
to voltage violations. The supplementary controller maintains
gates, which ensures the priority of the temperature control.
the ESS’s state of charge value within a certain range, which
Therefore, the extra voltage control will not undermine the hot
facilitates a secure, sustainable and efficient operation.
storage function of the BTs.
Energy storage system active and reactive power outputs
The voltage control algorithm switches the load on/off in
are determined by a droop control and the droop setting is
response to voltage deviations. The control algorithm assigns
obtained based on voltage sensitivity factors matrices.
a pair of voltage set-points, namely VON and VOFF , which
dynamically and linearly vary with the temperature of a BT. 1) Voltage Sensitivity Factors Matrices
For example, a BT will have a higher VON and lower VOFF Voltage sensitivity factors relate the change in voltage at
values if its temperature is higher than other BT temperatures. a bus to a change in active and/or reactive power(s) at other
The control algorithm continuously compares the measured buses in the network [19]. In a voltage sensitivity factor matrix,
voltage (V ) with the set-points. If voltage V is higher than a high voltage sensitivity factor implies that a change in active
VON , the voltage control generates a state signal SHV and the and reactive power at a bus drives a large change in voltage
load is switched on. In contrast, if voltage V is lower than at the corresponding bus.
VOFF , the voltage control generates a state signal SLV and the Voltage sensitivity factors matrices (10)–(11) were extracted
load is switched off. The linear variation of VON and VOFF fromthe Jacobian matrix in (8).
with temperature ensures that among a population of BTs,  ∂P ∂P 
   
following a voltage drop, the BT with the highest temperature ∆P  ∆δ
=  ∂δ ∂V  (8)

will be switched off first because it is most willing to be ∆Q ∂Q ∂Q ∆V
switched off because its temperature has already been high. ∂δ ∂V

Voltage Control
VON (V)
SHV
1.07
VON err1 1
0
1.03 T(°C) err1
Thigh Tlow
V
SHV
VOFF (V)
SLV
0.97 VOFF err2 1 SLV Logic Sfinal
0 Gates
ST
err2
0.93 T(°C)
Thigh Tlow
Distribution
network
Temperature control
ST
BTTemerature 1

0
Tlow Thigh T(°C)

Heater

Fig. 2. The control system of a flexible load.

SAMI et al.: A VIRTUAL ENERGY STORAGE SYSTEM FOR VOLTAGE CONTROL OF DISTRIBUTION NETWORKS 149

∆V = M · ∆P + N · ∆Q (9) and the other is below), the designated bus is the non-
voltage violation bus (YL or YH). ESS responds with
where enough power to push designated bus’s voltage deviation
#−1
to the limits, therefore reducing the other bus voltage
"  −1
∂P ∂P ∂Q ∂Q
M= − · · (10) violations (cells 4 and 9 in Table I).
∂V ∂δ ∂δ ∂V
• If no voltage violation occurs, i.e. all monitored bus
 −1
∂P ∂Q voltages are within limits. ESS takes no action (cells 6,
N = −M · · (11) 7, 8, 10, 11, and 13 in Table I).
∂δ ∂δ
2) Droop Control of Energy Storage System Using Voltage TABLE I
Sensitivity Factors E NERGY S TORAGE S YSTEM VOLTAGE C ONTROL S ETTINGS
A brief network analysis was carried out to identify the RH 1 CRH
most vulnerable buses with respect to voltage violations. These YH 2 CRH 6 0
buses are often the ones heavily loaded, connected to a large G 3 CRH 7 0 10 0
YL 4 CYL 8 0 11 0 13 0
amount of DG or connected through small capacity branches. RL 5 0 9 DYH 12 DRL 14 DRL 15 DRL
Then, these buses are equipped with remote monitoring de- RH YH G YL RL
vices to monitor and to send voltage values to the ESS
controller. ESS will respond by a droop control with respect to the
The ESS controller receives the voltages of the buses and designated bus i. The required voltage change at the designated
classifies the voltages into zones based on British standard EN bus i (∆Vi in p.u.) is determined first to calculate the required
50160 [18] as illustrated in Fig. 3 to the following: changes of the active power (∆PES in p.u.) and reactive power
1) Red zones (RH and RL) represent the voltage violation (∆QES in p.u) from ESS by using (12).
ranges, i.e. bus voltage violates/exceeds the ± 6% limits.
2) Yellow zones (YL and YH) represent the severe voltage ∆Vi = MiESS × ∆PESS + NiESS × ∆QESS (12)
deviation ranges, i.e. bus voltage largely deviates (equal or where MiESS is the voltage sensitivity factor relating the
larger than ± 3%) from the nominal value yet within the limits. change in the ESS active power to the change in bus i voltage
3) Green zones (GL and GH) represent the slight voltage and NiESS is the voltage sensitivity factor relating the change
deviation ranges, i.e. bus voltage marginally deviates (smaller in ESS reactive power to the change in bus i voltage.
than ± 3%) from the nominal value. Where CRH is charge ESS with respect to the bus with the
highest voltage violation, DRL is discharge ESS with respect
0.94 0.97 1 1.03 1.06
to the bus with the lowest voltage violation, CYL is charge
RL YL GL GH YH RH ESS with respect to the bus with the lowest voltage deviation,
Bus voltage (p.u.) DYH is discharge ESS respect to the bus with the highest
voltage deviation, 0 is No ESS power output.
Fig. 3. Classified bus voltage zones. The ESS reactive power response is prioritised above its
active power response in order to minimise the charging and
In the worst case, the network will suffer from high voltage discharging of the battery. The required reactive power can
deviations in both high and low voltage directions (i.e. two be obtained by setting the active power to zero in (12).
directions). To address this case, two buses with the two Alternatively, if the required reactive power is higher than
largest voltage deviations are selected and placed on the its rated value, the ESS will provide both active and reactive
vertical and horizontal axes of Table I. In the case of having power. The required active power can be obtained by setting
more than two monitored buses with voltage violations, it is the reactive power to its rated value in (12).
assumed that by mitigating the extremes cases, the less severe 3) Supplementary Control of Energy Storage System
voltage violations will also be released. One of these buses is When all monitored bus voltages are in the green zones
considered as the designated bus. The following rules were (Fig. 3), the ESS supplementary control restores the state of
applied to determine the designated bus and the charging/ charge to 50 ± 10%. The ESS charges/discharges using droop
discharging actions: control with respect to the monitored bus with the highest
• If both buses (in Table I) have a similar direction of voltage sensitivity factors. The ESS responds with enough
voltage diversions (i.e. both bus voltages are above/below power to push this bus voltage to the yellow zone (Fig. 3).
the nominal value), the designated bus is the bus with the This ensures that the consuming/absorbing ESS power will
largest voltage violation (RH or RL). ESS responds with not cause voltage violations. Only the active power of (12)
enough power to bring the designated bus voltage back is used and the reactive power is set to zero. Consequently,
within the limits (cells 1, 2, 3, 12, 14, and 15 in Table any forthcoming charging or discharging requirements are
I). expected to be met.
• If the two buses have the opposite direction of voltage
violations. ESS takes no action (cell 5 in Table I). C. Coordinated Voltage Control of VESS
• If the two buses have the opposite direction of voltage The coordination between demand response and energy
deviation (i.e. one bus voltage is above the nominal value storage system in the VESS is achieved by setting their
150 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 4, NO. 2, JUNE 2018

controllers with different time delay constraints. As a result, with the BS EN50160 standard [15]. In this study, a 3% of
they will not conflict with each other and cause voltage the energy demand in the distribution network was considered
hunting. The time delay constant coordination also considers as the electric vehicles charging loads. Electric vehicles loads
conventional voltage control equipment including the OLTC were assumed to be distributed in proportion to the peak load
and voltage regulators (VR). When a voltage violation occurs, at each load bus. An electric vehicle charging profile was
the voltage controllers of the demand response (DR) units obtained from [8].
respond first with a time delay constant τDR . If the voltage In the GB power system, DG connection to the distribution
violation continues, the energy storage system (ESS) with a network has to comply with engineering recommendation
time delay constant τESS (i.e. τESS > τDR ) will respond P2/6 [21] to ensure the security of the supply. Therefore, the
secondly. This procedure ensures that no voltage violation will total capacity of the DG connected to a distribution network
take place due to the uncertainty of demand response. Then is required not to exceed a level that allows the connection
if required, the on-load tap changer (OLTC) will take action of all units of DG at a rated output under minimum load
lastly with a time delay constant τOLTC (i.e. τOLTC > τESS ). conditions during an outage of the highest rated distribution
This, in turn, results in less OLTC actions. circuit (which is often the transformer). This is known as the
firm connection (FC) for generating plants. Accordingly, the
V. T EST S YSTEM UKGDS firm connection is limited to 37.74 MVA (i.e. 33
MVA transformer capacity plus 4.74 MVA minimum load).
In this paper, the performance of the proposed VESS voltage
control scheme was evaluated using a simplified medium- B. DG Allocations
voltage network from the United Kingdom generic distribution
Two types of DG were considered in this study. The first
system (UKGDS).
type is the domestic photovoltaic (PV). PVs were connected to
A. UK Generic Distribution System 11 buses which were assumed to be distributed in proportion
to the peak load at each bus. The PV penetration level at each
The 33 kV radial network has 16-buses and is supplied by
bus (PV%i ) is set according to (13).
two identical 33 MVA 132/33 kV transformers with OLTC.
The network is illustrated in Fig. 4 [20]. A VR transformer rated
PPVi
and a sub-sea cable are connecting bus 9 to bus 8. The network PV%i = (13)
Plmax
i
supplies a peak load of 38.94 MVA with a power factor of
rated
0.98 [20]. Half-hourly load profiles, DG generation profiles, where PPVi is the aggregated rated power of all PVs con-
and full network data were obtained from [20]. Based on the nected to bus i and Plmax
i is the max. total load of bus i.
load profiles, the network minimum load is 4.74 MVA with a The second type of DG is wind-farms. In addition to the
power factor of 0.98. The network voltage is required to be existing wind-farm at bus 16, four extra wind-farm locations
maintained within ± 6% of the nominal value in accordance were considered (Fig. 4). All wind-farms were located far from
the substation and at the end of individual feeders.
OLTC The network hosting capacity, defined as the total DG
3 transformer capacity under the minimum loading condition, was obtained
Wind farm
by using genetic algorithm (GA). GA maximises the total
2 wind-farms capacity at a given PV penetration level. The
5 Solar power Genetic Algorithm iteratively modifies a population of individ-
ual solutions. For each GA solution, MATPOWER [22] was
Energy storage
4 employed to find a load flow solution to check voltage and
Demand
response thermal limits constraints. Under 20% PV penetration level
1
6 7 Loads Bus index (i.e. 6.55 MW), 41.9 MW of the wind-farm’s capacity was
2
Monitoring allowed into the network. Hence, the network hosting capacity
device is 48.45 MW.
10 11
9
8 C. VESS Allocation
All network load buses (12-buses in total), were assumed
13 to have flexible loads except the main bus (i.e. bus no. 2).
33 kV The demand response penetration level at each bus (DR%i ) is
132 kV 16 12
defined in (14):
15
Plmax i
DR%i = DR (14)
14 Plmax
i

where Plmax
DR i is the aggregated max. power of all demand
33 kV
response units connected to bus i and Plmax
i is the max. total
load of bus i.
In this paper, a bitumen tank was used to demonstrate
Fig. 4. UK generic distribution system used. demand response capabilities. Under a 30% demand response
SAMI et al.: A VIRTUAL ENERGY STORAGE SYSTEM FOR VOLTAGE CONTROL OF DISTRIBUTION NETWORKS 151

penetration level, 9.8 MW of demand response aggregated the largest load) and bus 8 (which connects VR with several
capacity was connected to the distribution network. Assuming buses) were nominated to be monitored by the energy storage
all bitumen tanks have a typical power consumption of 40 kW, system controller. To validate selecting these buses, for the
245 bitumen tanks are connected in the distribution network. DG capacity of 60.25 MW and the lesser DR penetration level
In the presence of flexible loads (i.e. demand response (20% rather than 30%), buses 5, 7, and 8 had voltage violation
units), the distribution network hosting capacity for DG is problems. Bus 6 had the largest voltage sensitivity factors
increased. The hosting capacity for DG with demand response with respect to buses 5, 7, and 8, therefore the energy storage
was obtained by GA. That is, with 9.8 MW of demand system was connected to bus 6. Power and energy capacities
response aggregated capacity, 60.25 MW of DG is allowed of the energy storage system were determined through running
into the network. This 60.25 MW DG capacity is composed time series load flow of the distribution network with DG at
of, 20% PV penetration level (i.e. 6.55 MW) and 53.70 MW different periods of the year and checking for voltage viola-
of the wind-farm’s capacity. tions. Whenever a voltage violation occurs, the required energy
Fig. 5 shows the effect of different DG penetration levels storage system active and reactive powers were calculated to
(i.e. 0%–200% of the UKGDS hosting capacity for DG) on eliminate that violation according to (12). As a result, the
the UKGDS network maximum voltage deviation considering energy storage system power capacity was the maximum value
the cases with and without the voltage control schemes of of the calculated active power whereas its energy capacity was
the VESS. The 100% DG penetration level indicates the the integration of consecutive active power deployed/absorbed.
UKGDS hosting capacity for DG of 60.25 MW. Under low DG For the 9.83 MW capacity of demand response and 60.25 MW
penetration levels (i.e. less than 60%), the maximum voltage capacity of DG and through one summer week, one winter
deviation in the UKGDS network with and without the control week and one spring day period, the energy storage system
scheme of the VESS were similar. Under high DG penetration rated power and energy capacities were calculated to be 2.3
levels (i.e. 60% to 100%), the VESS voltage control scheme MW and 1.4 MWh. The maximum reactive power was limited
controlled the maximum voltage deviation to remain within to 0.8 times of the rated active power similar to the energy
the voltage limit (i.e. 0.06 p.u.) while the maximum voltage storage system installed in the CLNR project [5] to reduce the
deviation exceeded the voltage limit without the VESS voltage size of the converter installed.
control scheme. When the DG penetration level is higher than
the network DG hosting capacity (i.e. higher than 100%), the VI. C ASE S TUDY
maximum voltage deviation in the UKGDS network with and In this section the performance of the VESS coordinated
without the control scheme of the VESS were both breaching control was assessed against a base case with no VESS and
the voltage limits. With DG penetration levels higher than 100 the DG capacity was 60.25 MW.
% however, the voltage control scheme of the VESS reduced In the base case, theUKGDS network voltage was only
the maximum voltage violation more than the base case. Fig. 5 controlled by the OLTC and VR transformers. The controllers
therefore shows that the voltage control scheme of the VESS of the OLTC and VR discretely change the transformer tap
is able to reduce voltage violations caused by DG penetration position to regulate the transformer secondary voltage with a
levels higher than the UKGDS network hosting capacity (i.e. corresponding set-point and a bandwidth. Both the OLTC and
100% or 60.25 MW). VR have 20 tap positions (−0.85 + 0.05 as a per unit nominal
value). When the voltage diverts outside the bandwidth for
Maximum Voltage Deviation (p.u.)

0.1 a time longer than the controller time constraints (τOLTC

VESS and τVR ), the controller takes actions to return the voltage
0.09
No VESS
0.08 to the set-point by changing the transformer tap position in
0.07 proportion to the voltage diversion.
0.06 One voltage set-point for VR and two voltage set-points for
0.05 OLTC were determined, i.e. a high-load winter set-point and a
0.04 low-load summer set-point (Table II). To determine the OLTC
0.03 and VR set-points, one winter week and one summer week
0.02 load profiles (without DG) were used to adjust the set-points
0 50 100 150 200
DG Penetration Level (%)
so that no voltage violation will take place in the distribution
network.
Fig. 5. UKGDS maximum voltage deviations at different DG penetration TABLE II
levels. OLTC AND VR C ONTROLLER S ETTINGS

The energy storage system is utilised to compensate for Parameter Value (p.u.)
the demand response uncertainty. Its location was decided by Voltage set-point/winter 1.0265
OLTC Voltage set-point/summer 1.02
voltage sensitivity factors. Hence, the energy storage system Bandwidth 0.011
is connected to the bus having the largest values of voltage VR
Voltage set-point/all year 1.02
sensitivity factors with respect to the monitored buses. The Bandwidth 0.013
monitored buses are the most vulnerable buses with respect
to voltage violations. By analysing the UKGDS, bus 5 (with Power flow analysis with 1-min resolution was carried
152 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 4, NO. 2, JUNE 2018

out using MATPOWER. Bitumen Tanks and battery energy VESS coordination control scheme accounts for OLTC and
storage system models, and OLTC and VR controllers were VR controllers to eliminate any chance for controller conflicts
all implemented using MATLAB. To evaluate the proposed or voltage hunting among them. The time delay constraints
VESS control scheme performance over different seasons of for VESS elements and network transformers are specified in
the year, the following three periods were investigated. Results Table IV.
were compared with the base case in which no VESS was TABLE III
used. VOLTAGE S ENSITIVITY FACTORS (V LOTAGE P.U./P OWER P.U.)
Case 1: A spring day with high DG power output and low
network demand. Bus 5 Voltage Bus 6 Voltage Bus 8 Voltage
Case 2: A winter week with high wind energy generation. ESS Active Power 0.0805 0.564 0.0806
ESS Reactive Power −0.325 −0.681 −0.326
Case 3: A summer week with high solar generation.
Case 4: An autumn week with high wind and solar gener-
TABLE IV
ations and a medium demand. VESS AND T RANSFORMERS C ONTROL T IME D ELAY
Due to page limitations, the results of Case One will be
presented in detail and the results of the other cases will be Parameter Time Delay (min)
summarised in a table. τDR 1
τESS 2
Fig. 6 shows the total load, wind and solar generations as τVR 3
a base case without VESS. In this case a coincidence of high τOLTC 4
DG output and low load led to voltage violation in the first
five hours of the day (see Fig. 7). Fig. 8 shows the bus voltages with the proposed VESS. The
network complies with BS EN50160 and all bus voltages are
30 Total Load
within limits. Moreover, the proposed VESS control scheme
20
MVA

reduces the network OLTC and VR transformers actions in

10 all cases except the number of VR tap changes in summer
0 (Table V) and hence reduces their maintenance requirements
0 5 10 15 20 25 30 35 40 45
50 and prolongs their life. The number of tap changes of the
Wind
40 VR transformer with the voltage control scheme of the VESS
30
MW

20
increased slightly during the summer week compared with
10 the base case, due to the increase in voltage variations with
0
0 5 10 15 20 25 30 35 40 45 1.07
4 1.06
Solar
3 1.05
MW

2 1.04
1 1.03
Voltage (p.u.)

0 1.02
0 5 10 15 20 25 30 35 40 45 1.01
Time (half hour) 1
0.99
Fig. 6. Load, wind and solar generations for one spring day (Case One).
0.98
0.97
1.08 2 3 4 5 6 7 8 9
0.96
1.07 10 11 12 13 14 15 16
0.95
1.06 0 5 10 15 20 25 30 35 40 45
1.05 Time (half hour)
1.04
Voltage (p.u.)

1.03 Fig. 8. Voltage of each bus with the VESS for one spring day (Case One).
1.02
1.01 TABLE V
1 VESS C ONTROL P ERFORMANCE A NALYSIS
0.99
0.98
Test period Performance Indicator No With
0.97
VESS VESS
0.96 2 3 4 5 6 7 8 9
10 11 12 13 14 15 16 No. of voltage violation buses 44 0
0.95 Spring Day
0 5 10 15 20 25 30 35 40 45 No. of VR tap changes 32 21
Case One
Time (half hour) No. of OLTC tap changes 0 0
No. of voltage violation buss 7 0
Winter Week
Fig. 7. Distribution network bus voltages without VESS for one spring day No. of VR tap changes 191 162
Case Two
(Case One). No. of OLTC tap changes 4 0
No. of voltage violation buss 2 0
Summer Week
No. of VR tap changes 79 83
The VESS and its control scheme in section IV were Case Three
No. of OLTC tap changes 0 0
employed to control the network voltage. ESS monitors buses No. of voltage violation buses 2 0
Autumn week
5 and 8 voltages combined with its connected bus 6 and it Case Four
No. of VR tap changes 101 72
No. of OLTC tap changes 0 0
utilises the voltage sensitivity factors values shown in Table III.
SAMI et al.: A VIRTUAL ENERGY STORAGE SYSTEM FOR VOLTAGE CONTROL OF DISTRIBUTION NETWORKS 153

the voltage control scheme. An equivalent capacity of loads case one. The other cases showed similar behaviour. Limited
(i.e. 30% of the total load) in the base case was replaced by demand response is a result of reaching their temperature
bitumen tanks for demand response in the VESS. Therefore, limits and hence needs to be switched on/off to guarantee the
the response of the bitumen tanks combined with the low temperature performance of bitumen tanks.
demand (i.e. summer week demand) from the remaining loads In addition, it is noted that VESS can prevent the distribu-
(i.e. non-flexible loads) led to a higher total load variation than tion network reinforcements of a third 33MVA transformer and
the base case. This higher total load variation triggered slightly feeder. With the VESS, the maximum power flow through the
different voltage variations than the base case, which caused substation (37.6 MVA) was less than the network FC capacity
the slightly more number of tap changing actions. of 37.7 MVA. Whereas in the base case without VESS, the
Fig. 10 shows the distribution of buses voltages over the power flow exceeds (41.9 MVA) the network FC as shown in
spring day. The number of samples is 720 (i.e. 15 buses over Fig. 12.
48 time intervals). Fig. 10 shows that with the VESS voltage
control scheme, the voltage violations were eliminated and all 25
Without VESS
15
buses with very high voltages were reduced to the voltage With VESS
5
permissible limit (i.e. 1.06 p.u.). Other cases also showed

MVA
−5 −32
similar results. With the VESS voltage control scheme, the −15 −34
−36
number of buses with permissible high voltages increased (i.e. −25 −38
−40
between 1.03–1.06 p.u.). However, the ability of the UKGDS −35 −42
−44
network to host a greater DG generation capacity (i.e. 48.45 −45
0 5 10 15 20 25 30 35 40 45
MW without the VESS to 60.25 MW with the VESS) was not Time (half hour)
affected since all buses voltages were controlled in order to
remain within the limits. Fig. 12. Power flow through the substation in the UKGDS with and without
the VESS for one spring day (Case One).
160
No VESS VESS
140
Number of Buses

120 VII. C ONCLUSION

100
This paper introduces a VESS to provide voltage control
80
in a distribution network which facilitates the integration of
60
DG in future power systems. The VESS consists of flexible
40
loads and an energy storage system. Voltage control of each
20
0
element in the VESS was developed and coordinated in order
to minimise voltage deviations in the distribution network.
1.07
1.065
1.06
1.055
1.05
1.045
1.04
1.035
1.03
1.025
1.02
1.015
1.01
1.005
1
0.995
0.99
0.985
0.98
0.975
0.97
0.965
0.96

The local voltage control of bitumen tanks alters the power

Voltage (p.u.) consumption of flexible loads in response to voltage devia-
tions at the connected bus. The proposed distributed voltage
Fig. 10. The distribution of the UKGDS bus voltages for one spring day
(Case One). control of bitumen tanks has little impact on the primary
function of the loads. After the response of flexible loads,
Results in Fig. 11 show that most of the time energy storage the energy storage system local voltage controller monitors
system reactive power was sufficient to counteract over-voltage the two most vulnerable bus voltages and then determines the
caused by high DG following the limited demand response in charging/discharging actions using the droop control with a
droop setting obtained from the voltage sensitivity factors. The
0
ESSQ
energy storage system voltage control ensures a firm and linear
−0.5 response from the VESS against the voltage violations. In
MVr

−1
−1.5 addition, the VESS control scheme coordinates its components
−2 and the network inherent voltage control equipment through
0 5 10 15 20 25 30 35 40 45
0
the setting of time delays to avoid the voltage hunting with a
ESSP minimum ICT required.
−0.5
MW

−1 PVs, wind-farms and the VESS were optimally connected to

−1.5 the UKGDS distributed network. Case studies were undertaken
−2
0 5 10 15 20 25 30 35 40 45 to test the performance of the voltage control scheme of the
6 VESS in coordination with the controller of OLTC and VR
DRtotal
4
transformers under different scenarios of DG outputs. The
MW

2
0 results show that the VESS control scheme eliminates all
−2
−4 0
voltage violations and reduces the required number of OLTC
5 10 15 20 25 30 35 40 45
and VR actions, and consequently extends transformer life.
Time (half hour)
The proposed VESS represents an economic and technical
Fig. 11. Response from different VESS elements for one spring day (Case alternative to a substation upgrade in order to cap with the
One). DG integration in the distribution network.
154 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 4, NO. 2, JUNE 2018

ACKNOWLEDGMENT [19] Y. J. He, M. Petit, and P. Dessante, “Optimization of the steady voltage
profile in distribution systems by coordinating the controls of distributed
The first author would like to thank Marc Cheah and Chao generations,” in Proceedings of the 2012 3rd IEEE PES International
Long from Cardiff University for their interesting comments Conference and Exhibition on Innovative Smart Grid Technologies
(ISGT Europe), 2012, pp. 1–7.
and discussions. [20] Centre for Distributed Generation and Sustainable Electrical Energy,
“United Kingdom Generic Distribution System (UKGDS)” UK, 2006
[Online]. Available: https://github.com/sedg/ukgds.
R EFERENCES [21] Energy Networks Association, Engineering Directorate, “Engineering
Recommendation P2/6,” Energy Networks Association, Engineering
[1] National Grid “System operability framework 2016. UK electricity trans- Directorate, London, UK, 2006.
mission,” National Grid, UK, 2016 [Online]. Available: https://www. [22] R. D. Zimmerman, C. E. Murillo-Sanchez, and R. J. Thomas, “MAT-
nationalgrid.com/sites/default/files/documents/8589937803- POWER: Steady-state operations, planning, and analysis tools for power
SOF%202016%20-%20Full%20Interactive%20Document. pdf. systems research and education,” IEEE Transactions on Power Systems,
[2] ElectraLink. (2016, Oct. ). 40% of total renewable output no longer vol. 26, no. 1, pp. 12–19, Feb. 2011.
‘invisible’ with launch of new energy data analytics service[Online].
Available: https://www.electralink.co.uk/2016/10/40-total-renewable-out
put-no-longer-invisible-launch-new-energy-data-analytics-service-
electralink/.
[3] N. Mahmud and A. Zahedi, “Review of control strategies for voltage
Saif S Sami (S’14) received his BSc and MSc from
regulation of the smart distribution network with high penetration of
Baghdad University in 2004 and 2008 respectively.
renewable distributed generation,” Renewable and Sustainable Energy
Currently, he is a final year PhD student in the
Reviews, vol. 64, pp. 582–595, Oct. 2016.
School of Engineering, Cardiff University, Cardiff,
[4] UK Power Networks, “Smarter Network Storage - business model con-
U.K. His research interests include energy storage
sultation,” UK Power Networks, UK, 2013 [Online]. Available: https://
systems and demand response.
www.ukpowernetworks.co.uk/internet/en/community/documents/Smar
ter-Network-Storage-Business-model-consultation.pdf.
[5] P. F. Wang, D. H. Liang, J. L. Yi, P. F. Lyons, P. J. Davison, and P. C.
Taylor, “Integrating electrical energy storage into coordinated voltage
control schemes for distribution networks,” IEEE Transactions on Smart
Grid, vol. 5, no. 2, pp. 1018–1032, Mar. 2014.
[6] L. Wang, D. H. Liang, A. F. Crossland, P. C. Taylor, D. Jones, and N. S.
Wade, “Coordination of multiple energy storage units in a low-voltage
distribution network,” IEEE Transactions on Smart Grid, vol. 6, no. 6,
pp. 2906–2918, Nov. 2015. Meng Cheng received her B.Sc. degrees in electrical
[7] G. Strbac, M. Aunedi, D. Pudjianto, P. Djapic, F. Teng, A. Sturt, D. and electronic engineering from Cardiff University,
Jackravut, R. Sansom, V. Yufit, and N. Brandon, “Strategic assessment U.K. and North China Electric Power University
of the role and value of energy storage systems in the UK low carbon (Beijing), China, in 2011, and her Ph.D. degree
energy future,” Energy Futures Lab, Imperial College London, London, from Cardiff University in 2015. She is currently
UK, 2012. a Research Associate with Cardiff University. Her
[8] National Grid, “Future energy scenarios. GB gas and electricity trans- research interests include smart grids and dynamic
mission,” Nationalgrid, UK, Jul. 2016 [Online]. Available: http://fes. demand.
nationalgrid.com/media/1292/2016-fes.pdf.
[9] M. Behrangrad, “A review of demand side management business models
in the electricity market,” Renewable and Sustainable Energy Reviews,
vol. 47, pp. 270–283, Jul. 2015.
[10] A. Zakariazadeh, O. Homaee, S. Jadid, and P. Siano, “A new approach
for real time voltage control using demand response in an automated
distribution system,” Applied Energy, vol. 117, pp. 157–166, Mar. 2014. Jianzhong Wu (M’06) received his Ph.D. degree in
[11] E. Vrettos, K. L. Lai, F. Oldewurtel, and G. Andersson, “Predictive electrical engineering from Tianjin University, Tian-
Control of buildings for Demand Response with dynamic day-ahead and jin, China, in 2004. He is currently a Professor with
real-time prices,” in 2013 European Control Conference (ECC), 2013, the Institute of Energy, Cardiff University, Cardiff,
pp. 2527–2534. U.K. His research focuses on energy infrastructure
and smart grid.
[12] S. A. Pourmousavi, M. H. Nehrir, and R. K. Sharma, “Multi-timescale
power management for islanded microgrids including storage and de-
mand response,” IEEE Transactions on Smart Grid, vol. 6, no. 3, pp.
1185–1195, May 2015.
[13] D. Wang, S. Y. Ge, H. J. Jia, C. S. Wang, Y. Zhou, N. Lu, and X. Y.
Kong, “A demand response and battery storage coordination algorithm
for providing microgrid tie-line smoothing services,” IEEE Transactions
on Sustainable Energy, vol. 5, no. 2, pp. 476–486, Apr. 2014.
[14] M. Cheng, S. S. Sami, and J. Z. Wu, “Benefits of using virtual energy
storage system for power system frequency response,” Applied Energy, Nick Jenkins (M’81–SM’97–F’05) received
vol. 194, pp. 376–385, May 2017. his B.Sc. degree from Southampton University,
[15] M. Cheng, J. Z. Wu, S. J. Galsworthy, C. E. Ugalde-Loo, N. Gargov, W. Southampton, U.K., in 1974, his M.Sc. degree from
W. Hung, and N. Jenkins, “Power system frequency response from the Reading University, Reading, U.K., in 1975, and
control of bitumen tanks,” IEEE Transactions on Power Systems, vol. his Ph.D. degree from Imperial College London,
31, no. 3, pp. 1769–1778, May 2016. London, U.K., in 1986. He is currently a Professor
[16] M. Cheng, “Dynamic demand for frequency response services in the and Director, Institute of Energy, Cardiff University,
great britain power system,” Ph. D. thesis, Cardiff University, Cardiff, Cardiff, U.K. Before moving to academia, his
UK, 2014. career included 14 years of industrial experience,
[17] S. S. Sami, M. Cheng, and J. Z. Wu, “Modelling and control of multi- of which five years were in developing countries.
type grid-scale energy storage for power system frequency response,” While at university, he has developed teaching and
in 2016 IEEE 8th International Power Electronics and Motion Control research activities in both electrical power engineering and renewable energy.
Conference (IPEMC-ECCE Asia), 2016, pp. 269–273.
[18] Voltage characteristics of electricity supplied by public distribution
networks, British Standard EN 50160, 2007.