IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO.

1, JANUARY 2015 209

Distributed Voltage Control with Electric

Springs: Comparison with STATCOM
Xiao Luo, Zohaib Akhtar, Student Member, IEEE, Chi Kwan Lee, Member, IEEE,
Balarko Chaudhuri, Senior Member, IEEE, Siew-Chong Tan, Senior Member, IEEE,
and Shu Yuen Ron Hui, Fellow, IEEE

Abstract—The concept of electric spring (ES) has been pro- The novel concept of electric spring (ES) has been pro-
posed recently as an effective means of distributed voltage posed as an effective means of distributed voltage control [2].
control. The idea is to regulate the voltage across the critical (C) The idea is to regulate the voltage across the critical loads
loads while allowing the noncritical (NC) impedance-type loads
(e.g., water heaters) to vary their power consumption and thus while allowing the noncritical (NC) impedance-type loads
contribute to demand-side response. In this paper, a comparison (e.g., water heaters) to vary their power consumption and thus
is made between distributed voltage control using ES against contribute to demand-side response [3], [4] as well. This would
the traditional single point control with STATic COMpensator allow and facilitate large penetration of intermittent renewable
(STATCOM). For a given range of supply voltage variation, the energy sources without requiring huge amounts of energy stor-
total reactive capacity required for each option to produce the
desired voltage regulation at the point of connection is compared. age to act as a buffer between supply and demand [5]. The
A simple case study with a single ES and STATCOM is presented basic proof of concept of ES has already been demonstrated
first to show that the ES and STATCOM require comparable through hardware experimentation with the developed proto-
reactive power to achieve similar voltage regulation. Comparison types [2], [6]. Distributed voltage regulation through collective
between a STATCOM and ES is further substantiated through action of a cluster of ESs, each employing droop control has
similar case studies on the IEEE 13-bus test feeder system and
also on a part of the distribution network in Sha Lo Wan Bay, also been illustrated [7].
Hong Kong. In both cases, it turns out that a group of ESs In this paper, the focus is to compare the effectiveness
achieves better total voltage regulation than STATCOM with less of single point voltage control using STATCOM against dis-
overall reactive power capacity. Dependence of the ES capability tributed voltage control using a group of ESs. The basis for
on proportion of critical and NC load is also shown. comparison is total voltage regulation [root mean square of
Index Terms—Demand response, electric springs (ES), STATic the deviation of the actual voltages from the rated (1.0 p.u)
COMpensator (STATCOM), voltage control, voltage regulation. values] achieved and the overall reactive capability required
for each option in order to achieve that [8], [9].
A number of papers [2], [5]–[7] have been published
I. I NTRODUCTION recently on the ES concept and its control. However, none
OLTAGE control in medium voltage (MV) or low voltage
V (LV) distribution networks is typically exercised through
transformer tap-changers and/or switched capacitors/reactors.
of those papers have focused on the collective performance
of multiple of ESs considering realistic distribution networks.
This paper demonstrates the effectiveness of multiple ESs
Sometimes a STATic COMpensator (STATCOM) is used working in unison through case studies on an IEEE test
for fast and precise voltage regulation, especially for the feeder network and also a part of a real distribution system
sensitive/critical loads [1]. in Hong Kong. The voltage regulation performance and total
reactive power requirement of a group of ESs in case of dis-
Manuscript received December 19, 2013; revised May 21, 2014; accepted tributed voltage control is compared against the single-point
July 14, 2014. Date of publication August 28, 2014; date of current
version December 17, 2014. This work was supported in part by the control using a STATCOM. In both cases, it turns out that
Commonwealth Fellowship, the Hong Kong Research Grant Council under a group of ESs achieves better total voltage regulation than
Grant HKU10/CRG/10; in part by the University of Hong Kong Seed STATCOM with less overall reactive power capacity.
Funds under Project 201111159239 and Project 201203159010, and in
part by the Engineering and Physical Sciences Research Council under
Grant EP/K006274/1. Paper no. TSG-00927-2013.
X. Luo, C. K. Lee, and S. C. Tan are with the Department of Electrical II. E LECTRIC S PRING (ES) C ONCEPT
and Electronic Engineering, University of Hong Kong, Hong Kong (e-mail:
[email protected]; [email protected]; [email protected]). Voltage control in LV and MV distribution networks and
Z. Akhtar and B. Chaudhuri are with the Department of Electrical and demand-side management (DSM) have traditionally been
Electronic Engineering, Imperial College London, London SW7 2AZ, U.K. treated and tackled separately. Voltage control is usually
(e-mail: [email protected]; [email protected]).
S. Y. R. Hui is with the Department of Electrical and Electronic achieved by control devices discussed in the previous sec-
Engineering, University of Hong Kong, Hong Kong; and also with Imperial tion. DSM, on the other hand, is employed in a more
College London, London SW7 2AZ, U.K. (e-mail: [email protected]). distributed fashion (often at the appliance level) and is
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. predicated on intelligence or communication facility in the
Digital Object Identifier 10.1109/TSG.2014.2345072 appliance [10]–[12].
1949-3053 c 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
210 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 1, JANUARY 2015

Fig. 1. Electric spring set-up for smart loads.

Fig. 3. System response following decrease in reactive power consumption

of the intermittent source from 467 to 110 VAr. (a) Non-critical load voltage.
(b) Critical load voltage. (c) Electric spring voltage. (d) Reactive power exchange.

energy source. The controllable source is capable of injecting

variable active and/or reactive power which causes the voltage
Fig. 2. Simulation set-up with an intermittent source and an equivalent power across the C load to fluctuate. For simplicity both C and NC
grid.
loads are represented by resistors although they do not have
to be necessarily resistive. The parameters used for the system
Alternatively, an integrated approach to voltage control and and the ES are the same as in [2] and are not repeated here
aggregated demand action could be achieved by separating the due to space restriction.
loads into critical (C) loads requiring constant voltage and un- The above system is modeled in MATLAB/SIMULINK
interrupted supply and NC, impedance-type loads. At times of using a controllable voltage source representation for both
generation shortfall or network constraint, the voltage of the ES and STATCOM. Modeling and control of ES is discussed
NC loads is reduced while regulating the voltages across the in [13]. The magnitude of the controllable voltage representing
C loads. This addresses the generation shortfall or network the ES is controlled using a PI controller to minimize the dif-
constraint and also facilitates better voltage regulation of the ference between the actual and reference values of the voltage
C loads through manipulation of the supply impedance voltage across the C load. Phase angle of the voltage source is locked
drop. in quadrature to the phase angle of series current to ensure
One way to exercise this control is to use the so-called ESs there is no active power transfer. The STATCOM is modeled
which are power electronic compensators that inject a voltage by a controllable voltage source in series with impedance. Its
with controllable magnitude VES in series with each NC load control circuit is very similar to that of ES except for the
to regulate the voltage VC across the C load as shown in Fig. 1. adjustments due to its parallel connection to the C and NC
The voltage VNC across the NC loads is thus controlled (within load.
allowable bounds) and the active power consumed by them
modulated. The series combination of the ES and the NC load
B. Voltage Suppress Mode
thus acts as a smart load which ensures tightly regulated voltage
across the C load while allowing its own power consumption to The voltage across the loads is increased above the nominal
vary and thereby, participate in demand-side response. Adding value (216 V) by reducing the reactive power absorption of
the voltage VES in quadrature with the current flowing through the renewable source. This is to test the ability of an ES and
the ES ensures exchange of reactive power only like conven- a STATCOM to suppress the voltage and regulate it at the
tional voltage compensators including STATCOM. For further nominal value. At t = 1.0 s, the reactive power absorption
details about ESs the readers can refer to [2] and [5]. by the intermittent renewable source is reduced from 467 VAr
down to 110 VAr. Without any voltage control, the load voltage
increases from the nominal value of 216 V up to 224 V as
III. ES VERSUS STATCOM
shown by Fig. 3(a) and (b). Both STATCOM and ES are able to
A. Test System restore the voltage across the C load back to the nominal value
In order to compare the voltage regulation performance of as shown by the overlapping blue and red traces in Fig. 3(b).
a single ES against that of a STATCOM, a simple test system The ES achieves this by injecting about 115 V in series with
as shown in Fig. 2 has been considered. It comprises of a the NC load the voltage across which drops to about 185 V
power source acting as the main power grid and a separate as shown by the blue traces in Fig. 3(a) and (c). In order to
controllable power source to emulate an intermittent renewable suppress the voltage, both ES and STATCOM absorb reactive
LUO et al.: DISTRIBUTED VOLTAGE CONTROL WITH ELECTRIC SPRINGS: COMPARISON WITH STATCOM 211

Fig. 5. System response for different distribution of noncritical and critical

Fig. 4. System response following increase in reactive power consumption
loads (NC:C). Disturbance is increase in reactive power consumption of the
of the intermittent source from 467 to 1100 VAr. (a) Noncritical load voltage.
intermittent source from 467 to 1100 VAr. (a) Noncritical load voltage. (b) Critical
(b) Critical load voltage. (c) Electric spring voltage. (d) Reactive power exchange.
load voltage. (c) Electric spring voltage. (d) Reactive power exchange.

power (as indicated by positive sign of Q) from the system than the STATCOM. This is due to the fact that an increase
as shown in Fig. 3(d) with ES requiring to absorb about 100 in ES voltage will result in a reduction of NC load voltage
VAr more than the STATCOM. which causes a decrease in active power consumption of the
It is observed that the reactive power consumed by ES to (resistive) NC load. Hence, the ES needs to produce less reac-
restore the C load voltage to normal value is higher than tive power than an equivalent STATCOM to restore the system
the reactive power consumed by STATCOM to achieve the voltage due to the similar arguments about the X/R ratio as
same voltage. This can be explained from Fig. 1. An increase mentioned earlier for the voltage suppress case.
in ES voltage will result in a decrease in NC load voltage.
This causes a decrease in the active power consumption of D. Proportion of C and NC Loads
the (resistive) NC load. In order to have a higher overall
active/reactive power consumption for the smart load, ES has An ES injects a voltage is series with the NC load in order
to consume more reactive power. Note that the X/R ratio is to regulate the voltage across the C load. The proportion of
not large (about 2) in this case which is why both active and the C and NC load is therefore, quite important toward the
reactive power affect the voltage regulation. effectiveness of an ES both in terms of its voltage regulation
capability and also the amount of reactive power (and hence
its rating) exchanged with the system. The reactive capability
C. Voltage Support Mode of an ES is governed by the product of the voltage it injects
To investigate the opposite effect of what was described and the current flowing through it (which is the same as the
in the previous subsection, the voltage across the loads current through the NC load). If the injected voltage increases,
is reduced by increasing the reactive power absorption of the voltage across the NC load and hence the current reduces
the renewable source. This is to test the ability of an ES which limits the reactive capability of an ES and thus its ability
and a STATCOM to support the voltage and regulate it to regulate the voltage across the C load.
at the nominal value. At t = 1.0 s, the reactive power For low proportion of NC load, the fidelity of current is
absorption by the intermittent renewable source is increased restricted which limits the capability of an ES compared to
from 467 to 1100 VAr. Without any voltage control, the the case when the proportion of NC load is relatively high.
load voltage is seen to drop from the nominal value of To verify this, simulations have been conducted with different
216 V to slightly below 190 V as shown by the green trace proportions of NC and C loads. The results are shown in Fig. 5.
in Fig. 4(a) and (b). It can be seen that for high proportion of NC load
As before, both STATCOM and ES are able to restore the (NC:C = 9:1) shown by the black traces, the C load volt-
voltage across the C load back to the nominal value as shown age is restored back to its nominal value, with only 80 V
by the overlapping blue and red traces in Fig. 4(b). The ES injected by the ES. This results in little change (from 216
achieves this by injecting about 150 V in series with the NC to 202 V) in voltage across the NC load. Voltage regulation
load the voltage across which drops to about 150 V as shown is similar for equal proportion of C and NC (NC:C = 5:5)
by the blue traces in Fig. 4(a) and (c). In order to suppress loads shown by magenta traces. However, the voltage across
the voltage, both ES and STATCOM inject reactive power (as the NC load is lower (about 140 V) than before due to larger
indicated by negative sign of Q) into the system as shown injected voltage (160 V) by the ES. Based on public statis-
in Fig. 4(d) with ES requiring to inject about 150 VAr less tics in Hong Kong [14], about 50% of loads (such as heaters,
212 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 1, JANUARY 2015

(i.e., capacitive) mode with a NC load of R–L type, the total

reactive power of the smart load QSL is given by
QSL = QES + QNC (5)
  
−VES ± VC2 − (VES cos θNC )2 + VES sin θNC
QSL =
ZNC
(a)   2
± VC2 − (VES cos θNC )2 + VES sin θNC
+ sin θNC .
ZNC
(6)
Similarly, for the ES in voltage suppress (i.e., inductive)
mode, we can write

VNC = ± VC2 − (VES cos θNC )2 − VES sin θNC (7)
and
  
VES ± VC2 − (VES cos θNC )2 − VES sin θNC
(b) QSL =
ZNC
Fig. 6. (a) Phasor diagram showing relationship between voltages across
  2
noncritical load, critical load, and ES. (b) Variation of reactive power of ES ± VC − (VES cos θNC ) − VES sin θNC
2 2
and smart load with respect to ES voltage for R–L and R noncritical loads.
+ sin θNC .
ZNC
air-conditioners, etc.) in domestics and commercial buildings (8)
can be considered as NC.
For low proportion of NC load (NC:C = 1:9), it is not possi- From (3), (6), and (8) it is clear that the reactive power of
ble to restore the voltage across the C load back to its nominal the ES and the smart load are both dependent on NC load
value as shown by the cyan trace in Fig. 5(b). This is because impedance (ZNC ). A decrease in the value of ZNC (increase
of the low fidelity in current which restricts the reactive capa- in the NC load) will result in an increase in reactive power.
bility of the ES to less than 100 VAr [Fig. 5(d)] for a maximum Hence, a higher proportion of NC load will increase the
possible ES voltage of 160 V. This demonstrates that the volt- effectiveness of an ES.
age regulation capability of an ES is dependent on the relative
proportion of NC and C load. Lesser the proportion of NC E. Reactive Power Limit of Smart Load
load, lower is the voltage regulation capability of an ES. As the For a fixed NC load impedance (ZNC ∠θ NC ) and a target C
second generation of ES with embedded energy storage [15] load voltage (VC = 1.0 p.u.), all the terms on the right hand side
has emerged, there would be more flexibility in control which of (3), (6), and (8) are constant except the ES voltage (VES ).
would be demonstrated in a future paper. Hence, QES and QSL can be expressed as functions of VES only.
The reactive power exchange with the ES depends on the Fig. 6(b) shows the variation of QES and QSL versus VES for
injected voltage VES and also on the impedance of the NC load. VC = 1.0 p.u., and ZNC = 1.0 p.u. for two different power fac-
Consider the circuit shown in Fig. 1. For a resistive–inductive tors of the NC load. In all cases the ES is considered to be in
(R–L) type NC load with impedance ZNC ∠ θ NC , the voltages voltage support (i.e., capacitive) mode as indicated by the neg-
VC , VES , and VNC are shown on the phasor diagram in Fig. 6(a) ative sign of QES . For a purely resistive NC load, QES and QSL
when the ES is working in voltage support (i.e., capacitive) are equal and are shown by the black trace in Fig. 6(b). QES
mode. From the phasor diagram, we can write and QSL for an R–L NC load with 0.95 power factor are shown
by blue and green traces, respectively. The figure is drawn only
VC2 = (VNC − VES sin θNC )2 + (VES cos θNC )2 (1)
 for nonnegative values of VNC phasor represented by (2).
VNC = ± VC − (VES cos θNC )2 + VES sin θNC
2 (2) It can be seen that beyond a certain point, increasing the ES
VES VNC voltage will result in a decrease in reactive power magnitude
QES = VES INC sin(−90o ) = −VES INC = − (3) due to a decrease of the current. Hence, it is essential to impose
ZNC
a limit on the output of the PI controller which determines the
V2
QNC = VNC INC sin θNC = NC sin θNC . (4) ES voltage magnitude, so that the voltage injected by the ES
ZNC does not go beyond the maximum reactive power (magnitude)
Here, QES and QNC are the reactive powers of the ES point on the curves shown in Fig. 6(b). It may also be noted
and the NC load, respectively. For a purely resistive NC that the maximum values of the two reactive powers will occur
load, the reactive power of the ES and the smart load will at different values of VES if the NC load is not purely resistive.
be equal. However, they would be different if the NC is In such cases, the limits of the PI controller should be based on
not purely resistive. If the ES is working in voltage support the maximum value of QSL . Also, it can also be seen that the
LUO et al.: DISTRIBUTED VOLTAGE CONTROL WITH ELECTRIC SPRINGS: COMPARISON WITH STATCOM 213

Fig. 7. Variations of the (a) voltage across the critical load, (b) voltages Fig. 8. Variations of the (a) voltage across the critical load, (b) voltages
across the noncritical load and the ES, and (c) reactive power of the ES and across the noncritical load and the ES, and (c) reactive power of the ES and
STATCOM as the reactive power absorption by the renewable source (at bus STATCOM as the active power generation by the renewable source (at bus 2,
2, Fig. 2) is changed from 150 to 1100 VAr. Fig. 2) is changed from 0 to 900 W.

reactive power output of the smart loads would be maximum voltage suppression (Q < 200 VAr) on the other hand, ES is
at different values of VES depending on the power factor of seen to require more Q than STATCOM. This is due to the
the NC loads. change in power consumption of the NC load (when ES is
active) as explained earlier in Sections II-B and II-C.
F. Variable Active and Reactive Power From Renewable Next, the reactive power absorption is fixed at Q = 467 VAr,
Source while the active power (P) generated at bus 2 is varied from
0 to 900 W. Without any voltage control, the voltage across
In this subsection, the result of varying the reactive power the loads increases with increase in active power generation
absorbed and the active power generated by the renewable (P) at bus 2 as shown by the green trace in Fig. 8(a).
energy source connected at bus 2 (see Fig. 2) is shown. First, One important point to note from Fig. 8(b) is that as power
the reactive power absorbed is varied between 150 and 1100 VAr generation from the renewable source at bus 2 increases,
keeping the active power generation fixed at zero. Without any the voltage across the NC load (and hence the active power
voltage control, the voltage across the loads reduces as the consumed by it) reduces in order to regulate the voltage across
reactive power absorption increases. This is shown by the the C load to its nominal value of 216 V. In such cases, the
green trace in Fig. 7(a) about the nominal voltage of 216 V. NC load voltage has to be lower than its nominal value for
For Q < 467 VAr, the actual voltage is higher than nominal a nonzero ES voltage. Hence, the active power consumed by
requiring voltage suppression while for Q > 467 VAr, the actual the NC load cannot increase above its nominal value. This
voltage is less than the nominal requiring voltage support. restriction can be overcome if the load has nonunity power
Voltage injected by the ES and the voltage across the NC factor in which case the two voltages are not constrained
load are shown in Fig. 7(b). For Q = 467 VAr, the voltage to be in quadrature. Alternatively, the ES can be allowed to
injected by the ES is almost zero while the voltage across inject a voltage with any phase angle (not just ±90 degrees)
the NC load is equal to the nominal value of 216 V. On either with respect to the current requiring exchange of both active
side of Q = 467 VAr, the ES injects a positive voltage, result- and reactive power with the system which is possible through
ing in a reduced voltage across the NC load such that the incorporation of energy storage (i.e., a battery) into the ES.
vector sum of the two equals the nominal voltage (i.e., 216 V) This type of ES with embedded energy storage is more versatile
which is maintained across the critical load. in terms of its capability to control the voltage while ensuring
The reactive power exchanged by the ES is compared power balance and hence regulate the system frequency and is
against that of a STATCOM to regulate the C load voltage referred to as version 2 or generation 2 of ES (ESv2) [15]. The
at 216 V. It can be seen that for voltage suppression (Q < scope of this paper is limited to reactive power only version
467 VAr), both of the ES and STATCOM absorbs VAr from (ESv1) [5] to ensure a fair comparison against STATCOM
the system (as indicated by the positive sign) while for voltage which only exchanges reactive power with the system.
support (Q > 467 VAr) they inject VAr into the system.
It should be noted that over the range of variation of Q
IV. C ASE S TUDY 1: IEEE 13-N ODE T EST F EEDER
absorption shown in Fig. 7(c), the reactive power exchanged
by the ES and the STATCOM are very similar. For higher A. Test Network
levels of voltage support (Q > 900 VAr), a STATCOM requires After comparing the performance of a single ES against a
more reactive power than an ES with the difference between STATCOM, the focus is on the collective action of a group
the two growing for larger Q absorption. For higher levels of of distributed ESs and how that compares against a single
214 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 1, JANUARY 2015

Fig. 10. Voltage regulation with distributed ESs and STATCOM following
5% reduction of the source voltage at bus 650.

Fig. 9. IEEE 13-node test feeder network with distributed representation of are regulated within the acceptable limit (5%). Nonetheless,
the LV side.
the regulation gets poorer away from the STATCOM loca-
tion. With ESs the voltage regulation is more uniform across
STATCOM. To investigate this, the IEEE 13-bus test feeder the LV feeder. A degree of voltage regulation can still be
system shown in Fig. 9 is considered [16]. The network has ensured even if one or more ES are out of operation. In order
two voltage levels 4.16 kV and 480 V with a distribution to demonstrate this qualitatively, the ES connected with bus
transformer connected between node 633 and 634. 4 is deactivated. It can be seen that the voltage regulation
In the original IEEE 13-node test feeder, the LV side is is still better than having no control at all, but understand-
represented by an aggregated load at bus 634. For the purpose ably worse than that with a STATCOM. There are only four
of this paper, the LV side has been modified to distribute the ESs in this system. In a larger system, we can have multiple
total load (160 kW with 0.825 lagging power factor) among ESs deactivated without making the system prone to voltage
four newly introduced LV bus bars labeled as 1, 2, 3, and changes.
4. The aggregated load (160 kW) connected at node 634 is The overall voltage regulation achieved in each case is com-
split equally among these four new nodes. The ratio of C to pared in terms of the root mean square of the deviation of the
NC loads is assumed to be 50:50. The LV distribution line actual voltages from the rated (1.0 p.u.) values which is termed
conductor dimensions are chosen based on the current ratings as total voltage regulation and defined in
of the loads and the conductor data and the distance between
the LV bus bars are provided in the Appendix. All other circuit Total Voltage Regulation

parameters are exactly the same as the feeder is set up to study Nb  2
i=1 Vrated_(p.u.) − Vactual_(p.u.)
unbalanced operation. For this paper, we consider only one = (10)
phase of the system as unbalanced operation is not the focus Nb
here. where, Nb is the total number of buses where voltage reg-
ulation is considered. The results are shown in Fig. 11 for
B. Voltage Support Mode both voltage support and voltage suppress (discussed in next
subsection) modes. It can be seen that the group of ESs
The collective action of the distributed ESs has been com-
achieves better voltage regulation than a STATCOM at bus
pared with a STATCOM installed on the MV side at bus 633.
633 [Fig. 11(b)]. Moreover, the total reactive capacity required
A 5% step reduction in the source voltage at bus 650 is con-
for the ESs is about six times less than that required by the
sidered. The comparison is based on the total reactive power
STATCOM [Fig. 11(a)].
required by the four ESs in order to achieve an acceptable
voltage regulation at the LV buses. Voltage regulation at a
C. Voltage Suppress Mode
particular bus is defined in (9) as the normalized difference
between the rated voltage (1.0 p.u.) and the actual voltage in Similar exercise, as in the previous subsection, has been
the event of a voltage disturbance repeated for over-voltage (voltage suppress) condition. A 5%
step increase in the source voltage at bus 650 is simulated.
|Vrated − Vactual |
Voltage Regulation = × 100% . (9) The voltage regulations with ESs and a STATCOM are shown
Vrated in Fig. 12. As before, voltage regulation with a STATCOM
The voltage regulation achieved at different LV buses is gets worse away from its connection point.
shown in Fig. 10. Without any voltage compensation the volt- Without any voltage compensation, the voltage regulation
age regulation becomes progressively poorer away from the is better away from the MV bus (bus 633) due to the natural
MV bus (bus 633) due to the voltage drop in the LV feeder. voltage drop across the LV feeder. With a group of ESs, the
In this case, the voltage regulation turns out to be unaccept- voltage regulation is more uniform which results in less than
ably high (>5%). With a STATCOM providing perfect (0) half of the total voltage regulation achieved with a STATCOM
voltage regulation at bus 633, the voltages at the LV bus bars as shown in Fig. 11(b).
LUO et al.: DISTRIBUTED VOLTAGE CONTROL WITH ELECTRIC SPRINGS: COMPARISON WITH STATCOM 215

Fig. 13. Single line diagram of a part of the distribution network from Sha
Lo Wan Bay, Lantau Island, Hong Kong.

The network data are provided in the Appendix. The param-

eters of the distribution lines are practical values, but the loads
are arbitrarily set because the actual load data are confiden-
tial due to privacy policy. There are 23 purely resistive loads
connected to the 220 V network. Each load has a rating of
30 kW which is assumed to have a 50:50 split between C and
Fig. 11. (a) Reactive power required. (b) Total voltage regulation achieved
collectively by all the distributed ESs and STATCOM under voltage support NC load. An ES is connected in series with each of the 23 NC
and suppress condition. loads.

B. Voltage Support Mode

To validate the collective performance of the ESs and com-
pare it with the voltage control of a STATCOM, a 5% step
reduction in the 11 kV substation (substation A) voltage has
been simulated. Voltages at all the load connection points
across the distribution network at Sha Lo Wan Bay (shown
in Fig. 13) are monitored.
The three subplots in Fig. 14 correspond to the cases with no
voltage compensation, with a STATCOM regulating the volt-
age at the 11 kV substation (substation A) and ESs connected
in series with all the NC loads at 220 V level. The distribution
of voltage is shown in Fig. 14 along the 11 kV feeder (x-axis)
Fig. 12. Voltage regulation with distributed ESs and STATCOM following and also along each of the 220 V feeders (y-axis). Without
5% increase in source voltage at bus 650. any voltage compensation [Fig. 14(a)] the voltage regulation
Moreover, the total reactive power consumption by the ESs is poor (>5%) getting worse as we move further away along
is less than 20 times that of a STATCOM. Thus, for both the 11 kV feeder and also the 220 V feeders due to natural
under-voltage and over-voltage conditions, a group of dis- voltage drop in the lines.
tributed ESs is shown to achieve better total voltage regulation The STATCOM regulates the voltage at substation A which
than a STATCOM with a total reactive capacity much less than results in very good regulation at bus 1 [Fig. 14(b)]. However,
that of a STATCOM [Fig. 11(a)]. the voltage regulation is poorer (but much better than the case
The study on the modified IEEE 13-node test feeder network without voltage compensation) further away along the 11 kV
confirms the following. and 220 V feeders.
1) Better total voltage regulation is achieved with a group In the case with ESs, the voltage regulation turns out to
of distributed ESs compared to a STATCOM although be better, especially at the loads which are at the far ends of
both are able to ensure acceptable regulation. the 220 V feeder. As the ES regulates the voltage by manip-
2) Total reactive capacity required by the group of ESs is ulating the voltage drop across the supply impedance, larger
significantly less than that of the STATCOM. impedance (for distant loads) improves the effectiveness of
ESs which is apparent from Fig. 14(c).
The distribution of the voltage across all the load buses of
V. C ASE S TUDY 2: D ISTRIBUTION N ETWORK IN S HA L O
Sha Lo Wan Bay distribution system is captured in terms of
WAN BAY, L ANTAU I SLAND , H ONG KONG
their mean and standard deviation in Fig. 15 for voltage support
A. Test Network and voltage suppress modes (discussed in the next subsection).
Another case study has been performed on a part of the For voltage support mode, the distributed ESs provide much
distribution network at Sha Lo Wan Bay in Lantau Island of better (lower average) and tighter (lower standard deviation)
Hong Kong. The objective is to compare the voltage regulation voltage regulation than a STATCOM [Fig. 15(a)].
performance of a group of ESs against a STATCOM. The This is further substantiated by the total voltage regulation
11 kV substations and a part of the 220 V feeder network as shown in Fig. 16(b) which shows ESs achieve three times
shown in Fig. 13 is considered for this paper. better total regulation than a STATCOM. Moreover, the total
216 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 1, JANUARY 2015

Fig. 14. Voltage regulation with distributed ESs and STATCOM following
5% reduction in source voltage at substation A. (a) No compensation device.
(b) STATCOM. (c) ESs.

reactive power capability required for the group of ESs is Fig. 15. Voltage distribution at different parts of the Sha Lo Wan distribution
about 14 times less than that of the STATCOM as shown in network under. (a) Voltage support. (b) Voltage suppress modes.
Fig. 16(a).

C. Voltage Suppress Mode

Similar exercise as above has been conducted to compare
the collective performance of the ESs and a STATCOM under
voltage suppress mode. A 5% step increase in the 11 kV
substation voltage has been simulated. The voltage regulation
performance is shown in Fig. 15(b) in terms of the mean and
standard deviation of the voltages at all the load buses.
It can be seen that voltage regulation without any voltage
compensation is within the acceptable (5%) limits. In this case,
the voltage regulation actually gets better away from the 11 kV
bus (substation A) due to the natural voltage drop across the
11 kV and 220 V feeders. Similar to the voltage support mode,
ESs provide much better (lower average) and tighter (lower
standard deviation) voltage regulation than a STATCOM.
The total voltage regulation shown in Fig. 16(b) depicts that Fig. 16. (a) Reactive power required. (b) Total voltage regulation achieved
collectively by all the distributed ESs and the STATCOM under voltage
the group of ESs achieves about two times better total regu- support and suppress condition.
lation than a STATCOM. The total reactive power capability
required for the group of ESs [Fig. 16(a)] is about 30 times
less than that of the STATCOM. VI. D ISCUSSION
The above case study on the Sha Lo Wan Bay distribution The case studies presented in this paper confirm the follow-
network in Hong Kong demonstrates the effectiveness of dis- ing.
tributed voltage control through a group of ESs under both 1) A group of distributed ESs is able to achieve better volt-
voltage support and suppresses modes. A group of distributed age regulation than a STATCOM. The reactive power
ESs achieves much better total voltage regulation compared to capacity of a STATCOM is not limited until the current
a STATCOM with much less reactive capability. limits are violated. In principle, a STATCOM can inject
LUO et al.: DISTRIBUTED VOLTAGE CONTROL WITH ELECTRIC SPRINGS: COMPARISON WITH STATCOM 217

approximation is valid without introducing much error:

(ZL ∠θL + ZS ∠θS ) ≈ ZL ∠θL . (12)
From (11) and (12), we can write
VL ∠0 ≈ VS ∠δ + IC ∠θC (ZS ∠θS ) (13)
VL ∠0 − VS ∠δ
IC ∠θC ≈ . (14)
(ZS ∠θS )
The phase angle θ C will be either 90o or -90o depend-
ing on the type of reactive power compensation required
(inductive or capacitive). The phase angle θ S is con-
Fig. 17. Simple circuit showing an ideal source connected with a fixed
impedance load through some series impedance and a voltage compensation
stant for a given X/R ratio of the feeder. From (14), it
device in parallel to the load. is evident that the magnitude of the compensation cur-
rent (IC ) required to restore the load voltage (VL ) back to
any amount of current (within its rated capacity) and the nominal value, in case of a change in source voltage
thus, any amount of reactive power. On the contrary, magnitude (VS ), is inversely proportional to the source
the reactive power capacity of an ES is limited. As the impedance (ZS ). For a given change in source voltage, a
voltage injected by an ES increases the voltage across higher series impedance magnitude (for longer distance
the NC load and hence the current through the ES away from the source) will require a smaller compen-
(as they are in series) decreases. At some point their sation current (which implies less reactive power) to
product, which is the reactive power reaches the max- restore the critical load voltage. Therefore, the farther
imum beyond which the ES cannot inject/absorb more the load is from the voltage source, the easier it is to
reactive power. Hence, its voltage regulation capability regulate the voltage with a less reactive power exchange.
is limited. However, if there are multiple ESs distributed 3) A STATCOM regulates the voltage at the point of con-
in the system, they can share the burden and this would nection but the load buses downstream will still have a
not necessarily be a problem. The capability of ESs to natural voltage profile where the voltage at far end could
regulate the voltage also depends on the relative propor- still be low even if the voltage at STATCOM bus is reg-
tion of C and NC loads. Higher proportion of NC loads ulated at 1.0 p.u. On the contrary, a group of distributed
allows larger fidelity on the current and hence improves ESs with droop control also improves the voltages at the
the voltage regulation capability. far end resulting in a better total voltage regulation.
2) In general, it is easier to regulate the voltage at locations 4) STATCOMs do central voltage control typically at the
which are electrically farther away from a stiff voltage point of coupling with the MV/LV feeders. So the entire
source which in this case would be the upstream MV/HV downstream feeders are vulnerable to voltage problems
network. As the ESs are located farther away from the if the STATCOM is out of operation. The ESs provide
upstream MV network than a STATCOM, there is less distributed voltage control, and failure of one/two does
burden on the ESs and collectively, they require less not make the entire feeder system susceptible to voltage
reactive power than a STATCOM installed upstream. problems.
This can be explained analytically by considering a 5) For an R–L type NC load, better voltage regulation
simple circuit shown in Fig. 17. could be achieved in voltage suppress mode as both ES
An inductive load (ZL ∠θ L ) is supplied from an ideal (working in inductive mode) and the NC load consume
voltage source though a series impedance (ZS ∠θ S ) repre- inductive reactive power and thereby, aid each other.
senting a feeder. A reactive power compensation device Same is true for voltage support mode (ES in capaci-
is connected in parallel to regulate the load voltage tive mode) in case of an R–C type NC load. In voltage
(VL ∠0) to the nominal value in case fluctuations in the support mode with an R–L type NC load, the total reac-
source voltage (VS ). If IC ∠θ C is the current injected by tive power of the smart load is equal to the difference
the compensation device, the load voltage (considered between the reactive power produced by the ES and that
as the reference) can be expressed as follows: consumed by the load which reduces the voltage regula-
tion capability compared to the case of a purely resistive
ZL ∠θL
VL ∠0 = VS ∠δ load.
ZL ∠θL + ZS ∠θS
+IC ∠θC (ZL ∠θL ||ZS ∠θS ) . (11)
VII. C ONCLUSION
In power systems, the series impedance is typi- In this paper, a comparison is made between distributed volt-
cally much lower compared to the load impedance age control using ES against the traditional single point control
(ZL >> ZS ). Under normal operation, the voltage across with STATCOM. For a given range of supply voltage variation,
the load impedance is between 0.95 and 1.05 p.u. while the total voltage regulation, and the total reactive capacity
the voltage drop across the series impedance varies required for each option to produce the desired voltage regu-
in the range of ±0.05–0.10 p.u. Hence, the following lation at the point of connection are compared. A simple case
218 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 1, JANUARY 2015

study with a single ES and STATCOM is presented first to [2] S. Y. Hui, C. K. Lee, and F. F. Wu, “Electric springs: A new smart grid
show that the ES and STATCOM require comparable reac- technology,” IEEE Trans. Smart Grid, vol. 3, no. 3, pp. 1552–1561,
Sep. 2012.
tive power to achieve similar voltage regulation. Comparison [3] A. Brooks, E. Lu, D. Reicher, C. Spirakis, and B. Weihl, “Demand
between a STATCOM and ES is further substantiated through dispatch,” IEEE Power Energy Mag., vol. 8, no. 3, pp. 20–29,
similar case studies on the IEEE 13-bus test feeder system May/Jun. 2010.
[4] D. Westermann and A. John, “Demand matching wind power genera-
and also on a part of the distribution network in Sha Lo Wan tion with wide-area measurement and demand-side management,” IEEE
Bay, Hong Kong. In both cases, it turns out that a group Trans. Energy Convers., vol. 22, no. 1, pp. 145–149, Mar. 2007.
of distributed ESs requires less overall reactive power capac- [5] C. K. Lee and S. Y. Hui, “Reduction of energy storage requirements
in future smart grid using electric springs,” IEEE Trans. Smart Grid,
ity than STATCOM and yields better total voltage regulation. vol. 4, no. 3, pp. 1–7, Sep. 2013.
This makes ESs a promising technology for future smart grids [6] C. K. Lee, B. Chaudhuri, and S. Y. Hui, “Hardware and control imple-
where selective voltage regulation for sensitive loads would be mentation of electric springs for stabilizing future smart grid with
intermittent renewable energy sources,” IEEE J. Emerg. Sel. Topics
necessary alongside demand-side response. Power Electron., vol. 1, no. 1, pp. 18–27, Mar. 2013.
[7] C. K. Lee, N. R. Chaudhuri, B. Chaudhuri, and S. Y. R. Hui, “Droop
control of distributed electric springs for stabilizing future power grid,”
A PPENDIX IEEE Trans. Smart Grid, vol. 4, no. 3, pp. 1558–1566, Sep. 2013.
The parameters for the modified LV segment (buses 1–4) [8] J. Dixon, L. Moran, J. Rodriguez, and R. Domke, “Reactive power com-
pensation technologies: State-of-the-art review,” Proc. IEEE, vol. 93,
of the IEEE 13-node test feeder system are provided here. no. 12, pp. 2144–2164, Dec. 2005.
[9] T. J. E. Miller, Reactive Power Control in Electric Systems. Hoboken,
NJ, USA: Wiley, 1982.
[10] P. Palensky and D. Dietrich, “Demand side management: Demand
response, intelligent energy systems, and smart loads,” IEEE Trans. Ind.
Informat., vol. 7, no. 3, pp. 381–388, Aug. 2011.
[11] M. Parvania and M. Fotuhi-Firuzabad, “Demand response scheduling by
stochastic SCUC,” IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 89–98,
Jun. 2010.
[12] M. Pedrasa, T. D. Spooner, and I. F. MacGill, “Scheduling of demand
side resources using binary particle swarm optimization,” IEEE Trans.
The parameters for the section of the Sha Lo Wan Bay Power Syst., vol. 24, no. 3, pp. 1173–1181, Aug. 2009.
distribution network considered for this paper are given below [13] N. R. Chaudhuri, C. K. Lee, B. Chaudhuri, and S. Y. R. Hui,
“Dynamic modeling of electric springs,” IEEE Trans. Smart
for the 11 kV and 220 V sections. Grid, to be published [Online]. Available: http://ieeexplore.ieee.org/
Data for 11 kV section of Sha Lo Wan Bay network. stamp/stamp.jsp?tp=&arnumber=6873343
[14] (2013). “Hong Kong energy end-use data 2013,” Electr. Mech.
Serv. Dept., Hong Kong Special Admin. Reg., [online]. Available:
http://www.emsd.gov.hk/emsd/eng/pee/edata_1.shtml
[15] S. C. Tan, C. K. Lee, and C. K. R. Hui, “General steady-state anal-
ysis and control principle of electric springs with active and reactive
power compensations,” IEEE Trans. Power Electron., vol. 28, no. 8,
pp. 3958–3969, Aug. 2013.
[16] (2006). IEEE 13-Node Test Feeder [Online]. Available: http://ewh.
ieee.org/soc/pes/dsacom/testfeeders/

Xiao Luo received the B.Eng. and the M.Sc.

degrees in electrical engineering from Hunan
University, Hunan, China, and Hong Kong
Polytechnic University, Hong Kong, in 2009 and
2010, respectively. He is currently pursuing the
Ph.D. degree from the Department of Electrical and
Electronic Engineering, University of Hong Kong.
His current research interests include renewable
energy and power quality improvement.

Data for 220V section of Sha Lo Wan Bay network.

Zohaib Akhtar (S’14) received the B.Sc. (Hons.)

and the M.Sc. degrees in electrical engineering
from the University of Engineering and Technology,
Lahore, Pakistan, in 2008 and 2011, respectively,
and is currently pursuing the Ph.D. degree from
Imperial College London, London, U.K.
R EFERENCES His current research interests include renewable
[1] N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concepts and energy, power distribution, and smart grids.
Technology of Flexible AC Transmission Systems. Piscataway, NJ, USA:
IEEE Press, 2000.
LUO et al.: DISTRIBUTED VOLTAGE CONTROL WITH ELECTRIC SPRINGS: COMPARISON WITH STATCOM 219

Chi Kwan Lee (M’08) received the B.Eng. and Siew-Chong Tan (S’00–M’06–SM’11) received the
the Ph.D. degrees in electronic engineering from B.Eng. (Hons.) and the M.Eng. degrees in elec-
the City University of Hong Kong, Kowloon, trical and computer engineering from the National
Hong Kong, in 1999 and 2004, respectively. University of Singapore, Singapore, in 2000 and
He was a Post-Doctoral Research Fellow at 2002, respectively, and the Ph.D. degree in elec-
the Power and Energy Research Centre, National tronic and information engineering from Hong Kong
University of Ireland, Galway, Ireland, from 2004 Polytechnic University, Hong Kong, in 2005.
to 2005. In 2006, he joined the Centre of Power From 2005 to 2012, he was a Research Associate,
Electronics, City University of Hong Kong, as a Post-Doctoral Fellow, Lecturer, and an Assistant
Research Fellow. From 2008 to 2011, he was a Professor with the Department of Electronic and
Lecturer of Electrical Engineering at the Hong Kong Information Engineering, Hong Kong Polytechnic
Polytechnic University, Hong Kong. He was a Visiting Academic at Imperial University, Hong Kong. In 2011, he was a Senior Scientist at the Agency for
College London, London, U.K., from 2010 to 2011. Since 2012, he has Science, Technology, and Research (A*Star), Singapore. He was a Visiting
been an Assistant Professor in the Department of Electrical and Electronic Scholar at the Grainger Center for Electric Machinery and Electromechanics,
Engineering, University of Hong Kong, Hong Kong. His current research University of Illinois at Urbana-Champaign, Urbana, IL, USA, in 2009,
interests include applications of power electronics to power systems, advanced and an Invited Academic Visitor at Huazhong University of Science and
inverters for renewable energy and smart grid applications, reactive power Technology, Wuhan, China, in 2011. He is currently an Associate Professor
control for load management in renewable energy systems, wireless power with the Department of Electrical and Electronic Engineering, University of
transfer, energy harvesting, and planar electromagnetics for high frequency Hong Kong, Hong Kong. His current research interests include the areas of
power converters. power electronics and control, LED lightings, smart grids, and clean energy
technologies. He co-authored the book Sliding Mode Control of Switching
Power Converters: Techniques and Implementation (CRC, 2011).
Dr. Tan serves extensively as a reviewer for various IEEE/IET transactions
and journals on power, electronics, circuits, and control engineering.

Shu Yuen Ron Hui (M’87–SM’94–F’03) received

the B.Sc. (Hons.) degree in electrical and electronic
engineering from the University of Birmingham,
Birmingham, U.K., in 1984, and the D.I.C. and
Ph.D. degrees from Imperial College London,
London, U.K., in 1987.
He currently holds the Philip Wong Wilson
Wong Chair Professorship with the University of
Hong Kong, Hong Kong. Since 2010, he has con-
currently held a part-time Chair Professorship of
Balarko Chaudhuri (M’06–SM’11) received the Power Electronics, Imperial College London. He has
Ph.D. degree in electrical and electronic engineer- published over 200 technical papers, including about 170 refereed journal
ing from Imperial College London, London, U.K., publications and book chapters. Over 55 of his patents have been adopted by
in 2005. industry.
He is currently a Senior Lecturer with the Control Prof. Hui was the recipient of the IEEE Rudolf Chope Research &
and Power Research Group, Imperial College Development Award from the IEEE Industrial Electronics Society and the IET
London. His current research interests include elec- Achievement Medal (Crompton Medal) from the Institution of Engineering
tric power transmission systems, control theory, and Technology in 2010. He is a Fellow of the Australian Academy of
smart grids, and renewable energy. Technological Sciences and Engineering. He is also the recipient of the
Dr. Chaudhuri is an Associate Editor of the 2015 IEEE William E. Newell Power Electronics Award. He is an Associate
IEEE S YSTEMS J OURNAL and Control Engineering Editor of the IEEE T RANSACTIONS ON P OWER E LECTRONICS and the IEEE
Practice. He is a member of the Institution of Engineering and Technology T RANSACTIONS ON I NDUSTRIAL E LECTRONICS, and an Editor of the IEEE
(IET) and International Council on Large Electric Systems. Journal of Emerging and Selected Topics in Power Electronics.