IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO.

2, MAY 2014 967

Coordinated Control of Grid-Connected

Photovoltaic Reactive Power and Battery
Energy Storage Systems to Improve the Voltage
Profile of a Residential Distribution Feeder
M. N. Kabir, Y. Mishra, Member, IEEE, G. Ledwich, Senior Member, IEEE,
Z. Y. Dong, Senior Member, IEEE, and K. P. Wong, Fellow, IEEE

Abstract—Increasing penetration of photovoltaic (PV), as well as ZnBr Zinc bromide.

increasing peak load demand, has resulted in poor voltage profile
for some residential distribution networks. This paper proposes State solar PV generation process at th
coordinated use of PV and battery energy storage (BES) to address hour of th day in a year.
voltage rise and/or dip problems. The reactive capability of PV P Probability of being in the state at th hour
inverter combined with droop-based BES system is evaluated for of th day in a year.
rural and urban scenarios (having different = ratios). Results
show that reactive compensation from PV inverters alone is suffi- The probability of having the value “ ”
cient to maintain acceptable voltage profile in an urban scenario on next time of th day,
(low-resistance feeder), whereas coordinated PV and BES support is provided that the solar PV generation pro-
required for the rural scenario (high-resistance feeder). Constant, cess gets a particular value “ ” at th hour of
as well as variable, droop-based BES schemes are analyzed. The th day.
required BES sizing and associated cost to maintain the acceptable
voltage profile under both schemes are presented. Uncertainties in Solar irradiation constant.
PV generation and load are considered, with probabilistic estima- , , and Weighted factors for DNI, GHI, and DIF,
tion of PV generation and randomness in load modeled to charac- respectively.
terize the effective utilization of BES. Actual PV generation data and Probability density of the initial state “ ” of
distribution system network data are used to verify the efficacy of
the proposed method. th for the solar PV generation process.
Distance between backbone bus and
Index Terms—Battery energy storage (BES), distribution
respective house in the network model.
networks, droop control, photovoltaic (PV).
Maximum load power.
Maximum PV generated power.
NOMENCLATURE Sensitivity matrix.
Variation in the bus angle with active
DNI Direct normal irradiation. power.
GHI Global horizontal irradiation. Variation in the bus angle with reactive
DIF Diffusion irradiation. power.
SoC State of charge. Variation in the bus voltage with active
DoD Depth of discharge. power.
NiMH Nickel metal hydride cell. Variation in the bus voltage with
reactive power.
Vn Vanadium.
Maximum difference in real power in the
sensitivity matrix.
Manuscript received January 31, 2013; revised May 1, 2013 and August 28,
2013; accepted December 30, 2013. Date of publication January 09, 2014; date of Required power reduction for droop-based
current version May 02, 2014. Paper no. TII-13-0060. storage.
M. N. Kabir, Y. Mishra, and G. Ledwich are with the School of Electrical Constant droop coefficient.
Engineering and Computer Science, Queensland University of Technology,
Brisbane, Queensland 4000, Australia (e-mail: [email protected]; Variable droop co-efficient for th house.
[email protected]; [email protected]). Critical voltage above which droop-based
Z. Y. Dong is with the School of Electrical and Information Engineering, energy storage starts.
University of Sydney, Sydney, New South Wales 2006, Australia (e-mail:
[email protected]). Maximum allowable voltage (1.06 p.u.).
K. P. Wong is with the University of Western Australia, Crawley, Western Expected voltage level (1.0 p.u.).
Australia 6009, Australia (e-mail: [email protected]). Voltage at th house without any energy
Color versions of one or more of the figures in this paper are available online at
http://ieeexplore.ieee.org. supply from BES at a particular time .
Digital Object Identifier 10.1109/TII.2014.2299336 Total time duration for voltage dip in a day.

1551-3203 © 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.
968 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO. 2, MAY 2014

Round trip efficiency considering SoC and

DoD.
Total energy stored from each house using
droop-based storage.
Available energy to be delivered to each
house at a particular time.
Supplied energy from BES for th house.

I. INTRODUCTION
HE ENERGY sector has jumped into an era where
T increased energy demand is partly met through wide-
spread installation of photovoltaics (PVs) due to favourable
economical, technical, and environmental factors. Fig. 1 shows
the predicted growth of PV generation in Australia [1], which is Fig. 1. Predicted PV growth in Australia under rapid and moderate uptake
expected to grow from 320 to 1130 MW/year under different scenarios [1].
uptake scenarios. This trend would result in 16%–20% of total
electricity produced from rooftop PVs by 2031.
Government incentives combined with public awareness
make rooftop residential PV a popular choice for domestic power
generation. Residential rooftop PVs contribute around 43% of
total rooftop PV generation in Australia as shown in Fig. 2. PVs
can increase service reliability and hence defer the need for
immediate investment in grid reinforcement. This would extend
the possibility of utilizing PVs for local voltage regulation in a
way that is not possible with conventional centralized generators
[2]. The feasibility of small scale solar generation in urban and Fig. 2. Composition of rooftop PV generation in Australia, in February 2012 [1].
rural areas has been demonstrated in [3], [4].
Nevertheless, high-level penetration of PVs in a distribution
network brings several technical challenges. Active generation in Integration of energy storage systems to store excessive
the form of PVs has resulted in over-voltage issues toward the end energy instead of curtailing can be another solution to alleviate
of the distribution feeder [5], particularly, for overhead lines. the over-voltage problem [17]–[19]. Battery energy storage
Voltage control by adjusting the transformer tap is proposed in (BES) systems are compact and can play vital role in maintaining
[6]. Particular focus has been given to the realtime setting of the fast residential voltage profile in terms of “Time Shift” which is
on-load tap changer (OLTC) of distribution substations in [7].After encouraged by storage charging at off-peak time and discharging
instantaneous demand forecasting and measurements of power at the peak time. Addressing voltage fluctuations caused by PVs
injectedbydistributedgenerations(DGs),itispossibletodetermine using customer side energy storage systems has been introduced
OLTCsettingsthatwillenablehigher generationfrom DGswithout recently in [20]–[23]. Low-energy density and deep discharging
voltage violations [8], [9]. But, it requires a complex algorithm and of nonbattery-type energy storage systems, such as electro-
reliable communication for data exchange between DGs and loads. chemical capacitors, are ideal for high power and short duration
Moreover, frequent tap changing increases transformer stress and applications. Battery-type storage are suitable for high energy,
feeder losses. long duration applications such as residential distribution sys-
Keeping the grid voltage in the acceptable range can also be tems [23]. The inverter control of PV and BES system is
achieved by PV capacity limitation [10], [11], which may not be reconfigured to work as current source in phase with grid voltage
desirable since a significant amount of energy will be wasted. under grid-connected mode and as voltage source using droop
Reactive power control can also be effective to improve the scheme under islanded mode [20]. Hugihara et al. [21] have
voltage profile. Several research studies are performed on dif- proposed coordinated controller for medium and large scale
ferent centralized and decentralized reactive control strategies in commercial customers and uses lumped grid scale storage with
[12]–[14]. Stetz et al. [15] propose autonomous inverters for complex control modes. Considering high line losses in LV
voltage improvement using their reactive capability, by limiting systems due to complex control modes, lumped grid scale storage
the active power feed-in, which would result in the loss of is not a good option for residential networks.
customer revenue. Besides, reactive control strategies become Storage is also helpful in smoothing the variable output of
less effective in low-voltage (LV) distribution network com- PVs, which are affected by cloud transients (CTs) and partial
pared to medium-voltage and high-voltage networks [16]. shading. PV module topology with battery equalizer can be used
Hence, in suburban and/or rural areas, where the distribution to raise the active power output from partially shaded PV
network is more resistive, a more suitable approach to maintain modules [24]. Considering CT and seasonal variation to quantify
voltage profile is required. the solar resource potential for planning purposes are reported in
KABIR et al.: COORDINATED CONTROL OF GRID-CONNECTED PV REACTIVE POWER AND BES SYSTEMS 969

Fig. 3. Typical radial residential distribution feeder.

[25]–[27]. However, the model to account for variability in PV

generation and randomness in loads is not yet explored in detail.
Contributions: This paper proposes a coordinated control of
PV and BES system for voltage control of residential distribution
systems. Unlike [21], [22], a local droop-based control of BES
placed at each house is proposed and does not require advance
metering infrastructures. In this paper, integration of PV and
BES in each house is similar to [20], except that BES charging
and discharging is based on the house voltage. Coordinated
inverter control is different compared to [24], as the primary aim
is not to increase the active power injection from PV modules
and make over-voltage problem worst.
Another major contribution of this paper is the quantification of
the impact of line characteristics on the effectiveness of reactive
control, which has not been reported in the past. Although
Carvalho et al. [16] have shown that the reactive control is
effective for wide range of load and generation condition, its
quantification with respect to ratio of the line is missing. Fig. 4. Over-voltage problem with PV (at 12 midday).
This quantification will determine the optimum BES sizing by
utilizing the reactive capability of PV in combination with BES. However, power injection at unity power factor causes over-
Furthermore, modeling the effect of CT and partial shading voltage issues in a radial distribution feeder (Fig. 3) as shown in
using variability of PV as well as load is considered in this paper. Fig. 4. Although different regions in the world will have different
Previously, Hollands and Huget [25] and Tiba et al. [26] have feeder voltage level as well as power ratings of the distribution
only included probabilistic modeling of clearness index and does transformer, voltage violations are observed in most of the grids
not model it as “memory less” stochastic process. The net across the world irrespective of the voltage levels. Fig. 3 is a setup
generation profile (difference between PV generation and load) found in typical Australian distribution systems which is more
can be considered as discrete time Markov chain (DTMC) sensitive due to the prevalence of overhead lines.
process because the “future state” depends only on the “present The voltage rise is prominent toward the end of the feeder.
state” and independent of the “past states.” This paper discusses a Rural voltage profile is worse than that of urban due to the higher
detailed probabilistic modeling to find the effective utilization of resistive characteristic of rural feeders.
BES under variable PV generation. The reactive power capability of PV inverter can be utilized to
Synopsis: In Section III, the coordinated control algorithm of reduce the over-voltage problem with very little loss of real
PV and BES for voltage improvement is introduced. A sample power. For a given apparent power ( ) and instantaneous real
LV system and extensive simulation results are presented in power generated ( ) from an inverter, the range of allowable
Section IV. The economics of the PV and BES systems is
discussed in Section V highlighting the choices for battery reactive power generation is given by < .
selection and cost comparison. BES capacity utilization using The operation of PV inverters in nonunity power factor mode
variable generation and load is discussed in Section VI followed will reduce real power injection and allow reactive power
by validation of the efficacy of proposed scheme with real absorption. Real power grid injection from PV inverter can
network data in Section VII. be further limited (if required) by using a BES system, thereby
reducing over-voltage problem.
Control of power storage in batteries can be achieved using
either constant droop-based or variable droop-based methods.
II. COORDINATED CONTROL OF PV AND BES FOR
The constant droop method utilizes same droop co-efficient ( )
VOLTAGE IMPROVEMENT
for all houses to store energy. is a function of and
Existing regulations require grid-connected PV inverters to and is defined as in (1). Using , the amount of reduced
operate at unity power factor for maximum real power injection. power from th PV is calculated as in (2). With
970 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO. 2, MAY 2014

constant droop method, BES starts charging when the voltage of

the house goes above , as defined in (3)

whereas the variable droop-based BES uses different droop

co-efficient ( ) for different houses to ensure uniform
energy storage. The calculation of ( ) is based on
, which in turn is defined using Jacobian of the load flow
as in

Each element of voltage sensitivity submatrix ( ) Fig. 5. BES energy supply during peak demand to eliminate voltage-dip.
represents the expected variation in voltage of any house ( ) with
respect to a unit injection of active power ( ) in another house
( ). Positive signs in all the elements of indicate that Step 4) If the voltage still exceeds the upper acceptable range
any injection of at any house ( ) will increase other house (1.06 p.u.), reactive power is absorbed using the PV
voltage ( ) and vice versa. The droop coefficients ( ) are inverters. PV inverters absorb reactive power during
designed to keep the house voltages between 1 p.u. and peak solar radiation time, the amount of which can
(1.06 p.u. as per Australian standards), where the th PV is be proportional to excess voltage, i.e., the amount of
defined using (5) and (6). The amount of power stored in the th voltage exceeding the upper acceptable range (1.06 p.u.)
BES systems is calculated in (7). With variable of the respective house at that time.
droop-based methods, BES starts charging when voltage of the Step 5) Finally, the load flow is performed again. If all
house goes above a critical voltage ( ) as defined in (8) house voltages are within the limit and then wait for
the next hour and start again from Step 1). Otherwise
repeat Step 4).
For radial distribution systems, some specially designed
methods, such as the well-known backward-forward sweep
technique, are more efficient to compute the power flow than
the conventional Newton–Raphson method [28], [29] and hence
can be used here. Similar coordination of PV inverters and BES
can be utilized for improving the LV problem during peak
demand as shown in Fig. 5. When there is no sunlight
( ), the full capacity of PV ( ) inverter can be used for
reactive compensation, with a small correction in the dc bus
capacitor rating. PV inverters inject reactive power when voltage
goes below the lower acceptable value (0.94 p.u. as per Austra-
The BES sizing will be determined by using either the constant lian standards). Real power is supplied from the BES during the
or the variable droop-based methods. The coordinated control of peak demand period.
PV inverters and BES is required to alleviate the over-voltage The energy is supplied when the voltage of th house at
problem. The overall algorithm for determining the appropriate time step “ ” is below the lower acceptable limit (0.94 p.u.).
BES sizing can be summarized below: Total time duration (tvd) is defined as the time duration when
Step 1) House voltages are observed every hour without any voltage dips are observed and is calculated based on past data
reactive compensation of PV inverters or droop-based (typically 4–5 h/day). is calculated from
BES systems. by multiplying it with considering SoC and
Step 2) At a particular time step, if all the house voltages are DoD of the battery as in
within limit as well as below the critical voltage ( ),
no action is taken and all voltages are measured during
next hour. Supplied energy from the BES for the th house is then
Step 3) If the voltage of a particular house goes above , PVs determined as in
supply power to BES systems instead of supplying
power to grid. House voltages are updated by running
the load flow.
KABIR et al.: COORDINATED CONTROL OF GRID-CONNECTED PV REACTIVE POWER AND BES SYSTEMS 971

For constant droop-based BES, stored energy,

, and hence will be different for
all the houses. Whereas, in the case of variable droop-based BES,
, , and are the
same for all houses, assuming identical BES systems are used.
After supplying energy from the respective BES, house
voltages ( ) are again checked by running the load flow. If
is still below 0.94 p.u., reactive injection from PV inverter is
performed for the th house. The overall procedure continues
until all the house voltages are above 0.94 p.u.

III. TEST SYSTEM DESCRIPTION AND SIMULATION RESULTS

An overhead residential feeder supplying electricity to
12 houses by a 75-kVA 11-kV/240-V transformer is considered Fig. 6. PV generation profile and load profile of house H11.
as a test network (Fig. 3). All the houses are equipped with
rooftop PVs rated at 6 kVA each. In the urban scenario, the
distance between backbone bus and each house ( ) is 20 m and
ratio is considered unity. Whereas in the rural scenario, is
100 m, ratio is 2.8 for the backbone and 5.6 for the single
wire earth return line (from the backbone bus to the respective
house). Distribution networks in rural areas have weak char-
acteristics (in terms of line resistance and inductance) and are
reflected in the test system.
A typical summer load profile (shown for 24 h in Fig. 6) is
taken from an Australian Distribution Company [30]. A maxi-
mum load of 3 kW is considered for each house. Variation in the
residential load profile distribution of different houses is modeled
using white Gaussian Noise. The random load of H11 is shown in
Fig. 6. Realtime data (15 and 60 min interval) for PV generation
is used from the University of Queensland solar system in this
study [31]. In this paper, hourly data profile is considered for the
coordinated control of PV and BES.
A summer day is chosen to simulate the worst case scenario Fig. 7. Hourly voltage profile of last house (H11/12) for urban and rural scenario
(extreme PV generation will create worst over-voltage and in the normal situation.
summer peak loading will have worst voltage dip). Fig. 6 shows
PV generation profile for a typical Australian summer day So, the maximum reactive capability of a PV inverter during day
(December 6, 2012) with moderate CT. and night time is considered to be 3.6 and 6 kVAR, respectively.

A. Voltage Profile During High PV Generation and Peak B. Impact of Line Characteristics on the Effectiveness of
Loading Time Reactive Compensation for Voltage Improvement
In normal situation (without considering any kind of reactive Although reactive compensation is an attractive option for
power control or battery storage), the hourly voltage profile of the voltage improvement, the ratio of the distribution feeder
residential distribution feeder is plotted and is shown for urban limits its effectiveness. Simulations are performed to find the
and rural scenario is Fig. 7 for H11/H12. critical ratio ( ) of the distribution line for which
The voltages of last two houses in the feeder H11/H12 exceed reactive compensation can be effective for voltage improvement.
the allowable limit of 1.06 p.u. at 12 (midday). The H11 and H12 Incremental line losses as well as line loading are calculated
loads are 1.875 kW compared to the PV generation of 6 kW at for different values of ratio using repeated load flow
12 (midday). Similarly, during the peak loading time, i.e., at studies. For every value of ratio in the feeder, the
8 P.M. when there is no PV generation, the voltage is well below injection from PV inverters are varied from 0 to 3.6 kVAR until
the lower limit of 0.94 p.u. The voltage profile of H11/H12 is the network voltage violations are eliminated. If voltage viola-
worst in the rural case. tion persists even after maximum reactive compensation from
The reactive capability of PV inverters (if operated at nonunity PV inverters, then the real power loading is reduced until the
power factor) can be used to improve the voltage profile. For a acceptable voltage is achieved. Fig. 8 shows the percentage of
6-kVA PV inverter, 3.6 kVAR of reactive power ( ) is available feeder loading level with the increase in ratio. With higher
at 0.8 power factor. During the night, when , the ratio ( > ), loading should be reduced to 70% of its
full capacity of PV inverter can be used for reactive compensation. peak value to keep the voltages in the acceptable range. The
972 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO. 2, MAY 2014

Fig. 8. Permitted loading and incremental loss variation with ratio.

incremental line losses, i.e., increase in the line losses when

ratio increases from unity shows that losses are high for more
resistive lines for same reactive injection. For ratio greater
than 5, these losses are too high. Fig. 9. Voltage profile, PV inverter reactive support, stored energy, and supplied
energy when coordinated control of PV inverter and constant droop-based BES is
Therefore, the effectiveness of PV inverters for voltage control used for voltage improvement.
is limited for higher ratio cases. For the urban case
( ), the reactive capability of PV inverters is
sufficient for the voltage improvement, whereas for the rural
case ( > ), the reactive compensation alone is not sufficient
and storage is required to achieve improved voltage profile.

C. Coordinated Control of PV Inverters and BES for

Improvement in Voltage Profile in the Rural Case
Coordinated control of PV inverters and constant/variable
droop-based BES to improve the voltage profile for the rural
scenario is investigated. All the houses connected to the rural
distribution feeder are proposed to have BES for improved
voltage profile, whereas for the urban scenario, there is no need
of BES since the reactive compensation from the PV inverter is
sufficient to maintain an acceptable voltage profile.
The hourly voltage profile, reactive support from PV inverters,
stored and supplied energy from the BES, is shown in Figs. 9 and
10 for constant and variable droop-based methods, respectively.
Two houses connected to a single bus possess the same char-
acteristics and therefore, for plotting purposes, only six houses
are shown. The BES is charged during the day time and
discharges during the evening when the house voltage goes
below 0.94 p.u. Supply of stored energy usually starts at late Fig. 10. Voltage profile, PV inverter reactive support, stored energy, and
afternoon (5–6 P.M.). So, a gradual decrease in the stored energy supplied energy when coordinated control of PV inverter and variable droop-
profile is observed. based BES is used for voltage improvement.
Constant droop-based BES requires different sizes of BES for
different houses and the last two houses H11/H12 require 5-kWh All the houses require total of 40.5 kWh of BES for the constant
BES. Houses at the end of the feeder require larger BES capacity droop-based method compared to only 24 kWh if the variable
and hence higher investment, which may not be seen as a fair droop-based method is chosen.
solution in the community, whereas the variable droop-based If BES alone were to be considered for voltage improvement,
method requires 2 kWh of battery for all houses as shown in third all the houses will need much higher capacity of batteries. For
and fourth subplot of Fig. 10. Moreover, with variable droop- H11/H12 alone, the constant droop-based method will require a
based BES, the total storage capacity required to achieve the 20-kWh BES system (Fig. 11) and the variable droop needs
same improvement in the voltage profile is significantly reduced. 17 kWh (Fig. 12). This is excessively large and economically
KABIR et al.: COORDINATED CONTROL OF GRID-CONNECTED PV REACTIVE POWER AND BES SYSTEMS 973

TABLE I
BATTERY CHARACTERISTICS COMPARISON

Fig. 11. Voltage profile, stored energy in BES, and supplied energy from BES
when constant droop-based BES is used alone for voltage improvement.
batteries (ZnBr, Vn redox). The following criteria are used for
selecting the appropriate BES for residential purposes.
The first criterion is recharging ability. Alkaline batteries are
good for only 25–30 cycles when operating with higher DoD.
Therefore, these batteries are not considered due to lack of
recharging ability.
The second concern is safety. Byproducts of the battery
charging process such as hydrogen gas, sulfuric acid can cause
explosion, corrosion, or burning. If overheated, some Li-Ion
batteries may suffer thermal runaway which can lead to combus-
tion in extreme cases. Deep discharge may short circuit the cell,
making them unfavorable for residential premises.
The third criterion is the space and weight characteristics. Too
heavy and spacious battery equipment may require additional
structural support and may not be an attractive option.
The fourth factor is the round trip efficiency during charging-
discharging process, which will affect battery size and cost.
Therefore, high round trip efficiency is desirable. Lower self
discharge rate is another measure of the efficiency. Life cycle of
the BES is highly correlated with frequency of charging-
discharging and DoD. Residential BES system needs at least
one charging–discharging cycle every day.
Fig. 12. Voltage profile, stored energy in BES, and supplied energy from BES
The fifth factor is the cost. This is given a least priority as the
when variable droop-based BES is used alone for voltage improvement. cost is constantly changing due to technological advancement.
Table I compares Lead-Acid, Li-ion, NiMH, and flow batteries
infeasible. Hence, the coordinated approach is preferred com- (Vn Redox, ZnBr) [32], [33].
pared to using BES alone for voltage improvement. Traditionally, lead-acid batteries were the prime choice for
off-grid PV applications. Despite their long history and wide-
spread usage, lead-acid batteries are one of the lowest energy-to-
IV. ECONOMICS OF COORDINATED PV INVERTERS AND
weight and energy-to-volume battery designs making it big and
BES SYSTEMS
heavy for residential applications. Nevertheless, they are far
The section compares the cost of constant and variable droop- cheaper than any other BES system available in the market.
based BES systems versus a utility feeder reinforcement such as Also, lead-acid batteries work optimally with 40%–50% DoD
DSTATCOM. An appropriate battery has been selected for the only. In recent years, flow batteries are becoming popular for
cost comparison. small scale storage (range of 5–10 kWh) due to its efficient
performance in frequent deep discharging (up to 100%). Unlike
A. Battery Selection
other battery technologies, flow batteries are not fatigued by
There are wide range of existing batteries systems available for frequent deep discharging as the electrolytes are kept in separate
storage including lead-Acid, Li-Ion, Alkaline, NiMH, and flow chamber. They are also resistant from self-discharging [32].
974 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO. 2, MAY 2014

TABLE II PVs. PV generation can be defined as an independent and

COST COMPARISON BETWEEN CONSTANT AND VARIABLE DROOP-BASED BES SYSTEMS identically distributed (IID) random variable if an infinite num-
FOR VOLTAGE IMPROVEMENT
ber of samples can be taken which is not possible in practice. The
process can be considered as DTMC instead. Weather forecast-
ing is considered as DTMC which is a “memory less” stochastic
process. In such processes, it is assumed that the “future state”
depends only on the “present state” and independent of the “past
states.” Let us consider as the solar PV generation state at
th hour of th day in a year. So, the process is { where
; }. The probability of trans-
ferring into a state at time “ ” entirely depends on the state at
present time “ .” It is independent of the past states at ,
. When the process is in state “ ,” there is a fixed
Among all flow batteries available, ZnBr is selected to be the probability that it will be in state “ ” in next sample such as
most suitable for this study, as it is cheaper than Vn redox.

B. Cost Comparison
The cost of constant and variable droop-based ZnBr battery
Once the solar PV generation process gets a particular value
system is shown in Table II. Constant droop-based BES systems
“ ” at th time of th day, the probability of having the value “ ”
costs $2100 for the last two houses which is 10 times higher than
on next time ( )th of th day is . The summation of
the first two houses, whereas variable droop-based BES system
probabilities of all possible values that solar PV generation
costs $800 for each house. Overall, the cost of variable droop-
process can have is 1
based BES storage system is 40% of constant droop-based
storage for the given network.
The cost of storage for voltage improvement is compared with
the utility-based reactive power control devices such as
D-STATCOM. The capacity of a D-STATCOM is determined
iteratively using load flow studies by placing a D-STATCOM at The value of “ ” depends on solar irradiation and CTs as
the fifth bus in the network shown in Fig. 1 (two-third the shown in (13). There are three factors associated with the solar
distance, i.e., nearly 75% of feeder [34]). For the given network, irradiation, namely DNI, GHI, and DIF
65 kVAR of D-STATCOM is required, which will cost around
$3500 (assuming $50–55/kVAR [35]). Although one time in-
vestment in a D-STATCOM is less when compared to the total
investment in BES ($9600), BES provides long-term technical where is the solar irradiation constant and , , and are
benefits in terms of peak shaving. D-STATCOM can only the weight factors for DNI, GHI, and DIF, respectively. DNI has
regulate voltage with reactive compensation but cannot supply the most impact on concentrated solar power (CSP) projects and
real power during peak demand. On the other hand, new tech- PV tracking systems, which are designed to keep their surfaces
nological innovation in the batteries may bring down the cost perpendicular to the incident solar rays. In residential community,
significantly, making the proposed scheme economically usually fixed panel PV’s are used whose generation mainly
attractive. depends on GHI. If the residences are widely placed, which is
more often in suburban and rural case, DIF has almost zero impact.
V. BES UTILIZATION WITH VARIATION OF PV GENERATION For residential distribution system, “ ” can be simplified as
AND LOAD PROFILE

Deterministic worst case scenario helps to determine the

maximum BES capacity required with coordinated control of
PV and BES. However, this capacity of BES may not be fully Satellite-based time series records of reflected sunlight are
utilized on a daily basis. Modeling multiple scenarios of variable used to determine a cloud index time series. Moreover, for GHI
PV generation and load is necessary to gauge the utilization of the average annual irradiance mean differ from the long-range
BES system. In this section, the probabilistic estimation of PV distributed mean by only 5%. So, it is justified to consider same
generation and randomness in load profiles are considered over a solar PV generation process for all the residences of a particular
year (2012) to analyze the percentage of utilization of BES network. Moreover, while considering the GHI, variation is
capacity. For load profile, two distinct seasonal profiles (summer distinctly observed for summer and winter only.
and winter) are considered. The probability of state “ ” is defined for the transition to
Using the present values of PV generation, future PV genera- next sample (i.e., transition from th time to th time). Now,
tion values are estimated. The intermittent characteristics of solar it is required to define th step transition probabilities, where
energy demand probabilistic analysis of the generation from , of any “ ”th day, where ,
KABIR et al.: COORDINATED CONTROL OF GRID-CONNECTED PV REACTIVE POWER AND BES SYSTEMS 975

Fig. 13. Utilization of BES over the year 2012.

based on present state ( th time of “ ”th day). If the unconditional Fig. 14. Urban network from a Northern Suburb in Brisbane, Australia.
probability at any state “ ” is to determined, the probability
density of the initial state “ ” for a particular day “ ” is required

All unconditional probabilities need to be computed by con-

ditioning on the “ th” time to determine the probability on th
step transition

Fig. 15. Rural network from North Queensland.

Such estimated solar PV generation values of each hour over a

year (2012) are used to determine the percentage utilization of
BES capacity. BES capacity is the size determined from the worst
case analysis in Sections II and III. During the simulation,
random load profiles with two distinct seasonal variations
(winter (April–September) and summer (October–March) are
considered for every house in the network. Using estimated PV
generation values and random load profiles, load flow is run for
each hour of the year (total 24 * 366 times) to determine daily
BES utilization percentages for the year 2012 for all the houses.
The coordinated control algorithm with PV inverter and variable
droop-based BES is considered for all the simulations.
The results of BES utilization are summarized in Fig. 13. BES
system is underutilized ( < of BES capacity) for only 16.1%
of time (59 days/year). The BES sizing determined from the
proposed algorithm in Section III is well utilized for 50.7% of Fig. 16. Voltage profiles and reactive compensation profile.
time (185 days/year).
peak load to simulate severe over-voltage conditions. For the
rural case, network data is taken from rural areas in North
VI. VALIDATION OF PROPOSED METHOD WITH
Queensland (Fig. 15). In this scenario, houses are located much
REAL NETWORK DATA
farther from each other (80–200 m) and ratio is much
The proposed method is applied in a distribution system in higher ( ).
Australia. For urban case, a North Brisbane city location is The house located toward the end of the distribution trans-
chosen (Fig. 14). Houses are close to each other (distance varies former has the worst voltage profile. The voltage profiles of all
from 10–35 m) and ratio is close to unity (varies from 0.9 houses are monitored, but for plotting purposes, only the house
to 1.2). The network data including distance and ratio is toward the end of the feeder, i.e., P5883-I (urban case) and P17
taken from [30]. The cumulative load data is also taken from one (rural case) is shown in Fig. 16. Voltage profile is much worse for
of the Australian distribution company [30]. The peak load for rural case.
each house is the average value, calculated by dividing the The voltage profile after reactive compensation from PV
cumulative peak load with the number of customers/houses. inverter is shown in Fig. 16. Voltage improves after the PV
The average value of the house peak load is around 3 kW. reactive support in urban case; however voltage violation still
Installed PV generation in each house is considered twice the exists in rural case. Reactive compensation from PV inverter
976 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 10, NO. 2, MAY 2014

TABLE III
URBAN LV NETWORK PARAMETERS USED IN THIS STUDY [36]

presented. Although one time investment in D-STATCOM is

lower when compared with total investment in BES, BES
provides long-term technical benefits in terms of peak shaving.
Considering the intermittent characteristics of PV generation,
probabilistic estimation is also performed. The probabilistic esti-
mation of PV generation and randomness in load profiles are
considered over a year (2012) to analyze the utilization of BES.
Two distinct seasonal profiles (summer and winter) are considered.
It is found that BES capacity is well utilized for the voltage control
purposes for more than 50% of time in the year (185 days/year).
The proposed method is applied in the actual urban and rural
Fig. 17. Voltage profile, reactive compensation profile, and supplied energy distribution network in Australia and similar improvement is
profile for rural case in coordinated approach.
observed.
alone for rural and urban cases is shown in second plot of Fig. 16. APPENDIX
The maximum capability of PV inverter is used (3.6 kVAR
absorption during the presence of sunlight and 6 kVAR supply Urban LV network parameters (percentage on 1 MVA base)
during the evening) for the rural case, but acceptable voltage are shown in Table III [36].
profile is yet to be achieved. Coordinated control of PV inverters Rural network parameters are chosen to maintain higher
and variable droop-based BES is applied to the rural case. Fig. 17 ratio (5.6) using urban network data.
shows improvement in the voltage profile of the last house in the
feeder. REFERENCES
[1] Australian Energy Market Operator. (2012). Rooftop PV information paper
VII. CONCLUSION [Online]. Available: http://www.aemo.com.au/Electricity/~/media/Files/
Other/forecasting/Rooftop_PV_Information_Paper.ashx
This paper introduces the use of coordinated control of PV [2] W. Xiao, W. G. Dunford, P. R. Palmer, and A. Capel, “Regulation of
inverters and droop-based BES to keep voltage in the acceptable photovoltaic voltage,” IEEE Trans. Ind. Electron., vol. 54, no. 3,
range with high penetration of rooftop PVs in residential distri- pp. 1365–1374, Jun. 2007.
[3] F. Blaabjerg, Z. Chen, and S. B. Kjaer, “Power electronics as efficient
bution systems. There may be over-voltage issues in the feeder interface in dispersed power generation systems,” IEEE Trans. Power
due to the unity power factor real power injection from PV Electron., vol. 19, no. 5, pp. 1184–1194, Sep. 2004.
inverters in some residential feeders. Utilizing the reactive power [4] M. Thomson and D. G. Infield, “Impact of widespread photovoltaics genera-
tion on distribution systems,” IET Renew. Power Gen., vol. 1, pp. 33–40, 2007.
capability of PV inverter is proposed to overcome this issue. [5] Y. Ueda, K. Kurokawa, T. Tanabe, K. Kitamura, and H. Sugihara, “Analysis
However, it is also found that when the line resistivity exceeds a results of output power loss due to the grid voltage rise in grid-connected
certain critical value (this critical ratio is 4.5–5 for the given photovoltaic power generation systems,” IEEE Trans. Ind. Electron.,
vol. 55, no. 7, pp. 2744–2751, Jul. 2008.
network), reactive compensation alone becomes less effective. [6] L. Xiaoming, L. Qingfen, Y. Xianggen, and X. Jianghui, “A new on-load tap
Therefore, the reactive capability of PV inverters alone is changing system with power electronic elements for power transformers,”
sufficient to improve the voltage profile in urban case, where in Proc. PowerCon Int. Conf. Power Syst. Technol., Kunming, China, vol. 1,
2002, pp. 556–559.
ratio is close is unity, whereas droop-based BES combined [7] R. Echavarria, A. Claudio, and M. Cotorogea, “Analysis, design, and
with PV inverters are required in the rural case which has higher implementation of a fast on-load tap changing regulator,” IEEE Trans.
ratio. If BES system were to be used alone for the voltage Power Electron., vol. 22, no. 2, pp. 527–533, Mar. 2007.
[8] X. Liu, A. Aichhorn, L. Liu, and H. Li, “Coordinated control of distributed
improvement, larger BES capacity is required. energy storage system with tap changer transformers for voltage rise
Constant as well as variable droop-based BES schemes are mitigation under high photovoltaic penetration,” IEEE Trans. Smart Grid.,
investigated. Although both schemes show good performance, vol. 3, no. 2, pp. 897–906, Jun. 2012.
[9] T. Sansawatt, L. F. Ochoa, and G. P. Harrison, “Smart decentralized control
constant droop-based BES requires larger battery size and of dg for voltage and thermal constraint management,” IEEE Trans. Power
unequal investment in BES for the customers depending on their Syst., vol. 27, no. 3, pp. 1637–1645, Aug. 2012.
location in the feeder. Variable droop-based BES, however, [10] Y. Ueda, K. Kurokawa, T. Tanabe, K. Kitamura, and H. Sugihara, “Analysis
results of output power loss due to the grid voltage rise in grid-connected
requires smaller battery size and equal investment in BES from photovoltaic power generation systems,” IEEE Trans. Ind. Electron.,
all the customers in distribution feeder. The proposed coordinat- vol. 55, no. 7, pp. 2744–2751, Jul. 2008.
ed control algorithm alleviates both the over-voltage and the [11] R. Tonkoski, L. A. C. Lopes, and T. H. M. EL-Fouly, “Droop-based active
power curtailment for over-voltage prevention in grid connected PV
voltage dip problem in the residential feeders. inverters,” in Proc. IEEE Int. Symp. Ind. Electron., Bari, Italy, 2010,
Advantage of variable droop-based energy storage is dis- pp. 2388–2393.
cussed in terms of financial investment from customers and brief [12] Y. K. Lo, H. J. Chiu, T. P. Lee, I. Purnama, and J. M. Wang, “Analysis
and design of a photovoltaic system DC connected to the utility with a
comparison of financial and technical advantages between BES power factor corrector,” IEEE Trans. Ind. Electron., vol. 56, no. 11,
and D-STATCOM (reactive power compensation device) is pp. 4354–4362, Nov. 2009.
KABIR et al.: COORDINATED CONTROL OF GRID-CONNECTED PV REACTIVE POWER AND BES SYSTEMS 977

[13] R. Cardenas, R. Pena, P. Wheeler, J. Clare, and G. Asher, “Control of the [30] (2009, Jul.). Energex Regulatory Proposal for the Period July 2010–June
reactive power supplied by a WECS based on an induction generator fed by 2015 [Online]. Available: http://www.energex.com.au/__data/assets/pdf_
a matrix converter,” IEEE Trans. Ind. Electron., vol. 56, no. 2, pp. 429–438, file/0010/31789/ENERGEX_s_Regulatory_Proposal_2010-2015.pdf
Feb. 2009. [31] University of Queensland (UQ) Solar Photo Voltaic data [Online]. Avail-
[14] A. Cagnano, E. Tuglie, M. Liserre, and R. A. Mastromauro, “Online optimal able: http://uqld.smartersoft.com.au/user/reportEnergy.php
reactive power control strategy of PV inverters,” IEEE Trans. Ind. Electron., [32] All About Batteries [Online]. Available: http://www.allaboutbatteries.com/
vol. 58, no. 10, pp. 4549–4558, Oct. 2011. Battery-Energy.html
[15] T. Stetz, F. Marten, and M. Braun, “Improved low voltage grid-integration [33] A. Akhil et al., “DOE/EPRI 2013 Electricity Storage Handbook in Collab-
of photovoltaic systems in Germany,” IEEE Trans. Sustain. Energy, vol. 4, oration with NRECA,” Sandia National Laboratories, Albuquerque, NM;
no. 2, pp. 534–542, Apr. 2013. Livermore, CA, Tech. Rep. SAND2013-5131, Jul. 2013.
[16] P. M. S. Carvalho, P. F. Correia, and L. A. F. Ferreira, “Distributed reactive [34] F. Shahnia, R. Majumder, A. Ghosh, G. Ledwich, and F. Zare, “Voltage
power generation control for voltage rise mitigation in distribution net- imbalance analysis in residential low voltage distribution networks
works,” IEEE Trans. Power Syst., vol. 23, no. 2, pp. 766–772, May 2008. with rooftop PVs,” Electr. Power Syst. Res., vol. 81, no. 9,
[17] P. C. Loh, L. Zhang, and F. Gao, “Compact integrated energy systems for pp. 1805–1814, 2011.
distributed generation,” IEEE Trans. Ind. Electron., vol. 60, no. 4, [35] M. Molinas, J. A. Suul, and T. Undeland, “Low voltage ride through of wind
pp. 1492–1502, Apr. 2013. farms with cage generators: STATCOM Versus SVC,” IEEE Trans. Power
[18] H. Kim, B. Parkhideh, T. D. Bongers, and H. Gao, “Reconfigurable solar Electron., vol. 23, no. 3, pp. 1104–1117, May 2008.
converter: A single-stage power conversion PV-Battery system,” IEEE [36] DTI Distributed Generation Programme (Contractor: EA Technology),
Trans. Power Electron., vol. 28, no. 8, pp. 3788–3797, Aug. 2013. Network Losses and Distributed Generation, Contact Number: DG/CGI/
[19] H. Beltran, E. Bilbao, E. Belenguer, I. Etxeberria-Oradui, and P. Rodriguez, 00038/00/00, URN number: 06/1238; Jan. 2006.
“Evaluation of storage energy requirements for constant production in PV
power plants,” IEEE Trans. Ind. Electron., vol. 60, no. 3, pp. 1225–1234,
Apr. 2013.
[20] D. Velasco de la Fuente, C. L. T. Rodriguez, G. Garcera, E. Figueres, and M. N. Kabir is currently working toward the Ph.D. degree at the Queensland
R. O. Gonzalez, “Photovoltaic power system with battery backup with University of Technology, Brisbane, Australia.
grid-connection and islanded operation capabilities,” IEEE Trans. Ind. His research interests include grid integration of photovoltaic in distribution
Electron., vol. 60, no. 4, pp. 1571–1581, Apr. 2013. systems.
[21] H. Hugihara, K. Yokoyama, O. Saeki, K. Tsuji, and T. Funaki, “Economic
and efficient voltage management using customer-owned energy storage Y. Mishra (S’06–M’09) received the Ph.D. degree from the University of
systems in a distributed network with high penetration of photovoltaic Queensland, Brisbane, Australia, in 2009.
systems,” IEEE Trans. Power Syst., vol. 28, no. 1, pp. 102–111, He is currently working as a Lecturer at the Queensland University of
Feb. 2013. Technology, Brisbane. His research interests include distributed generation and
[22] M. Sechilariu, W. Baochao, and F. Locment, “Building integrated photo- distributed energy storage, and power system stability and control and their
voltaic system with energy storage and smart grid communication,” IEEE applications in smart grid.
Trans. Ind. Electron., vol. 60, no. 4, pp. 1607–1618, Apr. 2013.
[23] D. T. Ton, C. J. Hanley, G. H. Peek, and J. D. Boyes, “Solar Energy Grid G. Ledwich (M’73–SM’92) received the Ph.D. degree in electrical engineering
Integration Systems-Energy Storage (SEGIS-ES),” Sandia National Labo- from the University of Newcastle, Newcastle, Australia, in 1976.
ratories, Albuquerque, NM; Livermore, CA, Tech. Rep. SAND2008-4247, He is a Chair Professor of Electrical Asset Management, Queensland Univer-
2008. sity of Technology, Brisbane, Australia. His research interests include power
[24] L. F. L. Villa, T. Ho, J. C. Crebier, and B. Raison, “A power electronics systems, power electronics, and controls.
equalizer application for partially shaded photovoltaic modules,” IEEE
Trans. Ind. Electron., vol. 60, no. 3, pp. 1179–1190, Mar. 2013.
[25] K. G. T. Hollands and R. G. Huget, “A probability density function for the Z. Y. Dong (M’99–SM’06) received the Ph.D. degree from the University of
clearness index with applications,” Sol. Energy, vol. 30, no. 3, Sydney, Sydney, Australia, in 1999.
pp. 195–209, 1983. He had been an Ausgrid Chair Professor and Director of the Centre for
[26] C. Tiba, A. N. Siqueira, and N. Fraidenraich, “Cumulative distribution Intelligent Electricity Networks, University of Newcastle, Newcastle, Australia.
curves of daily clearness index in a southern tropical climate,” Renew. Currently, he is with the University of Sydney. His research interests include
Energy, vol. 32, pp. 2161–2172, 2007. smart grid, power system security and planning, renewable energy systems, and
[27] S. Sarkar and V. Ajjarapu, “MW resource assessment model for a hybrid computational intelligence and its application in power engineering.
energy conversion system with wind and solar resources,” IEEE Trans.
Sustain. Energy, vol. 2, no. 4, pp. 383–391, Oct. 2011. K. P. Wong (M’87–SM’90–F’02) received the M.Sc., Ph.D., and higher
[28] S. Ghosh and D. Das, “Method for load-flow solution of radial distribution doctorate D.Eng. degrees from the University of Manchester, Institute of Science
networks,” Proc. IEE Gener. Transm. Distrib., vol. 146, no. 6, pp. 641–648, and Technology, Manchester, U.K., in 1972, 1974, and 2001, respectively.
Nov. 1999. He has been a Winthrop Professor with the School of Electrical, Electronic and
[29] D. Shirmohammadi, H. W. Hong, A. Semlyen, and G. X. Luo, “A Computer Engineering, the University of Western Australia, Perth,
compensation-based power flow method for weakly meshed distribution Australia, since 1974. He is a Con-Joint Professor at the University of Newcastle,
and transmission networks,” IEEE Trans. Power Syst., vol. 3, no. 2, Newcastle, Australia. His current research interests include power system
pp. 753–762, May 1988. analysis, planning and operations, and smart grids.