You are on page 1of 8

1

Regulation of Distribution Network Voltage


Using Dispersed Battery Storage Systems: A Case
Study of a Rural Network
Matthew Zillmann, Student Member, IEEE, Ruifeng Yan, Student Member, IEEE,
and Tapan Kumar Saha, Senior Member, IEEE

Abstract—The drive for power networks throughout the world between reactive power and network voltage, using four-
to utilize clean, green renewable generation poses a number of quadrant inverters to provide VAR compensation for the
issues for distribution and transmission companies. Two of the network. The second method applied to energy storage
most popular forms of renewable generation, wind and
systems, is the variation of real power injection [1, 3, 5]. In
photovoltaic (PV) face large fluctuations in their generation
profile. This variation means that peak generation and peak doing so, power is back fed onto the network, reducing losses
customer demand are often greatly divergent. Battery energy and boosting the voltage at the point of connection. Hybrid
storage can help to buffer wind and PV generation, capturing a multistage techniques, offer a compromise between real and
portion of the energy produced during light load and exporting it reactive power injection, adapting dynamically to network
back onto the network as required. Moreover, energy storage can conditions and offering a more diverse method of control [6,
be utilized to load shift and regulate network voltage. This paper
7]. Other techniques, aimed at more specific applications
investigates the use of battery storage in regulating network
voltage, in particular, the different strategies which may be include harmonic cancellation [3, 8] and economic driven
employed in controlling the storage unit. Simulations are coupled control [9].
with real data, taken from a rural single wire earth return Distribution companies supplying rural Australia
(SWER) network, showing that battery storage is capable of experience a number of unique challenges due to the sparsely
boosting the network voltage. populated, long distance networks required. A cost effective
1
method of electrifying rural customers, employed throughout
Index Terms—Battery storage, energy storage, control
strategy, distribution system.
Queensland, is that of single wire earth return (SWER) supply.
SWER networks provide electricity using a single conductor,
I. INTRODUCTION normally high tensile strength steel cable, completing the
circuit using an earth return path. The high tensile nature of
T HE need to reduce the effects of global warming, along
with ever-changing customer loads, continues to place
the conductor allows for long span lengths, reducing the
number of poles required and increasing cost efficiency [10].
immense stress on modern-day power networks. Australian Due to the poor conductivity of the conductor and long
state and national governments have put forward renewable distances involved, SWER networks often experience issues
generation as the key to solving many of these problems; with voltage regulation, namely over and under voltages.
however, as we are coming to realize, this alone is not the This paper provides an investigation into the use of battery
solution. storage for network voltage regulation of a rural feeder. In
Renewable generation, such as wind and photovoltaic (PV) particular, this study looks at the capabilities of solitary
is largely variable, rarely coinciding with peak customer storage (i.e. storage without PV generation), as well as energy
demands. The ability to store energy generated during periods storage with PV. In each of these instances, different control
of light load for use during peak load periods allows strategies for the storage systems are proposed and analyzed,
renewable generation to be more effectively utilized. Using with the aim of selecting the most effective control strategy.
this technology, it is conceivable that network operators will This paper is presented as follows: Section II provides an
begin to load shift, flattening the short, sharp peaks in load and overview of the case study system, listing the operating
improving overall network performance. specifications of the feeder and battery storage units. In
Achieving these goals and optimally utilizing renewable Section III, the development of static and dynamic feeder
generation and energy storage is heavily reliant on the models, the battery storage model, and storage control
development of an intelligent control strategy. A variety of strategies are discussed. Using these simulations and the
control strategies have already been proposed by researchers, related control strategies, the effect of battery storage on
normally with the aim of reducing network load or regulation regulating network voltage is simulated and analyzed in
of voltage. The first method, put forward by a number of Section IV. To justify energy storage from a cost perspective,
authors, is the transfer reactive power between storage and the Section V addresses preliminary economic analysis of some
network [1-4]. This technique exploits the relationship options for improvement of voltage regulation. Finally, major
outcomes and conclusions are highlighted in Section VI.
M. Zillmann (matthew.zillmann@gmail.com), R. Yan
(ruifeng@itee.uq.edu.au) and T. K. Saha (saha@itee.uq.edu.au) are with
School of Information Technology & Electrical Engineering, University of
Queensland, Australia.

978-1-4577-1002-5/11/$26.00 ©2011 IEEE


2

II. CASE STUDY NETWORK specifications of the network. Comparison of the data
presented, highlights that the storage units are significant in
A. Geographical Network Layout
comparison to the feeder loading. Furthermore, with a total
Fig. 1 shows a single line geographical diagram of the storage capacity of 160 kWh at 5 kVA rating per unit (40 kVA
SWER network. This feeder is responsible for supplying at total), the storage is able to reduce loading for a sustained
total of 93 of rural customers, with loads ranging from pumps, period of time.
to household connections, to small commercial loads. Like Table I shows the calculated conductor impedance for
most SWER systems, this network is characterized by long, 3/2/2.75 ACSR/GZ. Of particular interest is the high
lightly loaded lines with a low concentration of customers per impedance nature of this conductor, which introduces
km of line. This system also services eight ZnBr flow energy significant losses into the network, especially under heavy
storage systems [11], as indicated in Fig. 1. loading conditions.
These storage units have been positioned along the
backbone, with the majority being located toward the end of TABLE II
the SWER. Placement at the end of the feeder: BATTERY ENERGY STORAGE SPECIFICATIONS [12]
• Aids in reducing current flowing through the backbone, Battery Operating Specification
and resultantly the voltage drop along the feeder. Number of Storage Units 8
• Ensures storage is in close proximity to a major load Energy Storage Medium ZnBr Flow Battery
center, reducing losses in back feeding. Battery Nominal Voltage 48 V
Rating (Battery and Inverter) 5 kVA
Capacity 20 kWh

C. Network Measurements
Feeder Load Profile
160
Power (kW)

140
120
100
80
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
(a) Feeder Load Profile
PV Generation Profile
1.5
Power (kW)

0.5

0
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
(b) PV Generation Profile
Fig. 1. A rural SWER network geographical diagram with annotations [12] Fig. 2. Comparison of the SWER feeder load profile [12] and 2 kW PV
generation profile [12]
B. Network Data
Feeder Voltage Profile
TABLE I
A RURAL SWER NETWORK OPERATING SPECIFICATIONS [12]
1.02
1
Voltage (PU)

Feeder Operating Specification


Operating Voltage 12.7 kV
0.98
Isolator Rating 2×100 kVA
Maximum Demand (Summer) 190 kVA 0.96
Customer Transformers 91
Number of Customers 93 0.94
SWER Line Length 114.088 km
Conductor Type (3/4/2.75 ACSR/GZ) 0 5 10 15 20
Calculated Conductor Impedance Time (hours)
Resistance @ 60°C 2 Ω/km Fig. 3. The SWER feeder voltage profile as measured at the SWER isolator
Inductive Reactance 0.899 Ω/km [12]
Capacitive Susceptance 2.06 × 106 S/km
Fig. 2 highlights a standard summer day load curve, as
Table I and Table II provide an overview of the measured at the isolator of the SWER. Also shown, is the PV
3

generation profile for a normal summer day, taken from an models it was possible to note that each produced similar
experimental solar array at the University of Queensland. By voltage profiles. However, based on the observation and
comparing the two figures, it is evident that the peak in PV understanding, the increase in current with decrease in voltage
generation is largely divergent from the morning and evening ensured that constant power loads produced the most severe
peak in customer demand. conditions on the study network. Therefore, the assumption of
Fig. 3 shows the measured voltage profile as taken from the constant power to characterise customer loads was made on
12.7 kV side of the SWER isolating transformer position. This the basis that it provided a worst-case analysis of the system.
demonstrates the voltage sag occurring during the morning
B. Static Model
and evening peak in customer demand. This real data is
utilized in setting the bus voltage at the isolator of the As previously mentioned, the static model was developed
simulated network. in SINCAL’s load flow modeling. This model, developed to
emulate the worst case operating conditions of the network,
III. NETWORK MODEL (i.e. peak demand with lowest isolator voltage and no energy
storage or PV), was used as a base case for comparison against
To allow for an extensive investigation of the SWER
the dynamic models.
system and its performance under different operating
To verify the accuracy of the base case model, voltage
conditions, two models of the feeder were developed. The first
readings from auto reclosers and power quality meters were
was a static load flow model constructed in the electrical
cross-checked against the simulated network. This contrasting
network modeling software SINCAL [13]. The second was a
of real and simulated values verified the accuracy of the
dynamic (time-based) simulation developed in MATLAB/
model.
SIMULINK [14]. By comparing the performance of the
dynamic model to that of the base static load flow, the C. Dynamic Model
accuracy of the dynamic model was confirmed. The dynamic model was implemented in MATLAB/
A. Network Modeling Assumptions SIMULINK, with all conductor data and loading matching
that used in the static mode. The real measured voltage profile
In designing the SINCAL and SIMULINK network
(shown in Fig. 3) was incorporated through the use of a
models, the following major assumptions were made:
controlled voltage source and lookup table to simulate the
• Given that the isolator voltage was simulated using a isolator voltage profile. This initial dynamic model, once
measured voltage profile, the effect of the upstream verified, was utilized as the basis for all subsequent studies.
three-phase network was neglected.
• The customer load was proportional to the customer D. Battery Storage and Photovoltaic Model
transformer rating, allowing for load allocation based Following substantial research into the methods of
on transformer rating. modeling battery performance, the technique proposed in [15]
• All customer loads could be modeled as constant was selected for its simplicity and accuracy in modeling a
power. variety of battery types. This formulation, shown in (1),
• To emphasize the improvement in voltage performance characterized the internal battery voltage, based on the state of
due to energy storage, the voltage regulator was charge of the battery.
removed from simulations. Moreover, the vast majority
of SWER systems do not contain voltage regulators. (1)
Therefore, removing the regulator generalizes the
system, making the results more widely applicable to Where:
other SWER systems. E = internal battery voltage (V)
These assumptions imposed limitations on the outcomes of E0 = battery constant voltage (V)
this case study. The main limitation was incurred as a result of K = polarisation voltage (V)
neglecting the effect of the upstream three-phase network. The Q = battery capacity (Ah)
use of energy storage would have aided in the reduction of it = energy discharged from the battery (Ah)
losses in the three phase system, further improving network A = exponential zone voltage (V)
performance. B = exponential time constant inverse ((Ah)-1)
The local power distribution utility’s standard for SWER
customer loads is to utilise constant impedance in the Fig. 4 provides a comparison between the simulated battery
modelling process as they feel that this best models the loads voltage (solid) and the actual measured battery voltage
presented by rural customers. The rationale behind this (dashed) with different discharging currents, for a single
selection is in the nature of rural customers to shed load during discharge cycle.
periods of low network voltage performance in order to avoid This model provided a means of approximating the varied
damage to their appliances and equipment. battery performance as it was charged and discharged
In the development of the author’s model three load throughout the day.
characteristics were tested: constant impedance, constant
current and constant power. By comparing each of the three
4

Battery Discharge Curve


70
15A
60A
Battery Terminal Voltage (V)

60
80A
120A
50

40 (a) (b)
Fig. 5. (a) Tiers of operation for simple storage control; (b) Tiers of operation
30 for reactive power storage control

20 The final control strategy tested in this paper, combined PV


generation and energy storage to improve voltage
10 performance, while more effectively utilizing the energy
generated by PV. In this instance, PV was used to charge the
0 battery during the dead-band of operation, rather than
0 50 100 150 200 250
Time (Ahours) discharging onto the network. Fig. 6 highlights the control
Fig. 4. Simulated battery discharge (solid) and measured battery discharge logic and tiers of operation implemented for this strategy.
(dashed) comparison [12]
When grid voltage is higher than the preset upper limit,
batteries will be charged by both PV power and grid power. If
A controlled current source was employed to simulate the
the voltage drops between the upper limit and the lower limit,
PV unit. This model utilized a lookup table and the measured
batteries will be charged by PV systems alone in order to
generation data, shown in Fig. 2, replicating the generation of
avoid further voltage decrease caused by charging from the
a 2 kW PV array on a normal summer day.
grid. Battery discharge only occurs when grid voltage is below
E. Storage Control Strategies the lower voltage limit.
The current strategy utilized for control of the energy
storage systems is a time-of-day based control. This technique
requires hard coding fixed charge and discharge times into the
battery controller, based on observations of the customer load
profile. While this method may provide a good approximation
of the peak and light load periods, it does not offer any
flexibility in control. Moreover, the current state of the
network is not taken into account; as a result the energy
storage may act to degrade network performance rather than
improve it as intended. This paper looks at a more intuitive
and intelligent approach to controlling the available storage
systems.
Three control strategies were adopted with the aim of (a) (b)
improving network voltage performance. These strategies Fig. 6. (a) Control logic schematic; (b) Tiers of operation for combined PV-
were closely based on the multistage strategies proposed in [6, energy storage control
7]. Each of the control techniques monitors the voltage at the
point of connection, either charging, discharging, or remaining IV. SIMULATION RESULTS AND ANALYSIS
idle depending on voltage.
The first control strategy, shown in Fig. 5 (a), highlights the Incorporating the storage and PV models into the dynamic
various tiers of operation for simple solitary storage control. model of the SWER provided a means of testing the various
This strategy operates by transferring real power to or from control strategies and their influence on network performance.
the network depending on the measured grid voltage. A dead- The following outlines the performance for each of these
band is introduced to this strategy (and subsequent strategies) strategies, highlighting any improvements in performance
to prevent the hunting effect occurring between the charge and from the base network.
discharge states. A. Base Feeder Performance Without Storage or PV
The second control strategy, shown in Fig. 5 (b), is a To provide a basis for comparison the base network
variation of the simple control strategy, which discharges performance, that is, the performance of the network without
reactive power at a constant power factor into the network. For energy storage or PV was simulated. From this simulation,
the purpose of this paper, and to satisfy Australian standards multiple voltage profile measurements were taken from
[16], the power factor of 0.85 leading was selected. Using this various positions in the network. This highlighted any voltage
strategy, the effect of reactive power injection was compared performance issues that may be occurring on the network.
to that of pure real power injection. Furthermore, it provided a basis for comparison for each of the
5

subsequent storage control strategies. Fig. 7 shows the voltage Voltage Profile with Simple Control
profiles as measured at three locations around the SWER in 1
Base
Fig. 1. Simple
Base Feeder Voltage Profile 0.98
1.04
M1 Discharge Discharge
0.96

Voltage (PU)
1.02 M4
M11
1 0.94 Charge
Voltage (PU)

0.98
0.92
0.96
0.9
0.94
0.92 0.88
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
0.9 Fig. 8. Network voltage profile at the end of the feeder (M11) for the network
with energy storage using simple control (solid), compared to the base model
0.88 performance (dashed)
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
Fig. 7. Base network voltage profile at selected locations From Fig. 8, the boost in voltage at the end of the feeder
can be noted. Of particular interest, is the boost during the
From this voltage profile, two major events were noted, morning voltage sag. For this period the voltage is boosted
during the morning and night peak demands. For these periods from 0.889 pu to 0.900 pu, hence shifting the performance
the network voltage dropped significantly, resulting in quality from the “critical” to “poor” classification. This boost would
of supply (QoS) and possible voltage stability issues. help to improve the QoS for the customer and insure the
Investigation of the Australian Standard on network voltages voltage stability of the network.
[17] provides an indication of allowable per unit voltage levels
through a network under various operating conditions. Based C. Feeder Performance with Reactive Power Injection from
on this standard, a system of voltage performance Storage Inverters
classification was utilized, indicating the urgency for Reactive power injection into the network via four-
correctional intervention. quadrant inverters was utilized, with the aim of further
boosting the network voltage during the morning and evening
TABLE III voltage sag. Fig. 9 provides a comparison of the reactive
VOLTAGE PERFORMANCE CLASSIFICATION BANDING [17]
power control to the base model and simple control voltage
Performance Voltage (pu) Duration/Freq. Intervention performance.
Acceptable 1.05-0.94 Indefinite None Voltage Profile with Reactive Power Control
Poor 0.94-0.9 Limited/Infrequent Required 1
Base
Critical 0.9-0 Limited/Infrequent Immediate
Simple
0.98 PQ
This classification system was employed in setting the
upper and lower voltage limits for each of the storage Discharge Discharge
0.96
Voltage (PU)

controllers. Each of the eight storage units was individually


configured, with the aim of minimizing the operating time
spent in the “poor” and “critical” classifications. For the 0.94 Charge
majority of units, this meant setting the “upper limit” to 0.96
pu and the “lower limit” to 0.94 pu. Therefore, the storage 0.92
systems began injecting power into the network when the
voltage fell below 0.94 pu. 0.9

B. Feeder Performance with Simple Control for Storage 0.88


The first control strategy implemented and tested on the 0 2 4 6 8 10 12 14 16 18 20 22 24
simulated network, was that of the simple control strategy Time (hours)
Fig. 9 Network voltage profile at the end of the feeder (M11) for the network
(real power only). The positioning of the eight energy storage with energy storage using reactive power control (solid), compared to the base
units was as shown in Fig. 1. From this simulation, the point model (dashed) and simple control performance (dotted)
of worst voltage regulation, the end of the feeder (M11), was
monitored and the voltage performance compared to that of From Fig. 9, the voltage performance when injecting
the base model (Fig. 7). This comparison is shown in Fig. 8. reactive power into the network was seen to decrease in
comparison to the simple control strategy. This reduced
performance was due to the high R/X ratio of the feeder (≈ 2.3
6

from Table I), as discussed in [1]. When inductive components charge the battery unit. This helps to significantly reduce the
only account for a small fraction of the line impedance, loading on the feeder as each of the storage units being
reactive power compensation will not have any significant charging.
effects on voltage magnitude, because this cannot reduce the PV & Sotrage Current Flow
resistivity of the line, which is a major factor of causing 0.5
Grid Charge Discharge
voltage drop. As a result, the injection of reactive power had a Storage
minimal impact on the network voltage. 0.4 PV
D. Feeder Performance with Combined PV-Energy Storage

|Current| (A)
The grouping of PV generation and energy storage was 0.3
investigated using the proposed combined control strategy.
Discharge
For this simulation, a 2 kW PV unit was placed with each of
0.2 PV
the eight storage units. The generation profile for each PV Charge
installation was assumed to be the same throughout, as given
in Fig. 2. Fig. 10 demonstrates the performance of the network 0.1
with PV only, compared to that of the PV-energy storage
combination.
From the PV only curve of Fig. 10, it is evident that the PV 0
0 2 4 6 8 10 12 14 16 18 20 22 24
generation acts to boost the network voltage during the 10:00 Time (hours)
AM to 2:00 PM period of the day. During this time the Fig. 11. Simulated current flow between the gird (solid), storage (dashed), and
PV (dotted) for one PV-energy storage unit, (assuming 12.7 kV coupling)
network was lightly loaded, with “acceptable” voltage
performance. As such, the PV alone did not act to significantly
improve the voltage performance of the network during peak E. Feeder Performance with Increased Penetration of Storage
times. Given periods of abnormally light loading or an To demonstrate the possibility for further improvements in
increased penetration of PV, this may act to degrade the performance, as well as investigating the effect of high
performance of the network, causing over voltage QoS issues. penetration energy storage, the network model was modified
Voltage Profile with PV and Energy Storage by repositioning and increasing the number of storage units.
1 An additional eight battery units were added along the end of
Base
PV & Storage the feeder backbone, with all existing storage units
0.98 PV only repositioned to be directly connected to the backbone. Fig. 12
shows the voltage profile at the end of the feeder (M11)
Discharge Discharge
0.96 comparing the performance with and without energy storage.
Voltage (PU)

Voltage Profile with Increased Penetration


0.94 1.06
Charge M11-Base
1.04 M1
0.92 M11-Storage
PV & Storage 1.02
0.9 1
Voltage (PU)

Base
0.98
0.88
0 2 4 6 8 10 12 14 16 18 20 22 24 0.96
Time (hours)
Fig. 10. Network voltage profile at the end of the feeder (M11) for the network 0.94
with combined energy storage and PV generation (solid), compared to the base
model (dashed) and PV only performance (dotted) 0.92
0.9
Fig. 10 also demonstrates that the performance of the
network with energy storage and PV is similar to that of the 0.88
0 2 4 6 8 10 12 14 16 18 20 22 24
simple control strategy, especially during the morning and Time (hours)
evening voltage sag. The main improvement in performance is Fig. 12. Network voltage profile at the start (M1) and end (M11) of the feeder,
seen during the charge cycle. For this period the battery is comparing the performance of the based model (dashed) and the adjusted, high
penetration model (solid)
charged from the PV and the grid, hence reducing the storage
unit’s loading on the network. This reduction in loading is
From Fig. 12 it can be seen that the voltage during the
demonstrated in Fig. 11.
morning sag is boosted by about 2.5%. This demonstrates the
Fig. 11 illustrates that the current flow to the grid during
possibility for increased penetration of energy storage to
the morning and evening operation is mainly from the energy
provide improved network voltage performance. It is
storage, with the PV having minimal effect on the voltage
important to note that the upstream network was not modeled
performance. However, during the charge cycle, the current
in this simulation. If the full feeder were to be modeled, the
drawn from the grid is reduced as the PV generation is used to
7

reduction in line current during this period would reduce the improvements to customers along the backbone, the voltage
voltage drop along the three-phase network, further improving drop on the original conductor would make customers at the
voltage performance at the end of the feeder (M11). This is a extremities of the system still prone to poor voltage
limitation of this study. performance. To achieve major improvement in the voltage
performance of the network would require the re-conductoring
F. Improvement in Voltage Performance
of the majority of the SWER backbone, given the low
Comparison of the three control strategies, as given in customer density per km of line and the high associated cost;
Table IV, highlights that the optimal control strategy for this is rarely the preferred option on SWER systems.
improving the network performance is the combined PV- The alternate proposal of installing a regulator, at a cost of
energy storage control. As previously mentioned, the $100,000 AUD, would be the most cost-effective of the three
performance during this period is similar to that of the simple options. However, it is standard policy within the local power
control strategy. This is due to the divergence between the distribution utility to allow, at most, three cascaded regulators
peak in PV generation and the peaks in customer demand. on any feeder to avoid “voltage hunting”. Given that there are
Furthermore, the high R/X ratio of the feeder limits the two closed-delta regulators prior to the SWER isolator on the
effectiveness of four-quadrant inverters injecting reactive three-phase backbone, the PV/battery systems were trialled as
power to boost the network voltage. As a result, reactive a slightly more expensive alternative.
power control is largely unsuccessful in regulating the
network voltage. TABLE V
Higher storage penetration level (16 units) showed a 2.6% ESTIMATED COST OF VOLTAGE IMPROVEMENT METHODS
voltage improvement compared with baseline case and 1.4 % Methods Cost of one unit (AUD)*
with the case of 8 storage units. Therefore, with more storage PV (2kW) and ZnBr Battery $30,000 - $35,000
integrated into the network, voltage profiles of the network (5kVA - 20kWh)
have been improved. However, due to the low voltage (0.926 Re-conductor from 3/4/2.75 $25,000/km
ACSR/GZ to 4/3/3.0 ACSR/GZ
pu) at the beginning of the feeder (M1), the voltage magnitude SWER Regulator (12.7kV) $100,000
at the end of the feeder (M11) can only be increased to 0.915
pu with 16 storage units. Unless more storage systems are * Data obtained from the local power utility and industries in Jan 2011.
installed to result in considerable reverse power flow, the
voltage at M11 will not be boosted over 0.926 pu. Therefore,
the downstream scheme of voltage control has its limitation, VI. CONCLUSIONS
and further improvement should consider the regulation of the This paper provided a case study of a real world SWER
upstream voltage. network with measured data, demonstrating the effect of
energy storage and PV on the voltage performance of the
TABLE IV
system. Three main control strategies were addressed, with
COMPARISON OF VARIOUS CONTROL STRATEGIES PERFORMANCE FOR
MORNING VOLTAGE SAG each utilizing different techniques to boost the network
voltage.
Control Strategy Min. Simulated Voltage From the simulations given, it was shown that the use of a
Isolator Voltage (M1) 0.926 pu simple, multistage control strategy with solitary energy
Base Model (M11) 0.889 pu
Simple Control (M11) 0.900 pu
storage could boost the network voltage for short periods.
Reactive Power Control (M11) 0.891 pu Moreover, when the storage was combined with a PV unit, the
PV-Storage Control (M11) 0.901 pu energy generated by the PV cell was shown to be more
16 Storage Units (M11) 0.915 pu optimally utilized by charging the battery for later export onto
the network. Increasing the penetration and repositioning
energy storage were shown to achieve more effective
V. PRELIMINARY ECONOMIC ANALYSIS regulation of the network voltage. However, due to the
There are several methods of improving voltage profiles of limitations of the study, the full extent of performance
the SWER system, for example energy storage, network re- increase was not demonstrated.
conductoring and additional regulator. Next, economic Of particular interest, was the finding regarding the
perspectives of these options are discussed to shed some light injection of reactive power into the network to boost the
on a cost-effective point of view. voltage. Simulation of this control strategy illustrated that, for
The estimated cost of the methods is shown in Table V. An this network, introducing additional reactive power into the
estimate cost of installation of a PV and battery storage system network had minimal effect on the voltage performance.
is approximately $30,000 - $35,000 AUD. Taking the higher Therefore, the following should be considered when
end of this range and considering that the case study network employing energy storage for use in transmission and
currently has eight systems installed puts the total cost at distribution applications:
around $280,000 AUD. (1) The energy storage capacity available to the network:
At this price it would be possible to re-conductor 11.2km of This will largely determine the strategy adopted for
the total 114km system with 4/3/3.0 ACSR/GZ Sultana, at a control of the storage units.
cost of $25,000/km AUD. While this may offer some voltage (2) The R/X ratio of the feeder: For SWER systems it is
8

advised, due to the high R/X ratio of the conductor generation system”, Electrical Power & Energy Conference, EPEC
2009. IEEE, pp. 1-6, 2009.
used, that reactive power not be utilized for the [6] T. Tran-Quoc, T.M.C. Le, C. Kieny, N. Hadjsaid, S. Bacha, C.
improvement of voltage performance. For feeders Duvauchelle, and A. Almeida, “Local voltage control of PVS in
with lower R/X ratio, this may not be the case. distribution networks”, 20th International Conference on Electricity
Distribution, CIRED, pp. 1-4, 2009.
(3) The use of solitary storage or in combination with PV [7] P.N. Vovos, A.E. Kiprakis, A.R. Wallace, and G.P. Harrison,
generation: The availability of PV for charging “Centralized and Distributed Voltage Control: Impact on Distributed
storage reduces the loading of storage on the network, Generation Penetration”, IEEE Trans. Power Systems, vol. 22, pp. 476-
while enhancing the effectiveness of PV. 483, 2007.
[8] M. Prodanovic, K. De Brabandere, J. Van Den Keybus, T. Green, and J.
(4) The positioning of the energy storage units: Where Driesen, “Harmonic and reactive power compensation as ancillary
the positioning of storage and PV is controllable, it services in inverter-based distributed generation”, Generation,
should be placed in close proximity to load centers, Transmission and Distribution, IET, vol. 1, pp. 432-438, 2007.
[9] B. Lu and M. Shahidehpour, “Short-term scheduling of battery in a grid-
preferably on the backbone of the feeder to reduce connected PV/battery system”, IEEE Trans. Power Systems, vol. 20, pp.
voltage drop. 1053-1061, 2005.
Future work will focus on following issues: [10] Electricity Authority of New South Wales, High voltage earth return
distribution for rural areas, 4th Edition, 1978.
(1) Further investigation and evaluation into the use of [11] Redflow, http://www.redflow.com.au/.
existing methods (e.g. re-enabling voltage regulator [12] Internal information, Ergon Energy, 2010.
at M4 in Fig. 1) should be carried out, comparing the [13] SINCAL,http://www.energy.siemens.com/hq/en/services/power-transmi
ssion-distribution/power-technologies-international/software-solutions/p
results achieved using these traditional methods to ss-sincal.htm.
that of the PV and Battery system. Moreover, the cost [14] MATLAB/SIMULINK, http://www.mathworks.com/products/simulink.
and inherent gains of such options should be [15] O. Tremblay, L.-A. Dessaint, A.-I. Dekkiche, “A Generic Battery Model
for the Dynamic Simulation of Hybrid Electric Vehicles”, Vehicle
thoroughly considered and analysed. Power and Propulsion Conference, VPPC 2007. IEEE, pp. 284-289,
(2) The effect of increasing the penetration of PV and 2007.
energy storage: Higher penetrations allow for [16] Grid connection of energy systems via inverters, Part 2, Australian
improved control of voltage, but may raise issues Standard AS 4777.2-2005, May 2005.
[17] Standard voltages, Australian Standard AS 60038-2000, Feb. 2000.
with reverse current flow, increasing the risk of
overvoltage occurring on the network. IX. BIOGRAPHIES
(3) Research into the placement of storage units: An
analytical solution to the optimal positioning of Matthew Zillmann (S’2010) is a final year Bachelor
storage units would allow for more efficient of Engineering (Hons.) student studying at the
application of storage and PV. University of Queensland, Brisbane, Australia.
During his studies, he has specialized in Power
(4) Operation of energy storage in urban environment: Systems Engineering. His interests include,
The case study addressed in this paper is of a rural renewable energy integration, applications of energy
feeder with few customers per km of line. Networks storage and challenges facing modern day power
systems.
with higher customer density may increase the
efficiency of back feeding power. Moreover,
utilization in this environment may require the
adaptation of more diverse control strategies. Ruifeng Yan (S’2009) received the B. Eng. (Hons.)
degree in Automation from University of Science
and Technology, Beijing, China, in 2004 and the M.
VII. ACKNOWLEDGEMENT Eng degree in Electrical Engineering from the
The authors would like to express their gratitude to S. Australian National University, Canberra, Australia,
in 2007. He is currently pursuing Ph.D. degree in
Richardson and G. Walden of Ergon Energy, and Redflow for power and energy systems at the University of
their assistance in completing this work. Queensland, Brisbane, Australia.

VIII. REFERENCES Tapan K. Saha (SM’97) was born in Bangladesh


and immigrated to Australia in 1989. Currently, he
[1] R. Tonkoski and L.A.C. Lopes, “Voltage Regulation in Radial is Professor of Electrical Engineering in the School
Distribution Feeders with High Penetration of Photovoltaic”, Energy of Information Technology and Electrical
2030 Conference, ENERGY 2008. IEEE, pp. 1-7, 2008. Engineering, University of Queensland, Brisbane,
[2] G.C. Pyo, H.W. Kang, and S.I. Moon, “A new operation method for
Australia. Before joining the University of
grid-connected PV system considering voltage regulation in distribution
Queensland, he taught at the Bangladesh University
system”, Power and Energy Society General Meeting - Conversion and
of Engineering and Technology, Dhaka, for three-
Delivery of Electrical Energy in the 21st Century, IEEE, pp. 1-7, 2008.
[3] E. Demirok, D. Sera, R. Teodorescu, P. Rodriguez, and U. Borup, and half years and then at James Cook University,
“Clustered PV inverters in LV networks: An overview of impacts and Townsville, Australia, for two and half years. His
comparison of voltage control strategies”, Electrical Power & Energy research interests include power systems, power
Conference (EPEC), 2009 IEEE, pp. 1-6, 2009. quality, and condition monitoring of electrical
[4] K. Turitsyn, P. Sulc, S. Backhaus, and M. Chertkov, “Distributed control plants. Dr. Saha is a Fellow of the Institution of Engineers, Australia.
of reactive power flow in a radial distribution circuit with high
photovoltaic penetration”, Power and Energy Society General Meeting,
2010 IEEE, pp. 1-6, 2010.
[5] V. Khadkikar, R.K. Varma, R. Seethapathy, “Grid voltage regulation
utilizing storage batteries in PV solar — Wind plant based distributed

You might also like