Engineering, 2013, 5, 208-214
doi:10.4236/eng.2013.51b038 Published Online January 2013 (http://www.SciRP.org/journal/eng)
Coordinated Voltage Control in Distribution Network with
Renewable Energy Ba sed Dist rib ut ed Gen eration
Mohd Khairun Nizam Mohd Sarmin1, Worawat Nakawiro1, Mohd Zamri Che Wanik2,
Mohd FadzilMohd Siam1, ZahrulFaiziHussien1, Ahmad Asrul Ibrahim3, Ahmad Kamil Mat Hussin2
1Smart Grid Sectio n, TNB Research Malaysia, M alaysia
2Green Technology Section,TNB Research Malaysi a, Malaysia
3Power system research group, Uni vers itiKeb angsaan Malaysia, Mal aysia
Received 2013
ABSTRACT
This paper presentsa voltage control strategy for power distribution systems with interconnected renewable energy
based distributed generators (DGs). The control strategy coordinating conventional voltage control devices and reactive
po wer from DG.A mixed-integer nonlinear programming problem was formulated and solved by particle swarm opti-
mization (PSO). The code is written using DigSILENT programming language (DPL) and implemented inside DigSI-
LENT po wer factory simulatio n software. All syste m constraints and operating limits are considered. The optimal pow-
er flow based approach can incorporate various uncertainties such as intermittent po wer characteristics and varying load
demand. The proposed method is tested using real distribution network to demonstrate its effectiveness. T he merits of
the proposed method over the classical local-based control are presented in the simulation results. It is demonstrated
that the proposed method is capable of keeping the system voltage within operating limit. Power losses is at the same
time is minimized in comparison to the losses using conventional method.
Keywords: Photovoltaic; D istributed G e ner a tion; Particle Swarm Optimisation; Di gSI LE NT; Coordinated Control
1. Introduction
Renewable resources has gained significant attention in
recent years due to the cost increment and adverse en-
vironmental impacts of conventional fossil fuels [1].
Currently, Photovoltaic (PV)energy systems are considered
as opti mum sol ution to the ele ctricity supply i n most r ural
zones in developing countries [2,3].
PV application is gaining a lot of attention in the de-
veloping countr y like Mala ysia. Feed in Tariff introduced
in Malaysia in 2011 [5] is a sign that the government is
serious in promoting PV as a new energy source in sup-
porting Malaysian sustainable growth. Typicallya re-
newable energy resource such as PV system generatese-
lectricity as distributed generator (DG) is connected to
distribution networks. DG application is highlighted of-
fers various benefits to the distribution network where it
is connected to but harve sting the benefits is not without
chall enges [ 6].
Voltage contro l is one of well-documented operational
challenges [7]. Currently the voltage control strategies
are also developed based on the assumption of unidirec-
tional power flow. However in some extreme conditions
due to the intermittent and non-dispatchable characteris-
tics the excess DG power could flow in a reverse direc-
tion to the transformer high voltage side. Therefore a
new scheme for voltage control is definitely needed in
modern power distribution systems which comprising
DG. Some of these control schemes can be broadly cate-
gorized as tabulated in Table 1.
From practical viewpoints, the offline planning ap-
proach should be deployed in complimentary with an
advanced online contro ller. This coord ination will enable
the system to achieve economic operation and guarantee
security when subject to unexpected disturbances.
In this p a pe r , a n o ffline ( d a y-ahead) planning approach
based on particle swarm optimization (PSO) is proposed.
The approach determines optimal settings of all control
devices including on-load tap changers, switched shunt
capacitors, commonly found in most power distribution
network. System state quantities are considered and op-
erational limits is incorporated. PSO code is written in
DigSILENT programming language (DPL) and the pro-
posed approach is implemented in DigSILENT Power
Factory [15].
2. Voltage and Reactive Power Control
The concept of power loses can be explained by referring
to Figure 1. In a simple radial networ k shown in Figure

M. K. N. M. SARMIN ET AL.
209
1 the voltage drop across the feeder can be approximated
by [16-17]:
2
LL DG
RPX QQ
UU
⋅+⋅ −
∆≈
()
(1)
where R and X are the line resistance and reactance, re-
spectively, and P and Q are the active and reactive power
generated from DG. It can be seen that any fluctuation in
reactive power will impact the voltage fluctuatio n. If the
constant power factor control is used, the ratio P/Q is
maintained. This tends to increase reactive power of DG
and there by aggravating the v olta ge rise.
When reactive sources available properly compensate
the reactive power demand, the feeder current will de-
crease according to:
22
2
LL DG
P QQ
IU
+−
=()
(2)
Therefore the feeder losses PLoss will also decrease be-
cause losses are directly related to current as follow:
2
Loss
P IR= (3)
3. Coordinated Voltage Control
Figure 2 shows three alternatives for coordinated voltage
control (CVC) in a distribution network with DG. The
baseline of this scheme is to control the on-load tap
changer (OLTC) position of transformer and substation
capacitors CS via SCADA as depicted in the red arrow.
However if the communication channel is available it
Table 1. Comparison of distribution network voltage con-
trol scheme.
Offline Planning
Online Control
Objective
Determine optimal
control set points to
achi eve min/max objec-
ti v e( s) whereby main-
taining all constr aints
(Optimal Power Flow)
Find the recourse of
controllers to achieve
th e t arget objective (i.e.
to elimina te voltage
violati on, line loa ding,
etc).
Time
scale Day-ahead Online and closed loop
Inputs/
Data
require-
ments
Offline : Complete
network info, load
forecast, predicted
power generation
(i.e.solar)
Offline : Histor ical data ,
operat ion rules and
expertise and/or simula-
tion studies.
Online: S el ected
measurements
Example
Opt imiza tion algorithm
• Mat hematical pro-
grammi ng (i.e.[ 8])
• Heuristic me-
th o ds(i.e.[ 9-10])
Intelligent system
• Multi agent ( i.e.[11])
• Ar tif icia l ne ur a l net-
work (i.e.[12])
• Fuzzy logic
(i.e.[13-14])
R+jX
P
DG
+jX
DG
P
L
+jQ
L
I
Grid U
1
U
2
Figure 1 . A simple network showi ng vol tage drop.
Figure 2. Coordinated voltage control with DG sources.
may be possible to control the capacitors and energy sto-
rage devices installed at feeders (see the blue arrows).
Moreover if DG can participate in ancillary service s, DG
reactive power output can also be controlled.
4. Optimization Formulation
The optimal coordination of all control devices can be
determined by solving the formulated optimal power
flow problem. The control variables consist of OLTC
position, status of substation capacitorsCs and feeder
capacitors CF and DG reactive po wer out put s (co ntinuous
variables). The objective can be set to minimize the
power loss Ploss as shown:
, ,,
( ,,,);
losst ttDtDGt
P ftT= ∀∈Min xud d
(4)
where all the subscripts trepresents the planning time
interval t in the set of all time intervals in a day T. The
vector xt represents the state variables as listed in (6)-(9)
at time t. The vector ut represents the control variables at
time tgiven by:
,,
,,, ,
tt tttiti

= uTap Cf Cs Qgθpv
(5)
where the subscripts i represent the index of fast (15
minutes; i = 1,...,4) response controllers. At the consid-
ered time interval t, Taptis the vecto r of all OLTC trans-
formers; Cftis the vector of all feeder capacitors; andCstis
the vector of all feeder capacitors. These discrete con-
tro llers ha ve a slo w resp onse t ime (i. e. 1 ho ur li ke in thi s
study). On the other hand, the continuous variables have
by nature fast response. Therefore, these controls are

M. K. N. M. SARMIN ET AL.
210
discretized with a smaller time step i (15 minutes; i =
1,...,4 for any planning hour t). At time t, the vector
Qgt,icontains reactive power output for all synchronous
DGs and the vector θpvt,i contains the power factor set
point for all solar PV co nverter.
The network should be operated within a narrow band
of voltage variation to ensure safety of power system
equipments and supply quality
min , maxit
U UUit≤≤∀∈ ∀∈ ,
B
NT
(6)
whereUi,tis t he volta ge o f b usi at time t; Uminand Umax are
mini mum and maxi mum allo wable voltages, respectively
and NB is the set of controlled buses.
The current flo w in all lines must be maintained b elo w
the rating limit
,, ,,Ljt Lj
II jt≤∀∈ ∀∈
rat
,
L
NT
(7)
where IL,j,tis the current flow on line j at time t; IL,j,rat is
the the rma l capacity of line j and NL is the set of all lines.
The apparent power transfer on the substation trans-
former is limited to prevent any overloading
,t
SS t≤ ∀∈
TX TX,rat
, T
(8)
where STX,t is the apparent power flow on substation
transformer at time t and STX,rat is the capacity rating of
the substation transformer.
When DGs are centrally dispatched, the operating re-
gion is restricted by the power factor operating limits
described as:
min,maxmin, maxkt kt
PFPF PF
kt
θ θθ ≤≤⇒ ≤≤
∀ ∈∀∈
DG
NT
(9)
wherePFmin and PFmax are minimum and maximum op-
erating power factor of DGs (here assumed common for
all DGs for simplicity), respectively. To consider induc-
tive and capacitive operating range of DGs, the power
factor angle
θ is preferred because it is positive for lag-
ging (inductive) and leading (capacitive) power factors.
The angle
θ is constrained in the range between the min-
imum and operating power factor angles [
θ
min,
θ
max ]. The
NDGrepresents the set of all DGs.
As mentioned earlier that the proposed control meth-
odology will provide the set of opti mal control set -points
for the day-ahead operation. Therefore the necessary
input data is forecast of the load demand and predicted
generation pattern to be dispatched for the next 24 hours.
The developed optimization algorithm will iteratively
search for the best control variables which resulting in
the minimum total power losses while maintaining all
operational bounds and system security constraints.
The heuristic optimization approach is preferred be-
cause of its ability to handle complex problems without
any need for an explicit mathematical model. For every
iteration, the power system analysis (DigSILENT) is
called and load flow simulation is performed to deter-
mine the power system states corresponding to the con-
trol variable s updated by the optimization algorith m. T he
outputs of this method are the optimal position of trans-
former tap changers and the optimal status of capacitors
for the next da y of ope ration. The exp lanation given ear-
lier can be depicted as in Figure 3.
5. PSO Implementation
PSO algorith m flow char t is d epicted in Figure 4. A par-
ticle iflies in the search space between two successive
iterations acco rding to:
( )()( )
11
i ii
t tt+=+ +xxv
(10)
The or iginal eq uatio n for velocity update is given by:
()( )
( )
( )
1
2
1() ()
() ()
i
ttctt
c tt
+= +−
+−
ii1pi
2g i
vvrx x
rx x (11)
The randomness in the search procedure is introduced
by two independent uniform random sequences, r1 and r2
in the range (0, 1). The weighting coefficients c1 and c2
are the acceleration coefficients which control the influ-
ence of cognitive and social parameters on the particle’s
velocity.
Figure 3. Co nceptual diagram of the developed OPF .
Figure4 . PSO algorithm flow chart.

M. K. N. M. SARMIN ET AL.
211
In thi s wo rk, t he sta nd ard PSO versi o n 200 6 was used .
The parameters w and c remain unchanged from [18].
The global best position in the swarm xg is considered as
xg∈{xp1, xp2 ,…….., xps} suc h that
( )()
12
fmin f(),f(),........,f()=
gp pps
xxx x
(12)
Constraint handli ng is the most important pa rt in sol v-
ing a constrained optimization problem. Fitness functions
are used to assign a qualitative measure to individuals in
the population. To properly handle constraints in PSO,
the fitness function should be carefully designed such
that it can help guide the swarm process to the promising
and feasible search space.
The self-learning penalty function proposed in [19] is
applied. The elegant feature of this technique is that it is
parameter-less and capable of adjusting the penalized
fitness function at different stages of the search process.
6. Distribution Network
This section presents the network model used for simula-
tions in this research work. A part of real distribution
network in middle region of peninsular Malaysia was
chosen as the test network and the schematic diagram is
shown in Figure 5. This is a distribution network sup-
plied by two substations. The system basically operates
as a radial network with options for feedback from the
other 33/11 kV substation in case of breakdown.
Since our assumption is to have high degree of DG
penetration, therefore 5x1 MVA synchronous DGs and
5x1 MW solar PV are assumed to be installed at various
loca tio ns thro ug hout t he net w or k. All DG s and sol ar P Vs
are connected at 11 kV busbars. The 11 kV busbar is
supplied by two 33/11 kV transformers namely Tr(L1)
and Tr (L2).
In the simulation, all the loads are assumed to be of
constant power type with the power factor of 0.9 induc-
tive. Average power factor was stated to be 0.9. A future-
istic load pattern with the peak of 12.8 MW (presently
about 5 MW) is assumed as shown in Figure 6(a). The
power output for each solar PV is illustrated in Fig-
ure6(b) . This data is obtained by fitting the real-meas-
urements from a test unit of 3 kW solar PV into the nor-
mal (Gaussian) distribution. This unit is installed on the
rooftop one of the building at UniversitiTenagaNasional
main ca mpus.
This assumption may not be accurate since the statis-
tical characteristics of solar energy is quite complicated
and varied due to several factors. Finally, the power gen-
eration dispatch of synchronous DG is included in the
study scope. Therefore, two generation levels are as-
sumed fo r all syn chro no us D Gs as s hown in Figure 6( c).
DigSILENT Power Factory [15] is used as the power
system analysis software and the PSO code is written in
DP L.
Figure 5. Schematic diagram of the test net work.

M. K. N. M. SARMIN ET AL.
212
7. Simulation Results
7.1. Losses and Cost Comparison
This section demonstrates the effectiveness of the devel-
oped OPF method based on PSO (named ‘PSO’ hereafter)
in co mparison with the existing control method based on
local information (named ‘Local’ hereafter). Two cases
are considered namely ‘noDG’ (no DG is connected) and
‘withDG’ (all DGs a nd PVs are co nnected). For the local
control method, solar PV and synchronous DG are as-
sumed to operate in the capacitive mode with 0.9 power
factor.
Figure 7 displays the total power losses for the daily
operation according to the generation and load pattern
shown in Figure 6. It is obviously shown that based on
the local approach synchronous DGs do not play a role in
reducing the losses during the night times. This is be-
cause they are not located close to the load center. On the
other hand, the solar PV with a fixed power factor con-
trol can help reduce losses significantly during day time.
This is due to their stra tegic locations near to load c e nters
in the feeders 5 and 6.
With the PSO approach, it is very clear that the power
losses are minimized throughout the day. This is due to
the optimal settings of all control variables. The total
energy losses (in terms of kWh) of the four cases in Fig-
ure 7 are calculated as shown in Table 2 demonstrates
the sa ving i n cost due to ener gy losse s. T he energy p rice
is assumed to be RM 0.3/kWh. It can be shown that the
savings due to introduction of PSO are 3.06% in case of
no DG and 2.94% in case with DG.
Figure 6. Input data (a) total load demand (b) solar power
output of each PV (c) power dispatch of each synchronous
DG.
7.2. Optimal operation
The optimal setting of all control variables at each time
interval is the output of the developed methodology. The
optimal tap position for the transformer Tr(L1) is shown
in Figure 8. Notice that two changes are required in the
local method whereby the total number of tap changes is
11 for the PSO method. This number of change is ac-
ceptable in real operational practices for a transformer
with similar r a ting [13]).
7.3. Voltage Profile
Voltage magnitudes at the two substations are compared
based on the local and PSO methods as shown in Figure
9. It shows that the substation voltage s based o n the local
control method a re maintained at a higher level to ensure
that voltage at the last bus of all feed ers is not belo w the
statutory limit (-5%). Ho wever, the s ubstation volta ge in
the PSO method is on lower average. This is beca use the
PSO optimizes all available reactive sources in the net-
work and help support voltage at different buses. More-
over the reactive current will be reduced due to the op-
timal reactive and voltage support. Therefore it reduces
the current magnitudes in most feeders and thereby re-
duces the total power losses.
Figure 7. Power losses for a daily operation based on dif-
ferent control strategies.
Table 2. Cost comparison.
Method Energy losses
(kWh)
Energy
Cost
(RM/kWh)
Cost of
energy
losses
(RM/d ay)
Cost
saving
(%)
local_noD G 21047.10
0.3
6314.13 3.06
pso_noDG 20402.21 6120.66
local_withDG 20428.17 6128.45 2.94
pso_withDG 19826.97 5948.09

M. K. N. M. SARMIN ET AL.
213
Voltage profiles at the last bus of the feeder 5 are
shown in Figure 10. It is clear that the voltage level of
the PSO method is lower than the local method. This
figure also shows possibility of experiencing an under
voltage problem when the solar power output highly
fluctuates.
8. Conclusion
This paper focuses on the offline (day-ahead) planning
approach. An optimal power flow (OPF) is formulated to
Figure 8. optimal control set points f or (a) OLTC transfor-
mer (b) substatio n capacitor(c) solar PV inverter.
Figure 9. S ubstation voltage s.
Figure 10. Voltage at one of the feeders.
optimize the desired objectives such by searching for the
optimal schedule of all control devices. Power system
state quantities are considered and operational limits is
incorporated. The optimal power flow based approach
can incorporate various uncertainties suc h as intermitte nt
power characteristics and varying load demand. It is
demonstrated that the proposed method is capable of
minimizing power losses and voltage deviation in com-
parison to the conventional method. However the main
obstacle for this approach is that it relies on sensors and
communication infrastructures which may not be readily
available in many power util ities.
9. Acknowledgement
This project is funded by TNB Research Malaysia
through research funding TNBR/SF 52/2012 and partly
supported by UniversitiKebangsaanMalaysia via re-
sear ch gra nt G GP M-2011-071. Dr.-Ing. WorawatNakawiro
was a principal researcher at TNB Research Malaysia
between Sept 2011 and Aug.2012.
REFERENCES
[1] T. J. Hammons, J. C. Boyer, S. R. Conners, M. Davies, M.
Ellis, M. Fraser, E. A. Holt, and J. Markard, "Renewable
energy alternatives for developed countries," IEEE
Transactions on Energy Conversion, Vol. 15, No. 4,
December 2000, pp. 481-493.
[2] C. Boccaletti, G. Fabbri, J.Marco, and E. Santini, "An
Overview on Renewable Energy Technologies for
Developing Countries: the case of Guinea Bissau,"
International Conference on Renewable Energies and
Power Quality, Santander, Spain, 2008.
[3] D. N. Nkwetta, M. Smyth, and Vu Van Thong,
"Electricity supply, irregularities, and the prospect for

M. K. N. M. SARMIN ET AL.
214
solar energy and energy sustainability in Sub-Saharan
Africa," Journal of renewable and sustainable energy,
Vol. 02, 23 Ma rc h 2010, pp. 16.
[4] Subiyanto, A. Mohamed, and M. Hannan, "Intelligent
maximum power point tracking for PV system using
Hopfield neural network optimized fuzzy logic
controller," Energy and Buildings, Vol. 51, 2012, pp.
29-38.
[5] "Renewable Energy Act 2011, Law of Malaysia, Act
725," 2011. www.seda.gov.my
[6] M. Z. C. Wanik, "Simulation and Management of
Distributed Generation: Green Energy Integration to
Electri cal Power Syst em," Lambe rt Academic P ublis hing,
Saarbrueken, Germany, 2011.
[7] T. Niknam, A. M. Ranjbar, A.R. Shirani, “Impact of
Distributed Generation on Volt/Var Control in
Distribution Network,” Proceedings of Power Tech
Conference of the IEEEPES, Bolongna, 23-26 June 2003,
pp. 7.
[8] M. B. Liu, C. A. Canizares and W. Huang, “Reactive
Power and Voltage Control in Distribution Systems with
Limited Switching Operations”, IEEE Transactions on
Power Systems, Vol. 24, No. 2, May 2009, pp. 889-899.
[9] Y.-Y. Hong, K.-L. Pen, “Optimal VAR Planning
Considering Intermittent Wind Power using Markov
Model and Quantum Evolutionary Algorithm,” IEEE
Transactions on Power Delivery, Vol. 25, No. 4, Oct
2010, pp . 2987-2996.
[10] Y.-Y. Hong and Y.-F. Luo, “Optimal VAR Control
Considering Wind Farms using Probabilistic Load Flow
and Gray-Based Genetic Algori thms”, IEEE Transactions
on Power Delivery, Vol. 24, No. 3, July 2009, pp.
1441-1449.
[11] M.E. Baran and I. M. El-Markabi, “A Multi agent based
Dispatching Scheme for Distributed Generators for
Voltage Support on Distribution Feeders”, IEEE
Transactions on Power Systems, Vol. 22, No.1, Feb. 2007,
pp. 52-59.
[12] G. W. Kim and K. Y. Lee, “Coordination Control of
ULTC Transformer and STATCOM based on an
Artificial N eural Net work,” IEEE Transactions on Power
Systems, Vol. 20, No.2, May 2005, pp. 580-586.
[13] R.-H. Liang and Y.-S. Wang, “Fuzzy-based Reactive
Power and Voltage Control in a Distribution System,”
IEEE Transactions on Power Delivery, Vol. 18, No. 2,
April 2003, pp. 610-618.
[14] D. H. Spatti, I. N. da Silva, W. F. Usida, R. A. Flauzino,
“Real-Time Voltage Regulation in Power Distribution
using Fuzzy Control”, IEEE Transactions on Power
Delivery, Vol. 25, No. 2, April 2010, pp. 1112-1123.
[15] User Mannual, DiGSilent Power Factory v. 14.1,
DigSilent GmbH, Gomaringen, Germany, May 2011.
[16] F. A. Viawan, “Voltage Control and Voltage Stability of
Power Distribution Systems in the Presence of
Distributed Generation”, PhD Thesis, Chalmers
University of Technology, Göteborg, Sweden, 2008.
[17] P. N. Vovos, A. E. Kiprakis, A. R. Wallace and P.
Harrison, “Centralized and Distributed Voltage Control:
Impact on Distributed Generation Penetration”, IEEE
Transactions on Power Systems, Vol. 22, No.1, Feb. 2007,
pp. 476-483
[18] J. Kennedy and M. Clerc, "Standard PSO 2006,"
2006.http://www.particleswarm.info/Programs.html
[19] Biruk Tessema and Gary G. Yen, “A Self Adaptive Pe-
nalty Function Based Algorithm for Constrained Optimi-
zation,” IEEE Congress on Evolutionary Computation,
Oklahoma State University, Stillwater, 2006, pp 246-253.