Sie sind auf Seite 1von 9

1226

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 3, SEPTEMBER 2012

Voltage/VAR Control in Distribution Networks


via Reactive Power Injection Through
Distributed Generators
Siddharth Deshmukh, Student Member, IEEE, Balasubramaniam Natarajan, Senior Member, IEEE, and
Anil Pahwa, Fellow, IEEE

AbstractThis paper demonstrates how reactive power injection from distributed generators can be used to mitigate the
voltage/VAR control problem of a distribution network. Firstly,
power flow equations are formulated with arbitrarily located distributed generators in the network. Since reactive power injection
is limited by economic viability and power electronics interface,
we formulate voltage/VAR control as a constrained optimization
problem. The formulation aims to minimize the combined reactive
power injection by distributed generators, with constraints on: 1)
power flow equations; 2) voltage regulation; 3) phase imbalance
correction; and 4) maximum and minimum reactive power injection. The formulation is a nonconvex problem thereby making the
search for an optimal solution extremely complex. So, a suboptimal approach is proposed based on methods of sequential convex
programming (SCP). Comparing our suboptimal approach with
the optimal solution obtained from branch and bound method,
we show the trade-off in quality of our solution with runtime. We
also validate our approach on the IEEE 123 node test feeder and
illustrate the efficacy of using distributed generators as distributed
reactive power resource.
Index TermsConvex optimization, distributed generation, distribution network, sequential convex programming, voltage/VAR
control.

I. INTRODUCTION

ECENTLY, there is growing interest in distributed generation at or near the point of power consumption [1]. Distributed generators (DGs) feeding power at the distribution network level improve network reliability and reduce overall energy loss. Additionally, DGs enable operators to increase their
power supply capacity within the existing infrastructure [1], [2].
Integration of DGs in a distribution system poses many challenges in terms of: 1) power quality; 2) voltage regulation; 3)
protection; 4) reliability; and safety issues [1][5]. However, a
well controlled integrated operation of DGs with the main grid
can not only meet the challenges but can contribute ancillary
services like voltage/VAR support [6], [7]. Motivated by this
idea, recent research has focused primarily on two aspects of
DGs contribution to voltage control: 1) effective interfacing

Manuscript received July 19, 2011; revised December 12, 2011; accepted
March 21, 2012. Date of publication June 08, 2012; date of current version
August 20, 2012. This work was funded by Department of Energy grant #:
DE-EEC0000555. Paper no. TSG-00254-2011.
The authors are with the Department of Electrical and Computer Engineering,
Kansas State University, Kansas, KS 66506 USA (e-mail: sidd@ksu.edu;
bala@ksu.edu; pahwa@ksu.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2012.2196528

of DG at point of coupling (PCC), and 2) managing and optimizing multiple DGs power contribution. In [8] and [9], power
electronics interface at PCC is used to implement a PI feedback
control for regulating the local voltage. [10] and [11] extend the
idea of local voltage control in presence of multiple DGs. Another approach to regulate local voltage is shown in [12] and
[13], where active and reactive power of DG is controlled by
power electronics interface at PCC. Even though, we can not
neglect the importance of efficient interface at PCC, these prior
efforts consider voltage regulation only at PCC and not across
the entire distribution network.
Another thrust area of research relates to the centralized
and distributed control of DGs to regulate distribution network voltage. The author in [14] compares centralized and
distributed approaches for regulating the distribution network
voltage by controlling DG capacity. In [15], it is shown that at a
particular instant, either voltage or power factor of network can
be regulated. Hence a method of selective switching between
power factor and voltage control is proposed with maximum
utilization for distributed generation resources. To incorporate
stochastic nature in distributed generation and time variation in
load, [16] employs probabilistic network configuration model
in finding out the effect of DG penetration on voltage regulation. A similar approach is taken in [17] where Monte Carlo
simulations are performed on various case studies to determine
the effect of DGs on voltage regulation in low voltage grids.
Considering DGs as ad hoc infrastructure for quick voltage
support, especially in emergency situations, [18] proposes a
multiagent based dispatching scheme for communication between DGs. Considering DGs presence in distribution network,
[19] formulates an optimization problem with the objective of
minimizing power losses. A genetic algorithm is presented in
[19] for controlling the taps of load tap changer (LTC), size of
substation capacitor, and voltage amplitudes of DG.
While many of the DGs are currently assumed to inject
only real power, advances in power electronics and need for
voltage/VAR support has motivated us to consider DGs as
distributive reactive power resource. In [20], a multiobjective voltage/VAR control problem is formulated assuming
DGs presence, and a Nondominated Sorting Genetic Algorithm (NSGA) is proposed to solve the optimization problem.
Similarly, authors in [21] propose a hybrid algorithm based
on the Ant Colony genetic algorithm for solving nonlinear
voltage/VAR control problem. A comprehensive comparison
of stochastic search methods is discussed in [22] for optimizing
daily voltage/VAR control problem. Even though variants of

1949-3053/$31.00 2012 IEEE

DESHMUKH et al.: VOLTAGE/VAR CONTROL IN DISTRIBUTION NETWORKS VIA REACTIVE POWER INJECTION THROUGH DISTRIBUTED GENERATORS

stochastic search methods have been proposed, optimality of


obtained solution can never be guaranteed [23], [24]. Furthermore, an exact stopping criterion is hard to obtain in such
methods, making actual run time calculations nondeterministic.
Unlike stochastic search methods, interior point based methods
[25], such as SCP [26] have been shown to scale effectively
and in a deterministic manner with problem size. Deterministic
runtime and guarantees in optimality are important for real time
voltage/VAR control. The authors in [27] and [28] present a
real time control framework for controlling of end-user reactive
power devices to mitigate low voltage problems at transmission
level. Steepest distance method is used in [27], and Newtons
method is used in [28] to calculate the optimal reactive power
injection. However, both these solutions are local in nature
[25].
Considering unpredictability and local optimality in search
methods, our paper aims to demonstrate a deterministic runtime
and close to optimal, optimization technique for voltage/VAR
support. For this, we specifically consider distribution networks
that consist of multiple low capacity DGs. For example, this
model accommodates the scenario with residential generators
such as small wind turbines, solar panels, etc. We first formulate the power flow equation that capture the impact of reactive
power injection by the DGs across the entire distribution network. Then, we treat reactive power as a vital resource not only
for voltage regulation, but also for phase imbalance correction.
As reactive power injection has a natural trade-off relative to
real power injection by DGs, it becomes critical to optimally use
the reactive power of DGs. We accomplish this by formulating
an optimization problem where the objective is to minimize aggregate reactive power injected at various PCCs, with basic constraints on: 1) voltage regulation across the nodes of network,
and 2) phase imbalance correction. We also place constraints on
minimum and maximum reactive power a DG can inject based
on economic viability and limitations of power electronics interface. Thermal limits on distribution lines are not considered in
our formulation. In most distribution networks with exception
of those in dense urban areas, the lines are loaded much below
their thermal capacity [29]. Hence, thermal limits can be excluded from the constraints. However, if needed it can be always
included in the problem formulation. The resulting optimization is nonconvex, making the quest for global optimal solution
very complex. Fig. 1 shows different approaches which can be
adopted to solve this problem. One of the direct approaches is
branch and bound (BB) method. The BB method gives a global
optimal solution but with disadvantage of exponential runtime
[30], [31]. Another direct approach is to directly apply interior
point method on nonconvex problem. This approach has polynomial run time but it only gives a feasible solution. In our approach, we transform the original problem to a convex problem
and use sequential convex programming (SCP) to determine the
solution. This is a suboptimal approach with polynomial runtime. We demonstrate the quality of our suboptimal solution
by comparing it with the global optimum obtained via branch
and bound method. Finally, using the IEEE 123 node test feeder
[32], we validate our approach. Also, it can be observed that
usage of DGs decreases the dependency on voltage regulator
and capacitor banks for voltage/VAR control. Our results and

1227

Fig. 1. Different optimal solution methods.

analysis illustrate that it is not only feasible but prudent to use


DGs to provide voltage/VAR support in distribution networks.
This paper is organized as follows: Section II formulates the
power flow equation with DGs located arbitrarily at various
nodes. In Section III an optimization problem is formulated to
calculate net reactive power injection required to meet the distribution network constraints. Section IV presents the results and
analysis of simulation done for radial network and IEEE 123
node standard test feeder case. Finally, conclusions and possible
future work are presented in Section V.
II. THREE PHASE DISTRIBUTION SYSTEM MODEL
A three phase distribution network can be modeled as radial interconnection of mathematical equivalents of the corresponding network components as given in [33]. For mathematical consistency, we assume all loads are three phase star connected loads. Similarly, all the generators are considered to be
star-connected with capability to generate power in each phase.
Although we have considered loads and generators to be three
phase star connected, this configuration allows us to accommodate single and two phase loads and generators by assuming the
loads and generators on unconnected phases to be zero.
In practical distribution networks, there is mixture of constant impedance, constant current and constant power loads. We
have considered all loads to be constant power loads in our analysis. This assumption of constant power loads results in a nonlinear mathematical model. Other forms of load models can be
easily incorporated by introducing additional linear terms in the
formulation.
Notation
We use normal-faces to define scalars and bold-faces to define matrices and vectors;
denotes element wise complex
conjugate operation of vector/matrix ;
denotes element
wise absolute value of vector/matrix ;
denotes element wise product of two vectors/matrices , . Superscript
at element
indicates that element
corresponds to phase
where
. Similarly
indicates a composite

1228

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 3, SEPTEMBER 2012

Fig. 2. Three phase radial distribution network with DG.

vector of elements corresponding to phase , , . Subscript


and
indicate generator and load at node .
A. Power Flow Analysis
We consider a radial distribution model as shown in Fig. 2,
where at every node there is a generation source and load. We assume that the load
represent the apparent power of lumped
load of the lateral connected at node . Similarly,
represents
the apparent power generated at the
lateral. At any node ,
the absence of a generator or load is captured by setting the corresponding term
or
to .
represents the ground
voltage, and is considered as reference for the system.
represents the neutral voltage at node . Considering voltage vector
at node with ground as reference, voltage at
node
can be represented as

Here,
is equivalent three phase impedance matrix between node and
. Defining equivalent admittance matrix
, and applying Kirchoffs current law,
current entering from a node into the network can be expressed
as

(6)
In matrix form, the overall voltage current relation for the entire
network modeled in Fig. 2 can be represented as

(7)

(1)
where
is the three phase current
flowing from node to node
, and the impedance matrix is
three phase impedance matrix between node and node
.
In block matrix form, (1) can be represented as
(2)
Typically, neutral and ground are connected, and hence they are
at the same potential, i.e.,
. Equating
and
in second row of (2), we get
(3)
from (3) into first row of (2), we get correSubstituting
sponding Krons reduction form
(4)

where

element of

is defined as

Next, the power injected at node can be expressed as


. Substituting the value of
from (7), we can
rewrite
as
(8)
is
row of
matrix. Representing the equivwhere
alent admittance matrix in real and imaginary part,
and three phase voltage in polar form
and substituting in (8) gives

where
(5)

(9)

DESHMUKH et al.: VOLTAGE/VAR CONTROL IN DISTRIBUTION NETWORKS VIA REACTIVE POWER INJECTION THROUGH DISTRIBUTED GENERATORS

Simplifying (9) and representing in real and reactive power


form,

(10)
In steady state, net power at any node is difference between
power generated and consumed at that node. Therefore,
(11)
Here,
is the three phase apparent power generated at node
;
is three phase apparent load at node ;
and
are
the real and reactive power generated at node ; and
and
are the real and reactive components of load at node .
Equating (10) and (11), the power flow equations that capture
the impact of reactive power injection from DGs in distribution
networks can be expressed as

1229

to determine the optimal reactive power injection by DGs desired to satisfy the following requirements: 1) voltage is maintained within safety limits; 2) power flow (12) are satisfied; 3)
individual DGs minimum/maximum reactive power constraints
are met; and 4) phase imbalance is mitigated. For phase imbalance correction, we assume that the phase angle varies from 0 to
, and we limit the phase difference between any two phases
at a node to be greater than
, where
is the tolerance of phase imbalance.
We formulate the optimization problem by assuming that the
net generation capacity, i.e.,
of individual DGs is fixed,
known or predictable. Future work will include stochastic
models for distributed generation capability. So, presently we
have constraints on vector sum of real and reactive power
generated by each DG. When we do not have sufficient distributed generation, we assume that the main grid can provide
the needed capacity without any limit. We also assume that real
and reactive component of load, i.e.,
,
at each node
is known.
The optimization problem can be expressed as

subject to

(12)
Equation (12) can be also obtained by following alternate
modeling approaches as discussed in [34], [35]. In the next section, an optimization problem, minimizing reactive power injection is formulated for voltage/VAR control in distribution
network.
The power flow equations in (12) constitute an essential constraint in quantifying the effect of individual DGs on entire distribution system.
III. VOLTAGE/VAR CONTROL OPTIMIZATION FORMULATION
Distributed generators connected to distribution network can
be used to provide reactive power as ancillary service. However, low power generators are mostly owned by residential
customers, and are paid only for the real power they inject in
the network. Additionally, these small DGs may have some
maximum and minimum value of reactive power injection constraints based on the power electronics interface and economic
viability. In this section, we formulate an optimization problem

(13)
where, in objective function
,
are the regressors indicating preference of generators in reactive power contribution;
choice of function
indicate the dislike or penalty for the
increase in reactive power contribution; equality constraints,
and
represent the power flow (12);
is limit on net
power generation capacity of individual generators; inequality
constraint,
represents 5% voltage regulation;
represents
tolerance on phase imbalance, and
captures the limits on

1230

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 3, SEPTEMBER 2012

maximum and minimum reactive power injected by individual


DGs.
It can be observed that, because of constraints , , , and
, (13) is a nonconvex optimization problem. This problem
can be solved using branch and bound method that guarantees
optimal solution. However, the prohibitive complexity associated with branch and bound method makes it impractical for
real time voltage/VAR control application. Alternatively, one
may choose to use suboptimal stochastic search methods based
on evolutionary/genetic algorithms. These methods also suffer
from high complexity, poor repeatability, with no certificate of
solution quality. Therefore we attempt to transform our nonconvex optimization problem into convex form, and compare
the quality of suboptimal solution with global optimal solution.
This approach is of low complexity, fast, and is best suited for
problems where suboptimal solution close to true optimal solution is acceptable.

and
and

are the Jacobian of power flow equation constraints


, defined as
(16)

where

and
(17)
where

A. Convex Form Analysis


The objective function in above optimization problem
(13) is formulated as a regression problem formulation.
In this work, we assume a quadratic penalty function, i.e.,
. This choice of
assures that the objective function is convex. Additionally
is symmetric so that large positive and negative reactive power
injectors are equally undesirable. Also assuming fair policy
for all the generators, we set all regressors
to 1. In standard
convex optimization problem, the inequality constraints are
convex function and the equality constraints are affine. However, equality constraints in our formulation, especially the
power flow constraints
and
are highly nonlinear. Also,
the phase imbalance correction constraint
in inequality
constraint is nonconvex. Therefore, we apply the method of
first order approximation on
and
to get affine equality
constraints.
The first step in our reformulation is to find a feasible solution

(14)

and
.
where
To obtain a feasible solution we apply the interior point algorithm on the original problem. This gives a local optimal point.
The next step is to obtain affine approximation of nonlinear
equality constraints, via Taylor series expansion. The first order
Taylors series approximation of power flow equation can be expressed as

(15)
where
;
is the trust region with radius
feasible point , defined as

around the

Similarly, the affine approximation of equality constraint


on net generation capacity of individual generators can be
expressed as
(18)
where

is the
The three phase imbalance correction constraint
only nonconvex inequality constraint in our optimization for, the differmulation. If we maintain a small trust region
ence between the three phases will not change sign. Therefore,
the maximum and minimum phase angle will remain the same
during an iteration. Thus, if
in our feasible solution,
can be restated as

(19)
Finally, the convex transformed form of the original optimization problem corresponds to

subject to Constraint

, and

of (13), and

DESHMUKH et al.: VOLTAGE/VAR CONTROL IN DISTRIBUTION NETWORKS VIA REACTIVE POWER INJECTION THROUGH DISTRIBUTED GENERATORS

1231

TABLE I
VOLTAGE (IN PU) AND PHASE (IN DEGREES) PROFILE

(20)
The next step in SCP is to solve (20) based on the basic feasible solution (14). The new solution is plugged in , , ,
and
constraints to get a new instance of (20), and is solved
again. Thus, (20) is solved iteratively till the solution converges
to an optimal point. In every iteration a new feasible solution
is obtained which is better than previous solution.Solving the
reformulated problem (20) by SCP is fast but finding a global
optimal solution to original problem is not guaranteed. In next
section, we show the complexity and convergence analysis of
our approach and compare it with global solver based on branch
and bound method.

TABLE II
REACTIVE POWER INJECTION PROFILE (IN KVAR)

IV. RESULT ANALYSIS


In this section, we show the effectiveness of our approach by
analyzing a small-scale radial distribution network and a modified IEEE 123-node test feeder system.
A. Radial Distribution Network
First, we consider a radial distribution case shown in Fig. 2
with the following setup.
1) Simulation Setup:
System base values: 4.16 kV and 100 kVA.
Number of nodes: 11 nodes (including the grid connecting
node 0).
Load profile: Constant power star connected spot load.
,
,
.
Generation profile: Three phase DGs with equal generation
capacity in all phases.
,
. We assume that there is no constraint on generation capability at the grid, both in terms of active and
reactive power. Thus our optimization problem takes grid
power as a variable and we get the optimal power drawn
from grid as a part of our solution.
Limit on reactive power injected by individual DGs:
(we have assumed 100% reactive power convertibility; however, it can be less as simulated in
Section IV-A4.)
Line impedance: Constant line impedance between all
nodes:
, as defined in (2). Further we assume that all
nodes are 1000 feet apart. The value of
(2) per mile is

Phase imbalance tolerance:


, i.e., phase difference
between any two phases is between 115 to 125 .
2) Simulation Result: Normally, in a distribution network,
voltage/VAR is controlled by use of voltage regulators and capacitor banks. In this paper, we claim that optimal injection of
reactive power by distributed generators can support network
constraints. To validate our claim, we first simulate the above
setup, without any distributed generation. We observe that the
problem is infeasible, i.e., voltage/VAR needs to be supported
by voltage regulator/capacitor banks. So, in the next step we
simulate the system with all the distributed generators set according to a set generation profile.
Tables I and II shows the voltage, phase and reactive power
injection profile obtained from our suboptimal approach.
Voltage profile indicates operation of radial network within
the 5% regulatory safety limit. Phase profile indicates phase
imbalance correction within tolerance of . It can be observed
from reactive power injection profile (unconstrained
),
that the grid acts as the major source of reactive power, with
small generators injecting minimum reactive power. The reactive power injected by distributed generators increases as we
move away from the grid. This again confirms that grid itself
can not support voltage/VAR control for distant nodes. The
objective function on the optimization problem (20) aims to
fairly distribute reactive power injection among all distributed
generators. This can be inferred from the reactive power injection by nodes closer to the grid. While constraints at these
nodes can be satisfied by the grid itself, some reactive power
continues to be injected by these nodes. This results in low
reactive power injection from generators at distant nodes. To
further illustrate the effectiveness of our approach, we constrain
the reactive power injection at node 10 to 8 kVAR or 0.8 pu.

1232

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 3, SEPTEMBER 2012

Table II also shows reactive power injection profile, obtained


after constraining reactive power injection on generator at node
10,
. To
compensate for the insufficient reactive power at node 10, it
can be observed that our approach increases the reactive power
injection on nodes 0 to 9. In next subsection, we analyze the
computational complexity of our approach and compare it with
global solver based on branch and bound algorithm.
3) Complexity Analysis: In this subsection, we analyze and
compare the computational complexity and runtime of our approach with global solver based on branch and bound method.
The branch and bound method is a nonheuristic approach, and
it certifies the quality of its solution to be -optimal [30]. The
computational complexity of branch and bound algorithm depends on: 1) complexity of computing the lower and the upper
bound functions; 2) number of rectangles to partition; and 3)
edge (variables) to partition. In the worst case if we assume,
that each side of rectangle has to be split in parts, and there
are such variables then complexity is
. For a problem with
4 nodes,
(3 voltage, 3 phase, 3 real power, and 3 reactive
power variables for each node) variables, the worst case complexity is
. Since variables in our case are continuous,
can have large value, and thus worst case runtime is relatively
very large for practical distribution networks.
The complexity for our approach depends on: 1) complexity
of interior point method; 2) complexity of SCP iteration; and 3)
number of SCP iterations. Interior point method solves the optimization problem by applying Newtons algorithm on sequence
of equality constrained problems. The worst case complexity
is
, where is the number of variables and
is the
number of constraints [25]. For our approach we use interior
point method first to get feasible point. Further it takes
to compute the affine approximation of nonconvex constraints.
Thus in our approach SCP iterations are effective to the order
of
. For problem with 4 nodes,
, and
(6 constraints per phase per node), the worst case complexity
is
. Thus our approach has far less computational
complexity compared to exponential time complexity of branch
and bound method.
4) Optimal Solution Comparison: In this subsection we compare the SCP based suboptimal solution with branch and bound
(BB) based global optimal solution. The simulation setup is similar to Section IV.A with three modifications. Firstly, to emphasize on reactive power drawn from DGs, we have increased
the distance between adjacent nodes. Following are the internodal distances in 1000 ft unit:
;
;
;
;
;
;
;
;
;
. Secondly, we reduce the
number of nodes with DGs with their increased generation capability, i.e.,
,
.
Finally, we include a constraint on power electronic interface
in converting total power to reactive power. That is, we assume that maximum of 60% of total generated power can be injected as reactive power by individual DGs. Table III shows the
three phase reactive power injection profile obtained from both
SCP based approach and BB method. The SCP based solution
requires, 0.006% (1892.32 kVAR) more reactive power compared to global minimum (1892.21 kVAR) obtained from BB

TABLE III
REACTIVE POWER INJECTION (IN KVAR) PROFILE (SCP VRS. BB)

Fig. 3. IEEE 123 node test feeder with distributed generators.

methods. Thus, with very small compromise in optimality, we


achieve significantly lower computational complexity and run
time. It can be observed that both the approaches, extract maximum reactive power (60% of total generation capacity) from
DGs. In the next subsection, we consider the IEEE 123 node
test feeder, as a test case to evaluate our approach on large scale
distribution network.
B. Simulation With Standard Test CaseIEEE 123 Node Test
Feeder
Fig. 3 shows a IEEE 123 node modified test node feeder with
arbitrarily located low power distributed generators. The following are the modification to the standard test case:
Node 149 is connected to grid, and nodes represented by
are with generation capability. Generation capacity is
arbitrarily single, two or three phase, each of 10 kVA as
shown in Fig. 3.
For simplicity in simulation, we do not consider shunt capacitors, voltage regulators, and transformers in the network; however, our approach is applicable to distribution
networks including mathematical models of these components.
Three phase switches are set according to [32].

DESHMUKH et al.: VOLTAGE/VAR CONTROL IN DISTRIBUTION NETWORKS VIA REACTIVE POWER INJECTION THROUGH DISTRIBUTED GENERATORS

TABLE IV
REACTIVE POWER INJECTION PROFILE (IN KVAR)

1233

losses is not considered while optimizing the control problem.


However, problem can be reformulated to include losses as an
additional objective. In our future work, we aim to address the
voltage/VAR control using stochastically varying loads and distributed generation. We will also include the impact of communication infrastructure on our control framework.
REFERENCES

All three phase loads are assumed to be constant power star


connected spot loads.
Our simulation setup is as same as subsection A, with additional constraint on reactive power injected by grid,
,
and
. Table IV
shows distributed reactive power injection profile required to
satisfy the network constraints (20). Once again our solution
considers grid as major source of reactive power. The generation profile is mixed with one, two and three phase generators
arbitrarily located in the network. As before, the solution results
in fair distribution of reactive power injection among all DGs.
V. CONCLUSION
In this paper, we present the efficacy of using distributed generators as a reactive power resource. Reactive power injection
by distributed generators is constrained by economic viability
and power electronic interface. So, it is important to optimally
utilize reactive power injected by distributed generators. An optimization problem is formulated with the objective to minimize injected reactive power, while satisfying constraints related to: 1) voltage regulation; 2) phase imbalance correction;
and 3) power flow. Thus distributed generators in our work
are treated at network level to address the voltage/VAR control. The optimization problem being nonconvex, we propose a
suboptimal approach based on sequential convex programming
(SCP). The proposed approach provides a near optimal solution
with much lower computation complexity (runtime) relative to
a global solver based on branch and bound method. Further we
have shown the practicality of our approach by simulating a
123 node IEEE test feeder. Since this paper focuses on reactive power minimization, the effect of voltage/VAR on system

[1] P. P. Barker and R. W. De-Mello, Determining the impact of distributed generation on power systems: Part 1Radial distribution systems, in Proc. Power Eng. Soc. Summer Meet., Jul. 2000, vol. 3, pp.
16451656.
[2] G. Pepermansa, J. Driesenb, D. Haeseldonckxc, R. Belmansc, and W.
Dhaeseleer, Distributed generation: Definition, benefits and issues,
Energy Policy, vol. 33, pp. 787798, 2005.
[3] F. Bastiao, P. Cruz, and R. Fiteiro, Impact of distributed generation
on distribution networks, in Proc. 5th Int. Conf. Eur. Electr. Market
(EEM), May 2008, pp. 16.
[4] T. Khoan and M. Vaziri, Effects of dispersed generation (DG) on distribution systems, in Proc. IEEE Power Eng. Soc. Gen. Meet., Jun.
2005, vol. 3, pp. 21732178.
[5] J. Driesen and R. Belmans, Distributed generation: Challenges and
possible solutions, in Proc. IEEE Power Eng. Soc. Gen. Meet., Oct.
2006.
[6] G. Joos, B. T. Ooi, D. McGillis, F. D. Galiana, and R. Marceau,
The potential of distributed generation to provide ancillary services,
in Proc. IEEE Power Eng. Soc. Summer Meet., 2000, vol. 3, pp.
17621767.
[7] F. Li, J. D. Kueck, D. T. Rizy, and T. King, A Preliminary analysis
of the economics of using distributed energy as a source of reactive
power supply, Oak Ridge National Laboratory (ORNL), Oak Ridge,
TN, Tech. Rep. (ORNL/TM-2006/014), 2006.
[8] H. Li, F. Li, Y. Xu, D. Rizy, and J. Kueck, Dynamic voltage regulation
using distributed energy resources, in Proc. 19th Int. Conf. Electr.
Distrib. (CIRED), Vienna, Austria, May 2007.
[9] M. Triggianese, J. Morren, S. de Haan, and P. Marino, Improved and
extended DG capability in voltage regulation by reactive and active
power, in Proc. Int. Conf. Power Eng., Energy, Electr. Drives (POWERENG), Apr. 2007, pp. 583588.
[10] H. Li, F. Li, Y. Xu, D. Rizy, and J. Kueck, Interaction of multiple distributed energy resources in voltage regulation, in Proc. Power Energy Soc. Gen. Meet.Convers. Del. Electr. Energy 21st Century, Jul.
2008, pp. 16.
[11] L. Huijuan, L. Fangxing, X. Yan, D. Rizy, and J. Kueck, Adaptive
voltage control with distributed energy resources: Algorithm, theoretical analysis, simulation, and field test verification, IEEE Trans.
Power Syst., vol. 25, pp. 16381647, Aug. 2010.
[12] J. Morren, S. de Haan, and J. Ferreira, Distributed generation units
contributing to voltage control in distribution networks, in Proc. 39th
Int. Univ. Power Eng. Conf. (UPEC), 2004, vol. 2, pp. 789793.
[13] D. Geibel, T. Degner, C. Hardt, M. Antchev, and A. Krusteva, Improvement of power quality and reliability with multifunctional PV-inverters in distributed energy systems, in Proc. 10th Int. Conf. Electr.
Power Quality Utilisation (EPQU), Sep. 2009, vol. 25, pp. 16.
[14] P. Vovos, A. Kiprakis, A. Wallace, and G. Harrison, Centralized and
distributed voltage control: Impact on distributed generation penetration, IEEE Trans. Power Syst., vol. 22, pp. 476483, Feb. 2007.
[15] D. N. Gaonkar, P. Rao, and R. Patel, Hybrid method for voltage regulation of distribution system with maximum utilization of connected
distributed generation source, in Proc. IEEE Power India Conf., Jun.
2006, pp. 510.
[16] S. Chun-Lien, Stochastic evaluation of voltages in distribution networks with distributed generation using detailed distribution operation
models, IEEE Trans. Power Syst., vol. 25, pp. 786795, May 2010.
[17] F. Demailly, O. Ninet, and A. Even, Numerical tools and models
for Monte Carlo studies of the influence on embedded generation on
voltage limits in LV grids, IEEE Trans. Power Del., vol. 20, pp.
23432350, Jul. 2005.
[18] M. Baran and I. El-Markabi, A multiagent-based dispatching scheme
for distributed generators for voltage support on distribution feeders,
IEEE Trans. Power Syst., vol. 22, pp. 5259, Feb. 2007.
[19] T. Niknam, A. Ranjbar, and A. Shirani, Impact of distributed generation on volt/VAR control in distribution networks, in IEEE Bologna
Power Tech Conf. Proc., Jul. 2003, pp. 714.

1234

[20] P. Hrisheekesha and J. Sharma, Evolutionary algorithm based optimal control distribution system dispersed generation, Int. J. Comput.
Appl., vol. 14, pp. 3137, Feb. 2010.
[21] T. Niknam, A new approach based on ant colony optimization for
daily volt/var control in distribution networks considering distributed
generators, Int. J. Comput. Appl., vol. 49, pp. 34173424, 2008.
[22] J. Olamaie and T. Niknam, Daily volt/var control in distribution networks with regard to DGs: A comparison of evolutionary methods, in
Proc. IEEE Power India Conf., 2006, p. 6.
[23] S. K. Johannes, Stochastic Optimization. New York: Springer, 2007.
[24] J. C. Spall, Introduction to Stochastic Search and Optimization. Hoboken, NJ: Wiley, 2003.
[25] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge,
U.K.: Cambridge Univ. Press, 2004.
[26] S. Boyd, Sequential convex programming, Stanford Univ., Stanford,
CA, 2010 [Online]. Available: http://www.stanford.edu/class/ee364b/
lectures/seq_slides.pdf
[27] K. Rogers, R. Klump, H. Khurana, A. Aquino-Lugo, and T. Overbye,
An authenticated control framework for distributed voltage support
on the smart grid, IEEE Trans. Smart Grid, vol. 1, pp. 4047, Jun.
2010.
[28] A. Aquino-Lugo, R. Klump, and T. Overbye, A control framework
for the smart grid for voltage support using agent-based technologies,
IEEE Trans. Smart Grid, vol. 2, pp. 173180, Mar. 2011.
[29] H. L. Willis, Power Distribution Planing Reference Book. Boca
Raton, CA: CRC, 2004.
[30] R. Horst and T. Hoang, Global Optimization: Deterministic Approaches. New York: Springer, 1996.
[31] S. Boyd and J. Mattingley, Branch and bound methods, Stanford
University, Stanford, CA, 2007 [Online]. Available: http://www.stanford.edu/class/ee364b/notes/bb_notes.pdf
[32] W. H. Kersting, Radial distribution test FeedersDistribution system
analysis subcommittee report, in Proc. IEEE Power Eng. Soc. Summer
Meet., 2001.
[33] W. H. Kersting, Distributed System Modeling and Analysis. Boca
Raton, CA: CRC, 2002.
[34] F. D. Torre, A. Dolara, S. Leva, and A. Morando, Faults analysis
theory and schemes of four-phase power systems, Proc. IEEE PowerTech Bucharest, pp. 17, Jul. 2009.
[35] F. D. Torre, S. Leva, and A. Morando, Symmetrical components and
space-vector transformations for four-phase networks, IEEE Trans.
Power Del., vol. 23, no. 4, pp. 17, Jul. 2008.

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 3, SEPTEMBER 2012

Siddharth Deshmukh (M07S10) received the B.E. degree in electronics and


telecommunication discipline from the National Institute of Technology, Raipur,
India, in 2004 and the M.Tech. degree from the Indian Institute of Technology,
Delhi, in 2006. Since Fall 2010, he has been working toward the Ph.D. degree
at Kansas State University, Manhattan.
His research interest includes network control and optimization, statistical
signal processing, and communication theory.

Balasubramaniam Natrajan (S98M02SM08) received the B.E degree in


electrical engineering from Birla Institute of Technology and Science, Pilani,
India, in 1997 and the Ph.D. in electrical degree from Colorado State University,
Fort Collins, in 2002.
Since Fall 2002, he has been a Faculty Member in the Department of Electrical and Computer Engineering, Kansas State University, Manhattan, where
he is currently an Associate Professor and the Director of the Wireless Communication (WiCom) and Information Processing Research Group. He was also
involved in telecommunications research at Daimler Benz Research Center,
Bangalore, India, in 1997. He has published a book titled Multi-carrier Technologies for Wireless Communications (Kluwer, 2002) and holds a patent on
customized spreading sequence design algorithm for CDMA systems. His research interests include spread spectrum communications, multicarrier CDMA
and OFDM, multiuser detection, cognitive radio networks, sensor signal processing, distributed detection, and estimation and antenna array processing.

Anil Pahwa (F03) received the B.E. (honors) degree in electrical engineering
from Birla Institute of Technology and Science, Pilani, India, in 1975, the M.S.
degree in electrical engineering from University of Maine, Orono, in 1979, and
the Ph.D. degree in electrical engineering from Texas A&M University, College
Station, in 1983.
Since 1983, he has been with Kansas State University, Manhattan, where
presently he is a Professor in the Electrical and Computer Engineering Department. He worked at ABB-ETI, Raleigh, NC, during sabbatical from August
1999 to August 2000. His research interests include distribution automation,
distribution system planning and analysis, distribution system reliability, and
intelligent computational methods for distribution system applications.
Dr. Pahwa is a member of Eta Kappa Nu, Tau Beta Pi, and ASEE.

Das könnte Ihnen auch gefallen