See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224056855

Optimal Phase Balancing in Distribution System Using Mixed-Integer Linear

Programming

Conference Paper · September 2006

DOI: 10.1109/TDCLA.2006.311368 · Source: IEEE Xplore

CITATIONS READS
24 1,965

4 authors, including:

Paulo M. De Oliveira-De Jesus Manuel António Matos

Los Andes University (Colombia) Institute for Systems and Computer Engineering, Technology and Science (INESC TE…
62 PUBLICATIONS 1,059 CITATIONS 175 PUBLICATIONS 5,583 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Paulo M. De Oliveira-De Jesus on 03 June 2014.

The user has requested enhancement of the downloaded file.

 1
2006 IEEE PES Transmission and Distribution Conference and Exposition Latin America, Venezuela

Optimal Phase Balancing in Distribution System

Using Mixed-Integer Linear Programming
H. M. Khodr, Member, IEEE, I. J. Zerpa, P. M. De Oliveira-De Jesús, Member, IEEE,
and Manuel A. Matos, Member, IEEE


Deregulation eliminates the boundary of the territory of the
situations since they can trip protection devices reducing the monopoly power industry. Competition forces utilities to im-
efficiency and damaging some appliances. Therefore, in order to prove power quality as well as to reduce investment and op-
achieve the balance of the networks it is necessary an improve-
eration costs. Feeder imbalance describes a situation in which
ment of the electric service leading to a reduction of the network
costs. The proposed model performs the optimal balancing of the voltages of a three-phase voltage source are not identical
electric networks by swapping loads among system phases. It is in magnitude, or the phase differences between them are not
assumed that all the loads have the same power factor and volt- 120 electrical degrees, or both. It affects motors and other
age drops are not considered. As a fundamental contribution of devices that depend upon a well-balanced three-phase voltage
this work, an additional restriction was added in the mathemati- source. Phase balancing is to make the voltages balanced at
cal model in order to avoid the introduction of additional phases.
each load point of the feeder. Phase swapping is a direct ap-
Index Terms
proach for phase balancing with the minimum cost. Phase
balancing can enhance utilities' competitive capability by im-
Index Terms--Distribution systems, phase balancing, phase swa- proving reliability, quality, and reducing costs. Therefore,
pping, unbalance feeders phase balancing optimization is nowadays receiving more
attention in the power industry, especially in today's deregu-
lating environments. The nonlinear effects, such as, voltage
I. NOMENCLATURE drops and energy losses, make the problem hard to solve. [1]
KCL: Kirchoff’s Current Law. Two options exist for balancing of three phases electrical
LINDO: Linear Interactive and Discrete Optimizer. systems. One is feeder’s reconfiguration at system level; and
MIP: Mixed – Integer Programming. the other one is phase swapping at the feeder level [2]. The
first one has as purpose to balance loads among feeders, there-
II. INTRODUCTION fore is not an effective technique to settle the unbalance prob-
he three phase electric systems need to maintain a bal- lem. Under normal operating conditions, the network is re-
T anced state of the in order to avoid problems that charac- configured to reduce the system's losses and/or to balance load
in the feeders. Under conditions of permanent failure, the net-
terize the unbalanced electric systems, as: unbalanced level of
voltages, overload of phases, protection devices trip, increase work is reconfigured to restore the service, minimizing the
of losses and decrease of the efficiency of appliances. zones without power. The network reconfiguration consists on
The increase of energy losses are one of the main problems modifying the topology of the system by switching remotely
that affect the electric service; its weight on the purchases of controlled sectionalizing switches. In this process, the nodes
energy and its effects on the quality, security and reliability of can be energized through different paths through the intercon-
the service are indicators to be considered, if the operating nection with other feeders (substations) and/or interconnection
efficiency and power quality of the service are the standard for of nodes belonging to a same feeder. Usually, distribution
the business. By many years, some efforts have been dedi- systems operate with a radial topological structure; conse-
cated and resources on behalf of its solution, and in spite of quently; the opening and closing of sectionalizing switches
these actions, the situation continues present and the tendency must be made considering this constraint. The problem con-
is toward seeking novel solutions at low costs. sists on determining an ordered switching list that allows to
reduce losses or to balance the load of the system without in-
fringing operational and topological constraints [3].
This work has been supported in part by Fundação para a Ciência e a Tec-
nologia Grant SFRH/ BPD/ 21035/ 20041 Phase swapping is an effective way to balancing a feeder in
H. M. Khodr is with INESC Porto on leave of Energy Conversion and De- terms of its phases, which consists of change the connection
livery Department, Universidad Simón Bolívar, Aptdo.89000, Caracas, Vene- of the loads or lateral branches among the phases of the line.
zuela (e-mail: [email protected]).
I. J. Zerpa. is with Ministry of Energy and Petroleum of Venezuela, Cara- Carrying out the appropriate swapping can be reduced the
cas, Venezuela (e-mail: [email protected]). unbalance level of the network, since the main cause of unbal-
P. M. De Oliveira-De Jesús andM. A. Matos is with the Faculty of Engi- ances in an electric network are the single-phase and two-
neering of the University of Porto and INESC Porto, Campus da FEUP, Porto
4200-465, Portugal (e-mail: [email protected]).
phase loads connected to a three-phase network.
1-4244-0288-3/06/$20.00 ©2006 IEEE
 2

The algorithm here proposed is based on [2] which are Each branch of the feeder is associated with a balance in-
characterized for reducing to fictitious nodes the lateral of the dex, for example, unbalanced flow, which is defined as the
main lines, so that to obtain the data of load of the lateral maximum difference of flows at this branch. The lower the
nodes, these should be added to carry out the reduction in the unbalanced flow is, the more balanced the branch is. [2]
loading data process to the program. This represents a con-
straint for balancing, for which here is proposed to carry out IV. MATHEMATICAL MODEL
the necessary changes to avoid this. These changes are: not to Considering a radial feeder as the one that is shown in Fig.
reduce the lateral nodes or fictitious nodes, but to introduce 1, with loads di,1, di,2, di,3 at node i, unbalance in the branch j,
the individual loads of each node and each branch; and the Uj, is defined as the maximum difference of phase current that
introduction of an additional constraint to the mathematical flow in the branch j.
model that avoids the introduction of additional phases in the
lateral branches. A tool has been programmed for balancing 1 2 3 4 5
electrical networks.
This program proposes the exchange of loads among 1 2 3 4 5
a I1,a I5,a
phases of the lateral of the main lines, to assure that the loads
average among the phases of the main lines differ not in large b I1,b I5,b
magnitudes, requesting the magnitudes of the currents that S
flow for each phase, and the location of the node in the net- c I1,c I5,c
work. To simplify the problem is assumed that all the loads
have the same power factor and voltage drops are not consid- d1,a , d1,b , d1,c d5,a , d5,b , d5,c
ered.
Fig. 1. An example of a radial feeder.
This proposal will contribute with savings of energy in the
system almost without investments, which will be a benefit for
In the Fig. 1, I j,ij , is the current of phase that circulates by
the business reducing its expenses by invoice of the service of
electricity, besides it will be more efficient and rational in the the branch j.
use of the energy resources complying thus with the tendency In the mathematical model integer decision variables will
of the present premises of the regulatory authorities on these be introduced to guarantee that one and only one load is as-
resources. signed to a phase. The model of MIP is stated as follow:

III. PHASE SWAPPING

Minimize Uj
In power distribution systems, single or two phase laterals
and unbalanced loading are common phenomena. Unbalanced Subject to:
feeders not only increase energy losses and the risk of over-
load situations, but they also affect system power quality and
electricity price. Even feeder system is designed as a balanced ^
max I j,a  I j,c , I j,b  I j,c , I j,a  I j,b ` Uj (1)
feeder based on the given load data, load prediction errors and
unbalanced load growth will induce feeder imbalance. The I j,ij ¦I k,ij  x iij,1 ˜ d i,1  x iij,2 ˜ d i,2  x iij,3 ˜ d i,3 (2)
purpose of phase balancing is to find the optimal phase swap-
ping scheme to balance an unbalanced feeder system with x iij,1  x iij,2  x iij,3 1 for all ij = a, b, c (3)
minimum cost.
Phase swapping balances system by retapping loads or lat- x ia,Ȧ  x ib,Ȧ  x ic,Ȧ 1 for all Ȧ= 1,2,3 (4)
erals to the phase lines. It requires each single-phase load 1 1 1
x iij,1  x iij,2  x iij,3  x iij,1  x iij,2  x iij,3 0 (5)
connecting to a phase line and each phase line has at most one
load at a node. Since unbalancing is usually not emergent I j,ij d C j (6)
condition as long as it does not cause device overloading or
extensive voltage drops or ground relay tripped, phase swap- x iij,Ȧ  ^0,1` for all i (7)
ping is carried out during maintenance or restoration periods.
[2] Where,
Utilities pursue balanced operating conditions in their dis- j is a branch.
tribution systems [1]. Phase balancing has several significant Uj is the unbalanced flow in branch j.
benefits, such as improving power quality and utilization fac- Ij,ij is the current that flow by the phase ij of branch j.
tor of existing facilities, and reducing energy losses and price. x iij,Ȧ is the decision variable for Z-th load of phase ij at
However, each phase swapping operation is associated with
certain costs on lineman expenses, maintenance expenses, and node i.
considered outage duration. The number of phase swapping Cj is the phase line capacity of branch j.
need to be compromised between benefits and costs.
 3

In the model, the equation (1) is the unbalanced flow on TABLA II

LINDO PROGRAM OF THE CASE OF STUDY
branch j; the equation (2) represents the KCL applied in the
node i; the equation (3) assures that each phase have a load
assigned; the equation (4) assures that each load is assigned to MIN 0.6U1+ 0.4U2
a single phase; the equation (5) assures that the two-phase SUBJECT TO
XA31+XB31+XC31=1
branches contain only two phases; the equation (6) is the line XA32+XB32+XC32=1
capacity constraint; the equation (7) is the decision variable XA41+XB41+XC41=1
that indicates if a load in determined node is connected to a XA51+XB51+XC51=1
XA52+XB52+XC52=1
phase or not. The objective of the model is to minimize the XA61+XB61+XC61=1
unbalanced flow on branch j. XA62+XB62+XC62=1
The constraint (5) was added respect the original model [2]. XA31+XA32<=1
XB31+XB32<=1
The inclusion of this constraint is the main contribution of this XC31+XC32<=1
paper. Without this restriction, the model is not able of distin- XA51+XA52<=1
guishing if several one-phase loads are always connected to XB51+XB52<=1
the same terminals. For instance, three one-phase loads might XC51+XC52<=1
XA61+XA62<=1
be connected at the same terminals A-B instead of the correct XB61+XB62<=1
A-B, B-C and A-C terminal connections. XC61+XC62<=1
The algorithm is solved using LINDO platform, which has XA51+XA52-XA61-XA62=0
XB51+XB52-XB61-XB62=0
the capacity to solve a model with almost 2000 nodes. XC51+XC52-XC61-XC62=0
-U1+IA1-IB1<=0
V. CASE OF STUDY -U1-IA1+IB1<=0
-U1+IB1-IC1<=0
Subsequently, the example shown in [2], with the results -U1-IB1+IC1<=0
obtained with the methodology employed there and comparing -U1+IC1-IA1<=0
-U1-IC1+IA1<=0
with the result obtained with the proposed algorithm. In Fig. -U2+IA2-IB2<=0
2 the system of six nodes of the example is shown. -U2-IA2+IB2<=0
-U2+IB2-IC2<=0
-U2-IB2+IC2<=0
-U2+IC2-IA2<=0
-U2-IC2+IA2<=0
IA1<=150
IB1<=150
IC1<=150
IA2<=100
IB2<=100
IC2<=100
-IA2-25XA51-20XA52-10XA61-15XA62+IA1=20
-IB2-25XB51-20XB52-10XB61-15XB62+IB1=30
-IC2-25XC51-20XC52-10XC61-15XC62+IC1=20
Fig. 2. Example of a six nodes system -10XA31-30XA32-15XA41+IA2=25
-10XB31-30XB32-15XB41+IB2=35
The data of the six nodes system loads are presented in the -10XC31-30XC32-15XC41+IC2=40
END
Table I. INTE XA31
INTE XB31
TABLA I INTE XC31
DATA OF THE ORIGINAL SYSTEM OF SIX NODES INTE XA32
INTE XB32
INTE XC32
Phase A B C INTE XA41
INTE XB41
Node INTE XC41
INTE XA51
1 20 A 30 A 20 A INTE XB51
2 25 A 35 A 40 A INTE XC51
INTE XA52
3 10 A 30 A N/A INTE XB52
INTE XC52
4 N/A N/A 15 A
INTE XA61
5 N/A 25 A 20 A INTE XB61
INTE XC61
6 N/A 10 A 15 A INTE XA62
INTE XB62
INTE XC62
 4

The problem was solved by LINDO in 275 iterations. Table TABLA V

DIFFERENCES BETWEEN RESULTS OBTAINED APPLYING BOTH ALGORITHMS
III shows the LINDO results obtained for the case of study.

Original Modified Modified vs.

TABLA III Algorithm Algorithm Original
CASE OF STUDY RESULTS Variables
Balancing Balancing Algorithm
OBJECTIVE FUNCTION VALUE
(%) (%) (%)
1) 13.00000
I1,A 72.73 81.82 9.09
VARIABLE VALUE REDUCED COST
XA31 1.000000 0.000000
XB32 1.000000 -6.000000 I1,B -26.92 -26.92 0.00
XA41 1.000000 0.000000
XC51 1.000000 15.000001
XA52 1.000000 0.000000 I1,C -4.55 -9.09 -4.55
XA61 1.000000 0.000000
XC62 1.000000 9.000000
U1 5.000000 0.000000 I2,A 14.29 42.86 28.57
U2 25.000000 0.000000
IA1 100.000000 0.000000
IB1 95.000000 0.000000 I2,B 0.00 0.00 0.00
IC1 00.000000 0.000000
IA2 50.000000 0.000000
IB2 65.000000 0.000000 I2,C -9.09 -27.27 -18.18
IC2 40.000000 0.000000

ǻI1 -86.67 -93.33 -6.67

From the results shown in Table IV can be observed that
the algorithm presented in [1] reduce the unbalances from ǻI2 -16.67 -16.67 0.00
57A to 16A. Nevertheless, with the changes introduced a ǻI -71.93 -77.19 -5.26
reduction from 57A to 13A is obtained, 3A less than the
4
original algorithm. Differences between results obtained ap-
Fig. 3 shows the six nodes system with the resulting phase
plying both algorithms are shown in Table V. With the pro-
current flow represented on it for modified algorithm.
posal method 5.26% less unbalanced flow was obtained. This
it would not be able to result thus always, but if worse comes
to worst will be obtained the same result that the original algo-
rithm.

TABLA IV
RESULTS OBTAINED APPLYING BOTH ALGORITHMS, IN AMPS

Original Modified
Original Algorithm Algorithm
Variables
System Phase Phase Fig. 3. Modified Algorithm Results
Balancing Balancing
I1,A 55 95 100
VI. CONCLUSIONS
I1,B 130 95 95 In this paper, it is proposed a general methodology to iden-
I1,C 110 105 100 tify risks that pass unnoticed when distribution feeders are
significantly unbalanced. The procedure proposes the neces-
I2,A 35 40 50 sary actions required to avoid accidents that involve to the loss
I2,B 65 65 65 of materials and personal damages. Unbalanced feeders could
cause several problems in the system and an increase on the
I2,C 55 50 40 costs of the utility. For that reason phase balancing is an ap-
ǻI1 75 10 5
propriate methodology to correct the disturbances.
The unbalanced flows in an electrical network are produced
ǻI2 30 25 25 by single and two-phase loads connected in three phase sys-
tems. This unbalance can be alleviated by swapping certain
ǻI 57 16 13
phases of the system. An optimal swapping strategy could
improve the performance of the network reducing notably
phase overloads and energy losses.
 5

VII. REFERENCES VIII. BIOGRAPHIES

[1] J. Zhu, G. Bilbro and M. Chow, "Phase balancing using simulated an- H. M. Khodr (M’99) received the Ph.D., M.Sc. and B.Sc. degrees in Electri-
nealing," IEEE Transactions on Power Systems, vol. 14, pp. 1508-1513, cal Engineering from the José Antonio Echeverría Higher Polytechnic Insti-
Nov. 1999. tute (ISPJAE) in 1997 and 1993 respectively. He is a Professor of Electrical
[2] J. Zhu, M. Chow and F. Zhang, "Phase Balancing using Mixed-Integer Engineering at Universidad Simón Bolívar (USB), Caracas, Venezuela. Pres-
Programming," IEEE Transactions on Power Systems, vol. 13, pp. 1487- ently, he is a Senior Researcher at INESC Porto, Portugal on leave of USB.
1492, Nov. 1998. His current research activities are concentrated in the planning, operation and
[3] H. Rudnick, I. Harnisch and R. Sanhueza, “Reconfiguration of Electric economics of electrical power systems.
Distribution Systems,” UTA Engineering Faculty Magazine, Chile, vol.
4, 1997. Ismaldo J. Zerpa M. obtained the title of Specialist in Electrical Installations
[4] Z. Quin, D. Shirmohammadi and W. Liu, “Distribution feeder reconfigu- and of Electrical Engineer in the Universidad Simón Bolívar, in 2003 and
ration for operation cost reduction,” IEEE Transactions on Power Sys- 2001 respectively. At present is performed in the Department of Energy and
tems, vol.12, pp. 730-735, May 1998. Petroleum, in the area of Regulation and Electric Rates.
[5] Z. Quin, D. Shirmohammadi and W. Liu, “Distribution feeder reconfigu-
ration for service restoration and load balancing,” IEEE Transactions on P. M. De Oliveira De Jesus (M’03) received the M.Sc. and the Electrical
Power Systems, vol.12, pp. 724-729, May 1997. Engineering degree in 2002 and 1.995 from Universidad Simón Bolívar, Ca-
[6] S. Civanlar, J. Grainger, H. Yin and S. Lee, “Distribution feeder recon- racas, Venezuela. Former Assistant Professor of Electrical Engineering at
figuration for loss reduction,” IEEE Transactions on Power Delivery, Universidad Simón Bolívar is presently pursuing Ph.D. studies at Faculdade
vol.3, pp. 1217-1223, Jul. 1998. de Engenharia da Universidade do Porto (FEUP) and work as researcher at
[7] W. Lin and H. Chin, “A new approach for distribution feeder reconfigu- Instituto de Engenharia de Sistemas e Computadores (INESC Porto), Portugal.
ration for loss reduction and service restoration,” IEEE Transactions on
Power Delivery, vol.3, pp. 870-875, Jul. 1998. Manuel A. Matos (M’94) was born in 1955 in Porto, Portugal. Currently, he
is Full Professor at the Faculty of Engineering of the University of Porto,
Portugal, and the Manager of the Power Systems Unit of Instituto de Engen-
haria de Sistemas e Computadores do Porto (INESC), Porto, Portugal. He also
collaborates with the Management School of the University of Porto. His
research interests include fuzzy modeling of power systems, optimization, and
decision-aid methods.

View publication stats