You are on page 1of 4

PHASE BALANCING USING HEREFORD RANCH ALGORITHM

M. GANDOMKAR Department of Electrical Engineering Islamic Azad University Saveh IRAN


Abstract: Unbalanced loads and feeders increase power system investment and operating costs. Single-phase lateral loads phase swapping is one of the popular methods to balance the systems. Phase swapping is a large scale combinational optimization problem and it is necessary to apply a proper optimization method. This paper presents a new approach to phase balancing on Hereford ranch algorithm (HRA) in a LV distribution network. The proposed method was tested for two sample LV feeders. Also to show its effectiveness, the results of suggested algorithm are compared with classical genetic algorithm. Keywords: Phase Balancing, Phase Swapping, Distribution Networks, Hereford Ranch Algorithm (HRA), Genetic Algorithm (GA)

1 Introduction
Due to single-phase, two-phase, unbalanced threephase loads and unbalanced impedance, the distribution systems operate in unbalanced fashion. Unbalanced feeders not only increase energy losses and the risk of overload situations, but they also affect power quality and electricity price. A severely unbalanced circuit can result in excessive voltage drops on the heavily phase. Even a feeder system is designed as a balanced feeder based on the given load data, load prediction errors and unbalanced load growth will induce feeder imbalance. The purpose of phase balancing is to find the optimal phase swapping scheme to balance an unbalanced feeder system with minimum cost[1]. There are two approaches for phase balancing: one is feeder reconfiguration at the system level; the other is phase swapping at the feeder level. Feeder reconfiguration has been extensively studied in the past several decades while phase swapping has been ignored. Since feeder reconfiguration is approximately primarily designed for load balancing among the feeders, most researchers do not consider phase balancing as an objective in feeder reconfiguration [2]. Phase swapping is normally carried out during maintenance and restoration periods. Phase swapping is a direct and effective way to balance a feeder in terms of phases. For the time being engineers already use phase swapping to balance

phase loading based on their experience and trial and error methods. The procedure is labor intensive, sub-optimal, and time-consuming. In phase swapping algorithm, control variables are the phase swapping options at each candidate node and the cost function is to minimize the power losses subject to load flow equations, capacity constraints, voltage and phase balancing requirements[2]. It is a large scale combinatorial optimization problem and is NP-complete (nondeterministic polynomial time complete). Thus its computing effort increases exponentially with the size of candidate set. It is formulated into MixedInteger Programming (MIP), which is suitable for the linear objective function [1]. But in many cases, the function may not well represent some decisionmaking criteria. For example, minimizing energy losses will make the objective function as a nonlinear integer function, which is difficult to be solved analytically. While intelligent computation methods, such as Fuzzy Logic, Genetic Algorithm and Simulated Annealing are powerful to solve nonlinear integer programming problems[2]. In these problems, existing genetic algorithm (simple genetic algorithm (SGA)) can achieve global optimum through the directional irregular searching but has several drawbacks such as excessive convergence time and premature convergence. High selective pressure for better convergence brings about an increasing diversity loss rate in population

and premature convergence is owing to high diversity loss rate [3,4]. START Generation initial population (population size=n) Evaluate fitness The m fittest strings are identified as the breeding males Selected strings as breeding males are effectively removed from population Calculate the relative fitness of each string One female string is selected to breed by Roulette Wheel Selection Set the selected female strings fitness as 0 Selected female is mated with a selected male from the list of males in a round-robin fashion and performs crossover of those

In this paper, to overcome the defects of existing SGA, Hereford ranch algorithm (HRA) is applied to solve the optimal phase movement schedule for phase swapping problem.

2 Problem Formulation
Solving the problem is to optimize the power losses with a proposed load re-phasing schedule, subject to certain constraints. For a LV feeder, the solution of the problem based on trial and error will be timeconsuming and sub-optimal. In practice, the possible re-phasing numbers are limited, too. To investigate the effects of voltage drops, conductor capacity and losses, a full unsymmetrical load flow model have to be considered in the phase balancing formulation. The losses function can be stated as:

F = ( I ai Rai + I bi Rbi + I ci Rci + I ni Rni ) (1)


2 2 2 2 i =1

Where:

I ai , I bi , I ci : current of i th segment of LV network


for A,B,C phases respectively Rai , Rbi , Rci : resistance of i th segment of LV network for A,B,C phases respectively I ni , Rni : neutral current & resistance of i th segment

l : maximum segment number of LV network


total number of generated offspring=n ? Yes Performs mutation process The children replace the parents No The problem can thus be stated as: Minimize F Subject to - The node voltage cannot violate between its upper and lower limit. - The conductors capacity cannot be over. - Load phase (single or double phase) must be considered. - Re-phasing number(s) must be considered

3 Hereford Ranch Algorithm


Generally, GA's ability to find a correct solution in a variety of problems, to preserve diversity for preventing premature convergence and to improve convergence time are must affected by parent selection algorithm for generating offspring. Sexual differentiation and sexual selective breeding are the attributes of almost all the real life for preservation and evolution of the species, but have been rarely used in most conventional GA applications. Sexual selection is the scheme that employs sexual differentiation and sexual selection algorithm outperformed roulette wheel and tournament selection algorithm in terms of accuracy, preserving

maximum generation? No Yes END Fig. 1. Flowchart of HRA

m+ f (2) n where, m is the number of selected males, f is the number of selected females, n is the total L = 1

diversity and fast convergence. HRA uses sexual differentiation and selective breeding in choosing parents for genetic string, and in this scheme, diversity loss rate can be directly affected by controlling male-to-female ratio. The diversity loss rate is

selective breeding algorithm. Table 2

to

conventional

genetic

population, and generally, the number of selected females for parents is half of total population. In that case, diversity loss rate L and diversity remain rate D will be

L = 1 MF 2

Comparison of HRA and SGA Loss Reduction (kW) # of phase Feeder 1 Feeder 2 movement SGA HRA SGA HRA 1 14.35 15.17 18.64 19.54 2 16.52 16.85 22.47 23.05 3 18.21 18.33 24.89 24.98 4 19.67 19.67 27.04 27.10 5 20.14 20.16 28.54 28.60 6 20.45 20.46 29.15 29.17 7 20.69 20.69 29.45 29.50 8 20.73 20.74 29.56 29.61

(3) (4)

D = (1 L) t
where, MF = m

is the male-to-female ratio, and

Fig.2 and Fig.3 show the convergence characteristics of HRA and SGA for feeder 1 and feeder 2, respectively. The characteristics prove that HRA converges faster than SGA from the initial stage of evolution process.
25 Loss Reduction(kW) 20 15 10 5 0 0 10 20 30 40 Generations 50 60 70
SGA HRA

t is the generation number.


In this paper, male-to-female ratio ( MF ) is given as 0.5 and the number of selected males is determined by given MF . If the number of selected males for parents is m , conventional HRA's flowchart is given in Fig.1. From selection procedure in Fig.1, the parent's string classified as a male is removed from the population, since it should not be subsequently classified as a female. Conventional HRA uses basic two-point crossover and the frequency of mutation is very low for stable convergence.

Fig.2 SGA and HRA comparison (feeder 1)

4 Test Results
Loss Reduuction(kW)

Two LV distribution feeders were analyzed by the SGA and HRA. The specifications of feeders before phase balancing procedure are shown in Table 1. Table1 PLOSS # of IA,IB,IC (A) (KW) Laterals Feeder 1 Feeder 2 135 224 170,345,260 380,210,310 108 205

35 30 25 20 15 10 5 0 0 10 20 30 40 50 Generations 60 70
SGA HRA

Table 2 shows the test results of SGA and HRA for two feeders. HRA is better than SGA for most of the cases for loss reduction values after phase movement. These results show the need of introducing sexual differentiation and sexual

Fig.3 SGA and HRA comparison (feeder 2)

5 Conclusion
In the deregulated environments, phase balancing problem, is becoming more important, because it can be improve circuit losses, line capacity, voltage drops and electricity price. In this paper, conventional GA and HRA are introduced to search optimal phase movement for phase balancing in LV distribution feeders. The proposed algorithms are applied to two feeders. From the various test results, it can be seen that all of two methods are proper for phase balancing problem and the ability to search of HRA is better than SGA for phase swapping problem. References: [1] J.Zhu,M.Y. Chow and F.Zhang, Phase Swapping to Balance a Radial Feeder System, presented in IEEE. PES.Winter Meeting 98 . [2] J.Zhu,G.Bilbro and M.Y.Chow, Phase Balancing using Simulated Annealing, IEEE. Transactions on Power System,Vol.14, No.4, Nov.1999, PP 15081513 [3] Z. Michalewiccz, Genetic Algorithm + Data Structure =Evolution Programs, Spring-verlag, New York 1992 [4] M. H. Jones, The Hereford Ranch Algorithm: An Improvement of Genetic Algorithm using Selective Breeding , Utah State University, Logan, UT, 1995

You might also like