International Journal of Grid and Distributed Computing
Vol. 9, No. 7 (2016), pp.33-42
http://dx.doi.org/10.14257/ijgdc.2016.9.7.04
ISSN: 2005-4262 IJGDC
Copyright ⓒ 2016 SERSC
Study of Fault Location Algorithm for Distribution Network with
Distributed Generation based on IGA-RBF Neural Network
Huanxin Guan, Ganggang Hao and Hongtao Yu
Shenyang Institute of Engineering, Shenyang, China
huanxin-guan@163.com
Abstract
The access of the Distributed Generation (DG) has to make the fault location problem
of distribution network extremely complex and affect the efficiency and accuracy of fault
location. According to the fault information that the Supervisory Control And Data
Acquisition (SCADA) of distribution network upload and considering the change of
configuration and logic relation of distribution network protection after DGs access, this
paper proposed a radial basis function (RBF) neural network based on Improved Genetic
Algorithm (IGA), which used the real-coded genetic algorithm with adaptive crossover
and mutation into the gradient-dropping algorithm as the RBF network learning
algorithm, and constructed a new switch function and fitness function. And then the
improved algorithm was applied to the fault location of distribution network containing
distributed power supply. The simulation results show that the RBF neural network based
on IGA not only has the advantages of simple structure and fast operation, but also has
better generalization performance. The analysis and comparison results show that the
optimized Improved algorithm can effectively improve the convergence speed and
precision, and it has good fault-tolerance to the lack or distortion of the fault information.
Keywords: Distributed generation; distribution network; fault location; IGA-RBF
neural network; gradient-dropping algorithm; self-adaptive
1. Introduction
In the current distribution network automation system at home and abroad, the feeder
terminal unit (FTU) and other devices are widely used for distribution system fault
location to provide a richer remote communication and remote measure data. With the
development of communication technology and computer technology, the feeder
automation technology is also increasingly intelligent, integrated, and now advanced
feeder terminal device FTU integrated remote metering, remote communication, remote
control, fault monitoring and many other features. So the main station remote FTU mode
has become the mainstream of the development of distribution automation system. Under
this situation, the research on the fault location method of the distribution network with
the full use of FTU information is booming, and the method to solve the fault location
mainly includes matrix method
[2,3]
, expert system
[5]
, genetic algorithm
[6]
, neural network
method
[7]
, etc.
The information sources used in distribution network fault location mostly come from
the outdoor FTU, which is influenced by weather, electromagnetic interference and other
factors. So FTUs upload information may appear distorted. The method proposed by
reference
[2]
is more accurate to the information required to upload. Although the matrix
algorithm is intuitive and fast, it has poor fault tolerance to the fault information. The
neural network has a strong self-learning ability, nonlinear mapping ability and fault
tolerance, which has been widely applied for fault detection and fault location of power
system, but its results of the training are not stable and easily fall into local optima. The
genetic algorithm is a global optimization algorithm based on the genetic mechanism of