Research Of Rough Set-Neural Network Fault Diagnosis Method About Distribution

The aim of this thesis is to improve on a practical electric power distribution fault diagnosis system. A new electric power distribution fault diagnosis algorithm in accordance with rough set-neural network is proposed after dissertating the current power distribution system and its supporting system and analyzing application characteristic of various kinds of artificial intelligence in fault diagnosis.Firstly, the requirements and significance of electric power distribution fault diagnosis system and the current diagnosis methods as well as the disadvantages are discussed in the thesis. On the basis of fault diagnosis network model, knowledge representation system of rough set theory is taken as a major tool to simplify the complex combine neural network and in which unnecessary properties are eliminated. The method overcomes some shortcomings, such as network scale is too large and the Rate of classification is slow. Based on rough set combine neural network model is presented. Then, a satisfying result is described by using data fusion.Reduction of attribute is the core of rough sets, finding all reduction of attribute is NP hard. The binary discernible matrix is ameliorated in theory, finding out all reduction of attribute is easy and fast. How to set up a good artifical neural is very hard. In the paper, figuring out core of data and coefficient about RBF network through K-means and SOFM. It is the very good neural network which is nicety and speedy for fault diagnosis.Finally, setting up a simple system of fault diagnosis about distribution by the Visual C++ and Matlab, and the results of an example are given.