Model based on rough set neural network using ant colony optimization algorithm and its assessment of fault-tolerance performance on fault diagnosis model

Abstract

In this paper is firstly presented that the associated rules are extracted based on rough set theory and are used as the theoretic basis of the connection mechanism of higher-order NN, which was composed with the feedforward NN, the composite NN model including optimization strategy of rough set calls rough set NN (RSNN). And the ant colony optimization algorithm (ACOA) was used to optimize solving and to improve the generalization performance of RSNN model. Hence the construction of presented model possesses theoretical significance and may ensure to take optimization performance and to enhance the generalization ability. Five assessment models are constructed, which use three learning algorithms (BP, GA, ACOA) and RSNN, and corresponding fault section diagnosis models for distribution network are built. Through the simulation and the analysis of fault-tolerance performance (FTP), it shows that the model based on the combination of ACOA and RSNN can effectively enhance the generalization ability and the FTP. The FTP of ACOA-RSNN can enhance 21.81%. (6 pages)