Application of GA-BP neural network on partial discharge of pattern recognition in GIS

In order to improve pattern recognition based on partial discharge detected with ultrasonic method, repetitiveness of partial discharge(PD) in different defects are measured and 34 steady defect characteristic parameters are got from the extracted 43 characteristic parameters. Then, the 24 effective characteristic parameters are filtered as input of neural network. Finally, an improved GA-BP neural network is proposed. After training, the result shows that the application of GA-BP neural network improves the recognition rate.