Diagnosis Of Faulty Porcelain Insulators Based On Artificial Neural Network

Corona pulses current method is a way which can be used to detect faulty insulators on ground. But its accuracy of detecting is low in application. In this paper we try to find a method based on artificial neural network to improve the accuracy of detecting faulty insulators. We took a lot of data of corona pulses from the laboratory and field by using the detecting device. We got the corona finger prints for both the good insulators and the faulty insulators by statistic of the data. We firstly smoothed these finger prints by wavelet transform. Then we expressed these finger prints by some statistic parameters such as skewness, kurtosis, number of peaks, asymmetry, and the cross-correlation factor. Lastly we put these statistic parameters into the artificial neural network for pattern recognition. It is noted from our study that smoothing the finger prints with wavelet is great benefit to the pattern recognition. We found from the result that the identification rate to the laboratory data is above 90%, to the field data, the identification rate is about 80%, and to the data assembled from laboratory and field, the identification rate is about 80%.