Abstract:

For the small number of samples in the classification of partial discharge pattern in high voltage electrical appliances and the poor recognition rate of conventional classification methods,a mixed algorithm is proposed based on probabilistic neural network hybrid algorithm and wavelet transform.Wavelet decomposition is performed on partial discharge signals in a laboratory simulation with the extracted wavelet energy coefficient as the feature parameter,and as the input of a probabilistic neural network for classification.The obtained results are better than that by the multilayer feedforward neural network (MLP) algorithm and the support vector machine method using sequential learning optimization.

Key words: probabilistic neural network;wavelet transform;partial discharge;pattern recognition