将二代小波与神经网络相结合进行局部放电故障分类。基于二代小波与信息熵理论，提取放电信号，以小波能谱熵与系数熵作为特征量。将提取的特征向量输入神经网络进行训练，训练时通过改进共轭梯度法自适应调整误差，得到最优训练网络。采用该文算法、经典神经网络以及小波神经网络，分别对放电模型产生的3种放电类型进行识别测试的结果表明：该文方法在识别准确率以及算法执行效率上，均优于经典神经网络以及小波神经网络。

Second generation wavelet (SGWT) and adaptive BP neural network were combined to classify partial discharge fault. Partial discharges (PD) signal was recognized based on SGWT and information entropy theory. Wavelet energy entropy and coefficient entropy were taken as characteristic quantity, and input into neural network for training. In the training process, the neural network could adaptively adjust error to obtain the optimal training network by using the improved conjugate gradient methods. Finally, the comparison between the proposed algorithm, classic neural network and wavelet neural network was carried out on the recognition test of three kings of PDs caused by discharge model, whose results showed that the recognition accuracy and execution efficiency of the proposed algorithm were better that those of classic neural network and wavelet neural network.