论文部分内容阅读
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。
The second generation of wavelet and neural network combined partial discharge fault classification. Based on the theory of second-generation wavelet and information entropy, the discharge signal is extracted, and the wavelet energy spectral entropy and coefficient entropy are taken as the characteristic quantities. The extracted feature vectors are input to the neural network for training, and the error is adjusted adaptively by the improved conjugate gradient method to obtain the optimal training network. By using this algorithm, classical neural network and wavelet neural network, the three types of discharges generated by the discharge model are respectively identified and tested. The results show that the proposed method is superior to classical neural networks in recognition accuracy and algorithm execution efficiency and Wavelet neural network.