PSO优化BP神经网络的串联故障电弧识别方法

Series fault arc identification method based on BP neural network optimized by PSO

运用db5小波对故障电弧信号进行4层分解,提取故障频段能量谱作为特征量,建立BP神经网络。用粒子群优化(PSO)算法优化BP神经网络,从而快速准确地对故障电弧特征量进行拟合,用训练后的神经网络对故障电弧进行预测,达到了较好的预测识别效果,验证了该串联型故障电弧识别方法的有效性。

By means of db5 wavelet,fault arc signal is decomposed into four layers,energy spectrum fault frequency band is extracted as characteristic quantity,establish BP neural network. Optimize BP neural network by particle swarm optimization（ PSO） algorithm,so as to fastly and accurately fit fault arc characteristic quantity,fault arc is predicted by trained neural network,achieve good recognition effect,verify effectiveness of fault arc recognition method.