粒子群优化自适应小波神经网络在带电局放信号识别中的应用

Application of Adaptive Wavelet Neural Network Based on Particle Swarm Optimization Algorithm in Online PD Pattern Recognition

XLPE中压电缆局部放电(PD)带电检测获得的信号可能源于电缆本体、电缆终端头,也可能来自于与之连接的开关柜中的电晕放电或表面放电等。由于不同来源的PD信号,对设备的危害不同,其判断标准也有所不同,故有必要对PD信号的来源进行识别。本文提出一种基于自适应小波神经网络的XLPE电缆PD识别方法,构建了一个4层自适应小波神经网络模型,对实验室获得的PD波形进行识别;提出使用粒子群算法先进行一次优化,后使用BP算法进行二次优化的训练方法;讨论了不同网络结构及小波函数对网络性能的影响。研究结果表明PSO-BP组合优化的自适应小波神经网络的训练效果明显优于单独使用BP算法,能够准确、可靠的对PD信号进行识别。

The(partial discharge, PD) signal of XLPE cable may come not only from the body of the cable and its termination, but also from the corona discharge or surface discharge of the switchgear connected with the cable. Different PD sources have different damages on the equipment, and their differentiating criteria are different too. Therefore it is necessary to recognize different PDs. A new pattern recognition method based on adaptive wavelet neural network is proposed and an adaptive wavelet neural network with four-layer is given to recognize PD source in this paper. Particle swarm optimization algorithm is used to optimize the network first, and then BP algorithm is used to make a second optimization, whose performance is remarkably better than that only using BP algorithm. Influences of different wavelets and different structures to the performance of the adaptive wavelet neural network are discussed. The results show that the adaptive wavelet neural network which is optimized by both particle swarm optimization and BP algorithms is able to recognize PD source accurately and reliably.