小波神经网络在配电网故障定位中的应用

Applied Research of Wavelet Neural Network with Momentum for Fault Location in Power Distribution

提出一种适用于配电网架空线路的单相接地故障定位新方法,该方法以暂态零序电流的小波能量、有功功率、无功功率作为数据融合的特征量,再结合紧致型小波神经网络(WNN)来进行故障定位,并针对小波神经网络存在收敛速度缓慢且容易陷入局部极小的问题,给出一种参数修正改进的算法,通过在权值调整式中增加动量项来选择学习步长,且以新方法初始化各个权值以提高网络学习效率。大量的Matlab仿真结果表明:此方法具有很好的单相接地故障定位性能,实验准确率基本可以达到100%,可应用于配电网故障定位。

A new method of single-phase-to-ground fault location for overhead lines of distribution network was presented.It was finished fault location by using the wavelet energy,active power and reactive power of the transient current as the characteristics and based on the close-type WNN.To against WNN which has low convergent speed and fall into local minimum easily,this paper given a parameter corrected algorithm,which was adopt in the momentum in the learning algorithms and initialize all the value,it helps improve network learning rate.A large number of Matlab simulation results showed that this method has a very good single-phase grounding fault location performance;the experimental accuracy could basically reached 100% and applied to distribution network fault location.