基于小波分析和BP神经网络的触电信号检测模型

Detecting model of electric shock signal based on wavelet analysis and BP neural network

针对从农村低压电网总泄漏电流中检测和判断触电电流信号的难题,该文提出一种基于小波变换和BP神经网络的触电信号检测方法。首先用触电物理实验平台对动物触电信号进行实测,选择合适的小波基和分解尺度对触电实验中总泄漏电流及触电电流进行小波多分辨分析,实现原始信号的预处理;再将预处理后的波形作为样本进行神经网络学习和训练,建立从总泄漏电流波形中提取触电电流波形的神经网络耦合模型,并用此模型对未训练的样本进行触电信号的检测,检测值与实际值的平均相对误差为3.93%,说明该方法能够从总泄漏电流中检测出触电电流信号,对于开发新一代剩余电流保护装置具有一定的参考价值。

It is difficult to exactly detect and judge a weak electric shock signals in the summation leakage current on the low-voltage electric power grid.Integrating the merit of wavelet transform with that of BP neural network,a novel detection of the electric shock current based on wavelet transform and BP neural network was introduced in this paper.First of all,the animal electric shock signals was tested by physical experiment of electric shock,and an appropriate wavelet base and decompose scale was chosen to analysis the summation leakage current and the electric shock current.And then,the wave shape of specimens pretreated by wavelet transform was trained by BP neural network.A neural network coupling model of extracting electric shock current from the summation leakage current was built,and used to detect electric shock current of the untrained specimens.The average relative error between detected value and actual value was 3.93%.The result indicated that this method can detect electric shock current in the summation leakage current,and can be a reference for the development of new generation residual current operated devices.