基于小波熵神经网络的直流系统环网接地故障检测

Grounding fault detection based on wavelet entropy and neural network for loop net of DC system

直流系统环网是接地故障检测中的一个关键因素。针对直流系统中环网对接地故障检测的影响,基于小波熵理论,提出一种新的检测环网接地故障的方法。该方法利用小波分析具有时频局部化特性和熵能对系统状态表征的特点,将小波分析和熵结合起来完成信号的特征挖掘。通过低频信号注入,采集环网状态,计算小波熵作为系统的特征参数,运用这些特征参数作为输入样本,训练BP神经网络,建立神经网络故障检测系统,以实现智能化的故障识别。仿真分析证明,环网发生接地故障前后的小波熵具有显著差别。

Loop net of DC system is an important factor in the grounding fault detection. A grounding fault detection method based on wavelet entropy is proposed for loop net of DC system, which combines wavelet analysis with entropy theory to mine the signal characteristics. The former has the ability of time-frequency localization and the latter has the ability of system state expression. By injecting low frequency signal and gathering loop net states,wavelet entropy is calculated as system characteristic parameter,which is used as input sample to train BP neural network. The neural network fault detection system is thus built to realize intelligent fault recognition. Simulation results show that there is distinct difference between wavelet entropies before and after grounding fault.