朱雁斌

(云南电网有限责任公司昆明供电局，云南 昆明　650011)



基于改进BP神经网络的接地故障定位研究

朱雁斌

(云南电网有限责任公司昆明供电局，云南 昆明650011)

摘要：考虑到经典的小波包频带能力特征提取方法不将频率时变特性进行考虑而统计全部的频带，而且经典的小波能量谱算法没有将每个分解频带的能量随着时间轴分布特性进行充分考虑，提出使用卷积小波包能量矩对单相接地故障信息的特征向量进行提取，从而为单相接地故障定位的判断提供依据。使用BP神经网络建立配电网单相接地故障的定位算法；由于常规BP神经网络容易陷入局部最小值，并且有算法收敛慢、训练时间长等问题，提出使用遗传算法对BP神经网络进行优化，同时为提高遗传算法优化效率，使用混合编码方式对遗传算法进行改进。最后通过实验验证所提出的改进型BP神经网络的接地故障定位算法的性能，结果表明，故障定位的精度有了较大的改善，验证了所提出方法的可行性。

关键词：接地故障；BP神经网络；改进遗传算法；故障定位

Abstract：The frequency band feature extraction method of classical wavelet packet does not consider the frequency time varying characteristics, and the classical wavelet energy spectrum algorithm also does not consider that the features of every decomposed frequency band energy are distributed along with time. The feature vector of single-phase earth fault information is extracted by convolution type of wavelet packet energy moment, which provides a reference for the determination of single-phase earth fault location. BP neural network is used to establish the location algorithm of single-phase earth fault. Because the conventional BP neural network is easy to fall into local minimum, and the algorithm is slow and the training time is long, the genetic algorithm is used to optimize BP neural network, and the hybrid encoding is used to improve the efficiency of genetic algorithm. Finally, the performance of the improved BP neural network is verified by the experiments. The results show that the accuracy of fault location is greatly improved and the feasibility of the proposed method is verified.

Key words：ground fault; BP neural network; improved genetic algorithm; fault location

0引言

近年来电力部门针对配电系统故障进行数据统计，发现其中60%以上的配电系统故障属于单相接地引起的。因此，对于电力部门来说，快速定位故障线路位置及时解决单相接地引起的配电系统故障是目前急需研究解决的主要问题之一，也是相关科学工作者研究的热点问题之一[1-4]。

文献[5]中使用卷积型小波包能量矩对接地故障时暂态电流信号提取特征向量，使用免疫粒子群优化对神经网络进行优化并建立精确快速识别接地故障的识别定位模型。……