神经网络与弱学习机制算法在单相接地故障判断中的对比分析

Comparative analysis of neural network and weak learning mechanism algorithm in single-phase grounding fault judgment

为了进一步提高单相接地故障识别准确率,对比研究了基于BP神经网络与弱学习机制的单相接地故障分类算法在判断单相接地故障准确率方面的问题。根据10 kV架空线路中所采集到的监测数据,使用‘相应电场下降百分比与电流上升突变值构建的向量空间’作为训练样本。采用‘基于终端吸引子的改进的全局寻优自适应快速BP学习算法’与‘基于决策树的弱学习机制算法’,对单相接地故障识别问题进行研究,并对比了两种算法之间的性能。结论为:相比于神经网络算法,弱学习机制算法能够更好地弱化向量识别时阈值问题,从而在进行单相接地故障分类时能将准确率提高到99%以上。

In order to further improve the accuracy of single-phase grounding fault recognition,the problem of single-phase grounding fault classification algorithm based on BP neural network and weak learning mechanism in judging the accuracy of single-phase grounding faults is compared and studied.According to the monitoring data collected from the 10kV overhead line,the vector space constructed by the corresponding electric field drop percentage and the current rise mutation value is used as the training sample.Using the improved global optimization adaptive fast BP learning algorithm based on terminal attractor and the weak learning mechanism algorithm based on decision tree,the problem of single-phase ground fault identification is studied and the performance between the two algorithms is compared.The conclusions are as following:Compared with the neural network algorithm,the weak learning mechanism algorithm can better weaken the threshold problem in vector recognition,so that the accuracy of single-phase ground fault classification can be improved to more than 99%.