基于改进BP神经网络的电气设备绝缘故障诊断研究

Research on Electrical Equipment Insulation Fault Diagnosis Based on Improved BP Neural Network

文章以变压器绝缘故障为例,首先采用改进谐波分析法进行在线监测,获取变压器绝缘的原始信号,并对已获取信号进行预处理工作。其次采用自适应学习率改进BP算法,并搭建3层前馈神经网络,构建新型神经网络模型。最后应用改进的BP神经网络模型,结合熵值法完成变压器信号训练,实现变压器绝缘故障诊断。实验结果表明,此方法可有效提升电气设备绝缘故障诊断的准确性,缩短整体诊断耗时,具有较高的实际应用价值。

This article takes the insulation fault of a transformer as an example.Firstly,an improved harmonic analysis method is used for online monitoring to obtain the original insulation signal of the transformer,and the obtained signal is preprocessed.Secondly,an adaptive learning rate is adopted to improve the BP algorithm,and a three-layer feedforward neural network is constructed to construct a new neural network model.Finally,an improved BP neural network model was applied,combined with entropy method to complete transformer signal training and achieve transformer insulation fault diagnosis.The experimental results show that this method can effectively improve the accuracy of insulation fault diagnosis in electrical equipment,shorten the overall diagnosis time,and has high practical application value.