神经网络在变压器故障诊断中典型算法研究

Research on Typical Artificial Neural Networks in Power Transformer Fault Diagnosis

为了随时检测变压器状态,及早发现并排除变压器可能存在的故障,笔者将3种不同的神经网络(即BP网络、GA-BP网络与RBF网络)应用于变压器故障诊断中,分别介绍了这3种网络的结构及原理,故障诊断采用MATLAB语言编程实现。大量实验数据结果分析表明,RBF网络在诊断准确率相比其他两种网络具有一定的优势。

In order to examine transformer momentarily to exclude possible fault in time, three kinds of artificial neural network were applied in the transformer fault diagnosis, such as the BP network, the GA-BP network, and RBF network. The structures and principles of the three networks were analyzed, and fault diagnosis was realized by MATLAB programming. The analysis of the massive experimental data indicate that the RBF network is better than the others in diagnostic accuracy.