基于复合神经网络的变压器故障诊断研究

Research on transformer fault diagnosis based on composite neural network

针对变压器工作环境复杂、故障诊断难的问题,提出一种基于QIA-RBF复合神经网络的变压器故障诊断方法。分析变压器常见的故障,对RBF神经网络进行设计,制定QIA免疫算法流程。开展变压器故障诊断研究,对不同诊断方式的诊断结果进行分析,结果表明复合QIA-RBF神经网络比单独的RBF神经网络对变压器故障的诊断精度高、误差小,能够满足变压器故障诊断需要。

Focusing on the complex working environment and difficulty in fault diagnosis of transformers,a transformer fault diagnosis method based on QIA-RBF composite neural network is proposed.Then the composite neural network is trained by the actual manual periodic inspection data,and the accuracy of its prediction results is compared with that predicted by bare RBF neural network.The results show that the composite QIA-RBF neural network has higher diagnostic accuracy and small error than the bare RBF neural network,which can meet the needs of transformer fault diagnosis.