两种前向神经网络在变压器故障诊断中的应用

Application of Two Forward Neural Networks in Transformer Fault Diagnosis

摘要:

针对依靠变压器油中溶解气体分析的传统故障诊断方法存在的不足以及未来智能诊断算法进一步发展的需要,在研究人工神经网络的基础上,介绍了两种前向神经网络在变压器故障诊断中的应用,并用大量的DGA样本数据做了仿真训练;首先讨论了几种常用变压器故障诊断方法的一些缺陷,通过分析现代智能诊断算法的局限性,得出改进人工神经网络的优势;然后结合两种前向神经网络的原理和结构,对变压器内部故障进行了分类和编码;分别设计了相应的故障诊断模型,在MATLAB中做了仿真测试,并给出了仿真程序;仿真结果表明,两种神经网络用于变压器故障识别较为理想,其中PNN网络在样本数量较多时效果更好,有效提高了变压器故障诊断的正确率。

Abstract:

Under the circumstance that the deficiency of conventional fault diagnosis method relies transformer oil dissolved gas analysis and the need for the further development of future intelligent diagnostic algorithms, this paper introduces the application of two forward neural networks in transformer fault diagnosis based on the research of artificial neural network and do simulation training with a lot of DGA sample data. Firstly, some flaws in the several common transformer fault diagnosis methods are discussed and it comes to the advantage of the artificial neural network after the analysis of the limitation of the modern intelligent diagnosis algorithm.Then the internal fault in the transformer are classified and coded with the consideration of principle and structure of two forward neural network .The corresponding fault diagnosis model is designed respectively and tested in Matlab with the simulation programs.The simulation results show that two kinds of neural network for transformer fault diagnosis identification is relatively ideal and PNN network behaves better when the Sample size is large, which improves the accuracy of transformer fault diagnosis.