变压器故障诊断的符号-神经网络系统的研究

A Study of Symbolic-Neural Network Systems for Transformer Fault Diagnosis

鉴于符号推理系统存在的问题和神经网络技术应用的局限性,提出了符号-神经网络系统的诊断方法,构造了基于变压器故障综合诊断思想的混合型专家系统的框架,并重点研究了特征气体法和改良IEC法的分解神经网络诊断方法。这种符号-神经网络系统将大大提高故障诊断水平和功能。

In view of the problems of the symbolic inference system and the application limit of artificial neural networks technology,the paper proposes a diagnosis method of symbolic-neural networks system. The layout of hybrid expert system is set up on the basis of tranformer synthetical diagnosis idea. The paper mainly studies the diagnosis method of decomposition ANN system for critical-gas analysis method and improved IEC method. The symbolic neural network system greatly raises the fault diagnosis level and function.