基于进化神经网络的变压器故障预测模型研究

Research on the Fault Prediction Model Based on Evolutionary Neural Network

分析BP神经网络算法,根据变压器故障诊断问题的特征,改进其算法,即引入了遗传算法全局搜索能力。由变压器故障诊断常规方法,收集故障和油气气体成分之间的数据样本,得到故障和油气气体成分之间的非线性关系模型,并且进行检测样本检验,证明其判断模型,有一定的外推能力,可以通过此模型进行变压器绝缘故障诊断。

BP neural network algorithm is analyzed in this paper, and the algorithm is improved according to the fault diagnosis problem characteristics of transformer, namely global searching ability of the genetic algorithm is introduced. Data samples between faults and hydrocarbon gas composition were collected by conventional fault diagnosis methods of transformer, nonlinear relationship model of faults and hydrocarbon gas composition was obtained, and detection of sample tests was performed. It proved that the judgment model had certain ability of extrapolation, and transformer insulation fault diagnosis can be performed by this model.