基于SOFM神经网络的变压器故障诊断研究

摘要

SOFM神经网络具有强大的非线性映射能力和高度的自组织和自学习能力，将SOFM神经网络应用于变压器的故障诊断。利用改进的罗杰斯三比值法获取变压器故障诊断的特征向量，建立了SOFM网络故障诊断模型，并对模型进行训练。为了检验模型的实际诊断能力，以变压器的4种典型故障诊断为例进行仿真实验。仿真结果表明：SOFM神经网络能够根据获胜神经元在竞争层的位置对变压器故障进行判断，诊断准确率高，收敛速度快，泛化能力强，表明基于SOFM网络的变压器的故障诊断是一种行之有效的方法。%Self-organizing feature mapping(SOFM)neural network has a strong nonlinear mapping ability as well as a powerful self-organizing and self-learning ability. It is applied to fault diagnosis of transformers. Improved Rogers three-ratio method is used to obtain the characteristic vectors of transformer fault diagnosis. First,a diagnosis model based on SOFM neural network is established and trained. To test the practical diagnosis ability of the model , 4 kinds of typical faults of transformers are taken as examples in the simulation experiment. The simulation results show that SOFM neural network can identify the fault types according to the location of winning neurons in the com-peting layer. And it has high accuracy,fast convergence speed and strong generalization ability,which indicates that the transformer fault diagnosis method based on SOFM neural network is effective .