反馈式Hopfield神经网络在输电线路故障诊断中的应用

APPLICATION OF HOPFIELD NN IN POWER TRANSMISSION LINE FAULT DIAGNOSIS

本文提出采用容错性能较高的Hopfield联想记忆（AM）神经网络（NN）来完成输电线路的故障诊断，建造了实现该问题的NN模型结构，并提出采用基于投影原理的伪逆学习算法。通过仿真和对NN模型系统容错性的测试分析表明，本文所建造的基于HopfieldAMNN能够对有干扰的实时输入信息情况具有较好的容错性，充分体现了反馈式NN在电力系统实际应用中的优势。

This paper proposed to adopt Hopfield Associatlve memory (AM) Neural Network (NN) with high fault-tolerance property for transmission line fault diagnosis, built AM NN model structure,and presented to use projectionbased fake converse learnign algorithm. Throughsimulation and testing analysis to NN model fault-tolerance ability, the results are shown that Hopfield AM NN built in this paper has good fault-tolerance property for distorted real-time input information sequence,this also demonstrated practical application superiority of feedback NN in power system.