基于邻域粒子群优化神经网络的变压器故障诊断

为了提高变压器故障诊断正判率,提出了一种邻域粒子群算法优化BP神经网络的电力变压器油中气体分析(DGA)方法,即通过相关统计分析和数据的预处理,选择变压器油中典型气体作为神经网络的输入,然后利用训练好的邻域粒子群算法优化后的神经网络进行变压器故障类型诊断。试验结果表明,该类方法具有很好的分类效果,较好地解决了变压器放电和过热共存时故障的难分辨问题,对故障类型的正判率较高。

Power Transformer Fault Diagnosis via Neural Network Based on Particle Swarm Optimization with Neighborhood Operator

In order to improve the correct judgement rate in power transformer fault diagnosis,this paper investigates a dissolved gas analysis method of transformer via neural network based on particle swarm optimization with neighborhood operator.Based on correlation analysis and data pretreatment,some typical gases in transformer oil are selected as the input of neural network for training,then the fault diagnosis is accomplished via the trained and optimized neural network.The experimental results show that this method gains good classification result,and can identify faults under the difficult situation where transformer overheat and partial discharge coexist.Moreover,this method works with a higher correct judgement rate.