使用多参量的变压器故障综合诊断技术

为全面综合诊断电力变压器故障,参考已有变压器故障综合诊断方法,结合变压器油中溶解气体数据和电力试验数据,利用自适应遗传算法优化小波神经网络和证据理论融合技术,提出了一种基于多参量的电力变压器故障综合诊断模型。通过故障特征参数的划分分别构建神经网络从不同侧面反映变压器的故障,同时结合证据的重要性、神经网络的输出改进证据体的基本概率分配赋值,充分体现证据体对单个故障模式识别的可信度。诊断结果表明,基于信息融合技术的变压器多参量故障综合诊断比基于单参量故障诊断的诊断性能较好。

Method of Fault Diagnosis for Power Transformer with Multi-parameters

According to the shortcomings of existing transformer fault comprehensive diagnosis methods,combining the data of dissolved gas in oil and the off-line electrical test data,one kind of transformer fault synthetic diagnosis model based on wavelet neural network(WNN) optimized by adaptive genetic algorithm(AGA) and D-S evidence theory fusion technique is proposed in this paper for completely synthetically diagnosing the transformer fault.According to the divisions of fault characteristic parameters,multi-neural networks are constructed to reflect the transformer faults from different aspects.The basic probability assignment(BPA) of evidence body is constructed combining the evidence importance and the outputs of neutral network in order to sufficiently display the believable value between the evidence bodies and the single fault model recognition.The diagnostic results indicate that the fault synthesis diagnostic method based on information fusion and multi-parameters has better diagnostic performance than method based on single-parameter,and a new idea is provided to diagnose other high reliability devices.