多神经网络与证据理论融合的变压器故障综合诊断方法研究

Study on Synthetic Diagnosis Method of Transformer Fault Using Multi-neural Network and Evidence Theory

电力变压器发生故障的部位多,故障原因、现象复杂,在故障诊断中,可以通过变压器不同方面的特征信号从不同侧面来反映变压器的故障。因而需要对变压器的多种特征信号进行综合处理和协同分析。该文结合色谱数据和电气试验数据,利用数据融合原理,将神经网络和证据理论进行有机结合,使两者优势互补,提出了多神经网络与证据理论融合的变压器故障综合诊断方法。诊断结果表明,运用提出的融合诊断算法,能充分利用色谱数据和电气试验数据的冗余、互补信息,使基于多种特征信号综合诊断结果的准确性和可靠性比基于单一故障特征的诊断得到有效的提高。

In the transformer fault diagnosis, the fault can be reflected by different characteristic signal from different side, due to complexity of fault reason and phenomenon of power transformer. Thus the synthetic disposal and cooperative analysis for multi-characteristic signal of transformer are needed. In this paper, a synthetic diagnosis method using multi-neural network and evidence theory for transformer fault diagnosis is presented, combining DGA data and routine electrical tests data, integrating two data fusion methods （ANN and evidence theory） by using their superiority and avoiding their disadvantages, The diagnostic results show accuracy and reliability based on multi-characteristic signal are improved effectively comparing with diagnosis based on single fault characteristic by using information from DGA data and routine electrical tests fully.