摘要:变压器是电力系统厂站中的非常重要的电气设备。它是电能传输和配送过程中能量转换、传输的核心，是国民经济各行各业和千家万户能量来源的必经之路。因此，能及时地、准确地对变压器各种异常状态或故障状态做出诊断，对保证整个电网能可靠、安全、有效的运行具有十分重要的意义。

　　　现有的变压器故障诊断方法中，三比值法、神经网络算法是常用手段，并在一定程度上取得了显著效果。由于变压器故障征兆的繁杂和故障类型的多样，仅仅利用单一的诊断方法难于全面确定故障性质和可能存在的故障。因此，针对变压器故障信息的不完整、不精确、模糊等特点，引入一种不确定性推理的证据理论。即以证据理论为基础，采用了一种以BP神经网络初步诊断和证据理论融合决策诊断相结合的故障诊断方法，并通过实例进行验证分析。结果表明，该方法可明显提高变压器故障诊断的可信度，降低其不确定性。

关键词：变压器；故障诊断；三比值法；证据理论；BP神经网路

ABSTRACT:The transformer is power system is very important in the factory station electrical equipment. It is in power transmission and distribution process energy conversion, the core of the transmission, is the national economy all walks of life and homes sources of energy the only road. Therefore, can timely, accurately to transformer various abnormal state or fault state make diagnosis, to guarantee the power grid can be reliable, safe and effective operation has the very vital significance. 

　　Current transformer fault diagnosis methods, three ratio method, neural network algorithm is commonly used method, and to a certain extent achieved significant results. Because of transformer fault symptoms multifarious and fault type of diversity, just use a single diagnosis approach is difficult to determine the fault comprehensive nature and possible failure. Therefore, in view of the transformer fault information is not complete, not accurate, fuzzy and other characteristics, introducing a uncertainty reasoning evidence theory. That is based on the theory of the evidence, adopted a preliminary diagnosis on BP neural network theory and evidence fusion decision with the combination of diagnosis fault diagnosis method, and is verified by practical examples analysis. The results show that this method can obviously improve the credibility of transformer fault diagnosis, reduce the uncertainty. 

Key Words: Transformer; Fault diagnosis; Three ratio method; Evidence theory; BP neural network