基于信息融合的大型油浸电力变压器故障诊断

SYNTHETIC INSULATION FAULT DIAGNOSTIC MODEL OF OIL-IMMERSED POWER TRANSFORMERS UTILIZING INFORMATION FUSION

摘要: 由于大型电力变压器具有互补性、冗余性和较强的不确定性等特点,该文将信息融合的基本思想引入到变压器的故障诊断中。在信息融合的基本框架下,利用反向传播人工神经网络和证据推理技术,建立了一种新型的油浸式电力变压器故障综合诊断的多级决策融合模型。该模型将油中溶解气体分析与常规电气试验的结论紧密结合起来,并充分借鉴现场的运行、诊断和维修经验,具有较强的知识表示及不确定性处理能力。

As the fault information of large power transformers has characteristics such as, complementarity, redundancy and uncertainty , the basic ideas of information fusion are introduced in this paper. In accordance with the basic principles of information fusion, a new type of multi-level comprehensive fault decision model is proposed, using back-propagation artificial neural networks and the technique of evidence reasoning. Within the diagnostic model, the dissolved gas-in-oil analysis (DGA) and the results of conventional electrical tests of power transformers are combined tightly. Also, the on-site experiences in operation, diagnosis and maintenance are highly utilized in the model. It has shown that the model possesses satisfactory capacity of knowledge representation and strong solving ability to deal with uncertain facts.