【作者】 杜剑光；

【导师】 粟福珩；

【作者基本信息】 华北电力大学（北京） ， 高电压与绝缘技术， 2003， 硕士

【摘要】 电力变压器故障诊断对电力系统的安全运行有着十分重要的意义。本文在广泛收集变压器故障诊断专家知识和经验的基础上，讨论变压器故障原因、种类以及检测与判断流程，总结现有的诊断方法，提出了基于信息融合的多层分布式推理机制的变压器故障诊断专家系统的思路与实现方法。该系统以国内外公认的油中溶解气体分析作为诊断内部潜伏性故障的主线，结合各种相应的试验项目（预防性试验、油务试验等）进行综合诊断，确定故障的大体部位。专家系统知识库包括：由针对各类试验项目组成的基本规则与复合规则；隐含分布于组合混沌神经网络中的诊断规则；各类变压器故障实例组成的源范例。采取正向推理与混沌神经网络模型相结合并与范例推理相互验证的推理机制弥补了单一方式的不足。本文深入研究了人工神经网络在变压器故障诊断中的应用，对目前最常采用的BP算法做出了较大改进，在原有算法中引入混沌动力学原理构造了混沌神经网络BP算法，并提出了训练样本集的预处理方法，在仿真训练中取得了满意的效果。结合菏泽电业局具体软件工程项目，开发了基于B/S架构的软件应用系统，达到了变压器远程诊断与状态检修的目的。更多还原

【Abstract】 Power transformer fault diagnosis is vital to power system safe operating. After discussing transformer fault cause, fault type and diagnostic procedure, this paper summarizes existing diagnosis method and puts forward idea of transformer fault diagnosis expert system based on multilayer distributed reasoning mechanism on the basis of collecting extensively expert knowledge and experience. The dissolve gas analysis method which adopted widely at home and abroad is used as dominant idea to diagnose internal incipient fault. Synthetical diagnosis on the basis of combining all kinds of relevant tests make a decision for the fault position. Tests include routine test for oil and preventive test for transformer. The knowledge base includes diversified rules, such as basic and complex rule composed of vary tests item, diagnosis rule that distributed impliedly over combined chaos neural network and the transformer fault case. It supply a gap of single way to adopt the reasoning mechanism cooperated with case-based reasoning and the combination of positive reasoning and chaos neural network model. After researching the application of artificial neural network to fault diagnosis of transformer, the back propagation algorithm is improved observably. Inducting chaos dynamics principium to create chaos neural network back propagation algorithm and using the pretreatment of training sample, the emulation program show that this method is effective. Combining the software engineering project of Heze power supply bureau, the system software is developed which based on browser/server structure. A condition based maintenance transformer fault diagnosis expert system software on internet is presented.更多还原