粗糙模糊神经网络在变压器诊断中的应用

Research on Application of Rough Sets and Fuzzy Neural Network in Power Transformer Fault Diagnosis

针对变压器故障的复杂性、模糊性以及模糊集理论、神经网络和粗糙集理论的优缺点,利用粗糙集理论的属性约简和规则生成能力和模糊神经网络在模式识别方面具有容错和分类优势。采用粗集理论对采集到的变压器油中溶解气体数据形成的规则进行约简处理,建立精简的规则集,根据规则集建变压器故障诊断的神经网络模型,采用自适应遗传算法优化神经网络连接的权值,通过仿真验证了该网络较好的诊断性能。

Considering the complexity and Ambiguity of transformer fauh diagnosis and the advantages and disadvantages of fuzzy set theory, neural network and of rough set theory, this article makes fully using the attribute reduction and rule generation of rough sets and pattern recognition and fault-tolerant Category advantages of fuzzy neural network. Using rough set theory dealing the reduction rules from collected data about gas dissolved in oil into a streamlined set of rules, transformer fault diagnosis model of the neural net- work is built in accordance with the Streamlining rules set, connection weights of neural network are optimized by genetic algorithm. At last the method is proved having superior diagnostic performance.