A Study On Transformer Online Detecting Hybrid Intelligent Expert System

The problem discussed in the dissertation is concerned about present power transformer faults diagnosis methods analysis and improvement. Owing to the multiplicity of transformer's types and fault characters, causation of fault is getting more complex and includes lots of redundancy. All of above determined that the study of transformer fault diagnosis methods is an arduous and complex task.Dissolved gas oil (DGA) is an effective way in area of charge oil electrical equipment fault diagnosis. It can precisely forecast the tendency of fault development as well. Now, become a widely spread method in online detection field in the world. IEC 3 ratio method is the most popular technique in DGA based methods. Though it simple and convenient but less efficiency. Artificial neural network is applied in various faults diagnosis area and achieved correspond progress in the study of transformer fault diagnosis. There is a key problem how to make the network getting generalize enough, that means avoid network tend to convergence on a local optimal point。Furthermore there is some inherent defect in traditional single artificial intelligent fault diagnosis method.Dissertation applies immune genetic algorithm in training neural network based on the present expert system and artificial neural network after compared several optimize neural network methods. Finishes the program under MATLAB IDE after analyze the genetic algorithm's principal of global search ability. According to the defects of genetic algorithm so bring immune mechanism into program make the algorithm more practical and then detail on the procedure of modeling and simulating. Dissertation has validated the efficiency and reliability of method which is applied in the dissertation after compared diagnosis result of IEC 3 ratio method, neural network trained by genetic algorithm and immune genetic algorithm. Hybrid intelligent method could learn from others strong points to offset one's weakness through integrate different methods. It is better than original one. Focus on the hybrid intelligent method application of transformer fault diagnosis. The study of hybrid intelligent expert system based on integration of artificial neural network and expert system was carry on in the dissertation. Reviewed several integration models of artificial neural network and expert system and established an effective model. Ultimately introduced the model design process and realize steps in detail. Example simulating result demonstrate the hybrid system's function and effect.