Font Size: a A A

Study On Blackboard-Style Expert System And Prediction Model Based On Genetic Algorithm For Insulation Fault Of Power Transformer

Posted on:2004-10-13Degree:DoctorType:Dissertation
Country:ChinaCandidate:R J LiaoFull Text:PDF
GTID:1102360095456611Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
A large electric power transformer is one of the key apparatus in the electric power system. Faults of a power transformer may have a great effect on stability of the power system to which the transformer is connected,and may cause aseries loss of power consumers.Therefore,there is great academic and engineering significance to do an earlier research on fault diagnosis technology and fault prediction technology on power transformers.By collecting a large amount of regulations, expert experience and actual fault data of power transformers (fault samples), this paper not only thoroughly studies the back propagation (BP) neural network and fuzzy theory, object oriental technology, blackboard style expert system, the basic principle, realization method of the genetic algorithm and the grey prediction theory, but also establishes the blackboard style expert system for the fault diagnosis of transformer insulation and the fault grey prediction system of transformer based on the genetic algorithm, and achieves some breakthroughs on the theory of the multi-expert cooperation diagnosis for the insulation fault and the fault combination prediction of transformers. It is effective by instance validation. The innovative achievements are concluded as follows:The blackboard style structure is applied to the insulation fault diagnosis expert system of transformer for the first time. The main body structure and reasoning mechanism for this kind of expert system are constructed. Combining with object-oriented technique, the insulation fault diagnosis blackboard style expert system for transformer is set up. It can correctly recognize the faults and approximately locate the fault positions, indicate the severity and its tendency. The validity of this expert system is proved to be effective by diagnosing the actual faults of transformers. The multi-expert cooperation diagnosis technology of the blackboard style expert system is studied, and the flow of the transformer insulation fault diagnosis is put forward in the cooperating diagnosis process. And the synthetically analyzing method is also proposed on condition that the expert opinions are not concordant. The effectiveness of this synthetic analyzing method is proved.The layered and distribution knowledge base is built for the first time by employing object-oriented techniques in the insulation fault diagnosis expert system for transformers, and the function to store expert knowledge in database is realized. Itovercomes the shortcomings of the traditional expert system by using production structures, which are less efficient and mistakes preferable, also improves the practicability and accuracy of the expert system.An improved model of fault grey prediction for transformers is established. The equal interval processing method of the original non-equal interval chromatographic data sequence is adopted, which convert non-equal interval chromatographic data sequence into an equal interval one. Meanwhile, the weakening operator is applied to reconstruct the transformer chromatographic data sequence, which lowers the influence of randomness of original chromatographic data sequence suffering from all kinds of random factors to some extend.The genetic algorithm is introduced to the transformer fault prediction, and the improved grey prediction model of the transformer faults based on the genetic algorithm is founded firstly. The parameters of this model are optimized by genetic algorithm directly, which improves the prediction precision and enlarges the scope of application.Aiming at overcoming the drawbacks of poor adaptability of a single model and the unstable prediction results affected by randomness of transformer chromatographic data, a combination model composed of the power function prediction model and the exponent function prediction model is proposed. By using the genetic algorithm, the weight coefficients of this combination prediction model is optimized and the most optimal coefficients combination model is achieved.Practical examples show t...
Keywords/Search Tags:expert system, multi-expert cooperation, blackboard structure, genetic algorithm, fault prediction, combination prediction
PDF Full Text Request
Related items