Research On The Predicting Algorithm Of Power Transformer's Fault

As the most important substation equipment, power transformer plays an important role in the power system. Its states during operation directly affect the stability, safety and reliability of the whole power system. Therefore, effective fault prediction on the power transformer can provide maintenance personnel with the necessary reference to avoid under-maintenance or excessive maintenance, and can guarantee the power supply's reliability and user's safety as well. How to improve the predicting accuracy is the key point but also the difficulty and many researchers have done some research and exploration in this respect. Due to the simple structure and easy to modeling, BP network is preferred. Meanwhile, De Julong, the famous professor in the field of grey algorithm at home and abroad, proposed GM algorithm with broad applied prospects so widely used in population forecast, aerospace camera's forecast, props'durability prediction and other fields. Based on Extensive literature review, this paper put forward some improved methods and obtained a series of related predicting model through in-depth study on GM algorithm and BP algorithm. By means of simulation and comparison, the paper verified the correctness of these theories, the effectiveness of the method and practical of these prediction models.The major work of this paper includes:(1) Summarized the existing predicting technique of transformer fault.(2) In-depth studied GM and BP algorithm.(3) Improved GM's predicting model of the transformer's fault and compared the results with the existing improvement techniques, and realized combinatorial optimization of the improved models, named "Internal Combination"of GM.(4) Improved BP's prediction model of the transformer's fault and introduced the genetic to optimize the BP, and in the light of shortcomings of genetic algorithm itself, niche technique, fuzzy control theory and artificial immune, etc. are introduced to realize the further optimization, and realized combinatorial optimization of the improved models, named "Internal Combination"of BP .(5) Set up combinatorial model between the improved GM model and BP model, that is"External Combination"(6) Put forward"Multi-grade Combinatorial Optimization"based on GM and BP.(7) Summarized the achievements of the research and put forward some further studying directions.