Fault Diagnosis Of Oil-immersed Power Transformer Based On Intelligent Information Fusion

The electric power relating to the national economy is keeping on development. Power Transformer is one of the most important electrical equipments in the electric system. Dissolved Gas-in-oil Analysis (DGA) is an important method to diagnose internal fault of transformer and it offers a main basis to find the general incipient faults of the transformer indirectly. Therefore International Electro technical Commission recommends Three Ratios Method. But in practice, the uncertainty existing in the diagnosis could not be eliminated, due to the ambiguity of the inference and insufficient standard for judgment via the above means. So how to use the DGA data to find the transformer faults timely is the recently research focus. Therefore, the methods for the diagnosis of potential faults concealed inside power transformers have attracted many researchers' interest, such as: Expert System, Fuzzy Theory, Artificial Neural Network (ANN), Information Fusion, etc. In this paper, the author use different mathematics method to study the relation between DGA data and fault type.There is complicated non-linear relations beween fault symptom space and faults pace, the mathematical model of diagnosis system is difficult to build up. However, ANN provides a new way for this problem due to its advantages such as parallel processing, self-adaptation self-study, association memory, non-linear mapping, etc. Thereby, Back-propagation Neural Network (BPNN) is applied to fault diagnosis system in the paper. 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. Further more there is some inherent defect in traditional single artificial intelligent fault diagnosis method.In the paper Fuzzy Neural Network (FNN) method is constructed based on Fuzzy Theory and ANN aiming at the shortcomings of BPNN. This network has combined advantage and fuzzy reasoning method. It has overcome the problem of difficult to determine fuzzy regulation in transformer fault diagnosis. The results of verification show that it improves the shortcomings of BPNN such as slow convergence and local minima, system, which can effectively overcome the shortcomings of traditional diagnose methods that is making a mistake or defective result.On the basis of information fusion, a synthetic diagnosis model for power transformer is proposed based on TNFIN and Dempster-Shafer's evidence theory. The DGA is used as dominant idea to diagnose internal incipient fault in this model. Synthetical diagnosis on the basis of combining all kinds of relevant tests makes a decision for the fault position; improves the reliability and accuracy of fault diagnosis.