Fault Diagnosis Of Power Transformer Using Integrated Artificial Intelligence

The applications of hybrid algorithm based on GA-BP and rough set theory in the fault diagnosis of power transformer are studied in this paper.The application of hybrid algorithm which combines improved genetic algorithm and error back-propagation algorithm in artificial neural network training is studied first. As the back-propagation neural network(BPNN) is adopted to diagnose the dissolved gases of transformer, convergence rate becomes slow, convergence precision becomes bad and even being out of convergence as more sampled data are trained and more complicated relation between input and output becomes. This paper presents a improved hybrid algorithm based on GA-BP which makes good use of searching virtue in overall range using genetic algorithm and great capability of searching in local range using error back-propagation algorithm. This hybrid algorithm is applied to transformer fault diagnosis. The superior efficiency of the method has been verified by the simulation results of practical examples.Conventional IEC three ratio method and new IEC ratio criterion are reduced according to the method of decision table in rough set theory respectively, simplified diagnosis rules are proposed. Besides being identical with original IEC method, the codes of conventional IEC three ratio method are increased and bounds of new IEC ratio criterion are augment by using this method, codes imperfectness of conventional IEC three ratio method and code name absence of new IEC ratio criterion are improved to some extent, scarcity of codes in conventional IEC three ratio method is offseted, absolutization of sort and boundary in new IEC ratio criterion is overcomed. This is important in practice for its flexibility and enhanced error tolerances. Fault diagnosis capability of IEC is enhanced. Synthetic diagnosis for power transformer using rough set theory is studied in this paper. A method of synthetic diagnosis for power transformer is put forward, in which the effective information from electrical tests and analysis result of the dissolved gas in the oil can be fully utilized. One fault probably cause several omens,but one omen probably caused by several faults, it is frequent that there is several faults in a transformer at the same time. Confirmation of fault position based on all sorts of fault omens can provide great significative guidance to repair people. Fault diagnosis can be described a matter of pattern recognition, which can be resolved by the approach of decision table in rough set theory, this method can be deal with imperfection information through avoiding omitted and wrong information. The accurate diagnosis results can be obtained directly from the set of complete fault samples, and the satisfactory diagnosis results can also be got from the set of incomplete fault samples by using the approach, moreover, it can deal with compound faults.