Transformer Fault Assessment And Prediction Base On Grey Theory

With economic development of our country, the need of electric power is rapidly growing.It is hard to keep balance of distribution between energy resources and load centre in ourcountry, so it is necessary to construct transmission power system in far distance and largevolume. As core equipment in transmission power system, the state on transformer straightlyinfluences the stability, dependability and security of transmission power system. Lots ofresearch organizations, institutes of science and universities do abroad and deeply research ontransformer.The research consists of fault prediction of transformer, state inspection of transformer,state assessment of transformer and fault diagnoses of transformer. The method of researchhas all kinds of intelligent algorithm, such as neural network, artificial intelligence, geneticvariation, etc. The intelligent algorithm has a good effect on transformer.In terms of grey system theory, author strongly studies its original outstanding feature andfundamental modeling principle, carefully research improved ideology about variousimproved method and root of inaccuracy during proceeding of establishing model. Based onsuch theory research, faced on the problem of great inaccuracy in condition of processingburst data when predicting transformer fault, put forward an improved method. To solve theproblem, that is only decision on category of transformer fault, and scarcity of quantitativeassessment while diagnosing fault of transformer, advance a concept and method of faultassessment, and then found a concept and method of predictive assessment by an applicationon power transformer in a factory.The prime task and innovative method contains:(1) From modeling principle of grey prediction model, find out the root of inaccuracy,which is the method of burst process from discrete format to continuous format. According totrend limit of varying parameter, and compared to trend limit of developing graph,demonstrate discrete prediction model is exact form of grey prediction model bylimited-reasoning, it is a special GM (1,1) model in condition of limit and is able tocompletely imitate exponential function.(2) After deeply researching grey theory and the principle of time-dependent parameters,put forward the improved method of discrete prediction model, that is unfixed initial valueand equal dimension rolling of DGM (1,1) model by applied to burst data in fault predictionof transformer. The effect of its prediction is compared with common unimproved DGM (1,1)model in condition of both burst data and not-burst data. The analysis shows the effect of theimproved DGM (1,1) model is better than unimproved model about burst data. (3) After deeply studying grey target theory and entropy weight theory, entropy weighttheory is introduced into grey theory in order to assess transformer fault. With purpose ofsolving the problem, weights of parameter coming from unstable principle and from acceptedexperience, entropy weight theory is introduced into assessment model to obtain scientificweight of parameter by rule of distributed data. So as to overcome difficulty, that is onlydecision on category of transformer fault, and unable to measure the degree of fault, choosegrey target theory as modeling theory of assessment, choose target centre degree as evaluatingquantitatively degree of fault, and avoid only making the decision on category of fault whendiagnosing fault of transformer. Then establish new assessment method, which is able toassess quantitatively degree of transformer fault and classify fault of transformer. The183groups of fault data on transformer are compared with the description of the fault, the result ofanalysis indicates that the method is effectively capable of assessing fault of powertransformer.(4) As an application of transformer fault in a factory, and combination with above thedemonstrated method of fault prediction and fault assessment, advance a concept of predictiveassessment and explicitly describe the method of predictive assessment. The result ofapplication analysis illustrates that the method of predictive assessment is effectively capableof assessing fault of power transformer.