Fault Diagnosis Method Of Transformers Based On Multi-source Information Fusion

As the core device of energy transmission and conversion in power grid,the operation stability and reliability of power transformer play a vital role in the operation safety of the whole power grid.Therefore,it is very important to diagnose transformer fault effectively.Although transformer on-line detection and diagnosis technology has made a large-scale development in recent years,the source of monitoring fault diagnosis data is relatively single,only for a state of transformer research,not to combine the state of many parts of transformer,which has a certain one-sidedness.In view of the above problems,this paper mainly does the following work:This paper reveals the relationship between the status of insulating oil,the main insulating component of transformer,and transformer faults,and analyses the mechanism of characteristic products in the aging and deterioration process of transformer insulating oil.Based on the content of these characteristic products,a transformer fault diagnosis model based on the status of insulating oil is constructed.Least squares support vector machine(LS-SVM)based on multi-classification is applied to transformer fault diagnosis.By constructing one-to-one fault classification mode,multiple binary LS-SVM classifiers are constructed to realize multi-classification of transformer faults.At the same time,the BAT algorithm is used to optimize the parameters of the classifier,and the classifier with the optimal parameters is obtained.Then the fault diagnosis example is analyzed by collecting the data indicators based on the insulation oil state.The experiment shows that the method is accurate and effective for transformer fault diagnosis,and it also proves that the transformer fault characteristics based on the insulation oil state can be in one.To some extent,it reflects the fault state of transformer.To overcome the shortcomings of the basic bat algorithm,such as local optimum and lack of intelligence in practical optimization,a bat algorithm with Levy flight characteristics is proposed to optimize LS-SVM model.The Levy flight feature is used to replace the bat individual’s search process for the best location.Especially in the global search,the bat generates a larger search range,iteration and matching mode to avoid falling into the local optimum to the greatest extent.The example shows that the improved bat algorithm has faster convergence speed,higher fault diagnosis accuracy and better effect.The relationship between transformer insulation paper status and transformer fault is explored,and the characteristic product of transformer insulation paper aging under external action is analyzed.Based on these characteristic product parameters,a transformer fault diagnosis model based on insulation paper status is constructed.At the same time,the transformer fault diagnosis based on the insulation oil-paper state is constructed by fusing the insulation oil state characteristic index and the insulation paper state characteristic index.The fault diagnosis example shows that the characteristic parameters based on insulation paper state can reflect the fault state of transformer to a certain extent,and the insulation oil paper state index obtained after fusion with the insulation oil state index has a higher accuracy in fault diagnosis.On the basis of collecting a large number of documents,technical standards,guidelines and regulations,expert experience and data generated in the actual operation process of power transformer,this paper puts forward the principle of establishing fault diagnosis model of multi-source information transformer,and according to this principle,extracts the reflection of transformer fault from three aspects: oil chromatographic analysis test,oiling test and electrical winding test.The fault diagnosis framework of transformer with multi-source information is constructed based on the state feature.Through the case diagnosis and comparative analysis of collected multi-source information data,it is concluded that the accuracy rate of transformer fault diagnosis based on multi-source information reaches 96%,and the accuracy rate of high-energy discharge fault diagnosis has been significantly improved.