Fault Prediction Of Transformer Based On Dissolved Gas Analysis

Transformer is one of the most important equipment in the whole power system.Its importance is self-evident.The power transformer is the most expensive and complex electrical equipment in the entire power system.Once an accident occurs,the direct and indirect economic losses caused are enormous.Therefore,fault prediction study of power transformers is particularly necessary.To a certain extent,various faults can be discovered in advance,which will help electric power system personnel to prepare as early as possible,and the faults that grow in the early stages will be eradicated as early as possible to avoid transformer failures and ensure the development of the national economy.Based on the actual situation and characteristics of China's power system and the actual online monitoring data,this paper establishes a new transformer fault prediction model.The principle of generating dissolved gas in transformer oil,the source of gas production,the content standard,and the corresponding relationship with the specific fault type inside the transformer are studied.The main and secondary gas components corresponding to different fault types of the transformer are summarized.The cause of the failure is briefly analyzed.Combining four common single prediction methods(artificial neural network prediction,time series prediction,grey prediction,and regression prediction),introducing an induced ordered weighted average operator and Markov theory,a new transformer oil dissolved gas prediction model is established.Based on the data that has been mastered by the laboratory to verify the model,the example analysis shows that the model is superior to the four-single prediction methods and compared with the methods in related literatures to prove the accuracy advantage of this prediction model,that is,the model has certain significance for the prediction accuracy improvement.Based on principal component analysis and cost-sensitive combination kernel correlation vector machine,a new transformer fault diagnosis model is constructed.Based on the dissolved gases in seven common oils,a new set of variables is formed by combining the three-ratio method for data reconstruction,and the principal component analysis of the reconstructed variable set is performed.Eight key features are used as model inputs to establish the cost-sensitive mechanism.The combined vector machine with optimized kernel function parameters is introduced,and finally constitutes a fault diagnosis model.Combined with the existing data samples for case analysis and comparison,the results of the table name have a higher accuracy of fault diagnosis accuracy than other methods.Combining the predictive model and the diagnostic model to form a transformer fault prediction model,and introducing model evaluation parameters at the same time,through an example analysis that the predictive model has a certain fault prediction accuracy,may consider guiding the actual transformer operation and maintenance work.