Application Of Dissolved Gas Concentration Prediction In Insulating Oil Of Zhaoqing Power Supply Bureau

In the power supply network,transformer is the foundation to ensure the stable operation of the entire power system,and is also one of the most critical and core components in the power network.Because the power transformer has the characteristics of complex structure,great technical difficulty,expensive equipment and large core function,the impact of transformer failure in the power system will be very great.In order to avoid transformer tripping and power failure accidents and discover potential transformer faults in advance,it is particularly important to predict transformer fault types and risks.Therefore,it is imperative to carry out research on transformer fault detection and prediction.Based on the analysis of dissolved gases in insulating oil and the research status of transformer fault diagnosis,this paper analyzes the gas generation principle and dissolution principle of dissolved gases in insulating oil,and sorts out the relationship between dissolved gases content under normal working conditions and fault conditions of transformers,thus determining the "gold standard" for transformer fault prediction and diagnosis.Then,the analysis method of dissolved gas in transformer insulating oil is introduced,and the common methods for predicting dissolved gas concentration are analyzed.On the basis of mutual information variable selection theory,the feasibility of mutual information feature selection(MIFS)and normalized mutual information feature selection(NMIFS)in dissolved gas concentration prediction is studied,and an improved NMIFS algorithm for dissolved gas concentration prediction of insulating oil is proposed.The rationality of the algorithm is verified by the detection data of transformer insulating oil dissolved gas in Zhaoqing Power Supply Bureau.Through combing the transformer fault diagnosis theory and combining the theory of Multi-Kernel Relevance Vector Machine(MK-RVM),a transformer fault prediction diagnosis model is proposed.The model uses principal component analysis(PCA)to extract principal components,then reconstructs the prediction sample matrix,and uses kernel function to optimize MK-RVM,thus obtaining the best MK-RVM prediction model.Finally,combined with three transformer fault cases in Zhaoqing Power Supply Bureau,the rationality and feasibility of the prediction method of dissolved gas concentration in insulating oil and transformer fault prediction and diagnosis method proposed in this paper are explored.The case application and analysis show that the dissolved gas concentration prediction method based on the improved NMIFS algorithm and the transformer fault prediction method based on MK-RVM have better adaptability and accuracy.