Research On Some Key Techniques Of Current-Carrying Fault Precaution And Prediction For Electric Equipment

Due to the large capacities, high voltage levels and long distances of power systems, the safe and stable operation of electric equipment becomes more and more important. The research and applications of the fault diagnosis for electric equipment have drawn increasing attentions. It is asignificant task for us to improve the performance of diagnostic systems, and then systems can be more effectively to find, locate, and analyze faults.This paper is specific to current-carrying faults of electric equipment. On the basis of summing up the present technologies and application situations of fault diagnosis, this paper is committed to fault precaution and prediction, which means that faults can be found and forecasted in their earlier stage. Through the research on mechanism analyzing, system modeling, data mining, and so on, an online diagnosticscheme for electric equipment current-carrying faults has been presented. The overall implementation of the scheme includes data colletion, signal transmission, data analysis, fault precaution, and fault prediction. Application results on a real110kV distribution substation verify the feasiblility and validity of the proposed method.The detailed research work and contribution are shown as follows:(1)Early warning for current-carrying faults:Principal component analysis (PCA), wavelet de-nosising translation method, and K-means clustering algorithm are introduced, and a PCA-based precaution approach is preposed. First, temperature data are de-noised by wavelet de-nosising translation method. Then the principal component’s eigenvalue (PCE) is extracted based on PCA, and PCE is the criterion to judge whether there is a fault. Finally, K-means clustering algorithm is applied to locate fault contacts. Experimental results verified that the PCA-based approach can find faults at their early stage and has better performance compared with another fault detection methods.(2)Modeling for current-carrying faults:The development mechanisms and characteristics of current-carrying faults are analyzed base on heat transfer theory, and a temperature model for fault contact is established to describe the temperature variation. Then, the model parameters estimation can be transformed and solved as a nonlinear equation with one variable. Temperature fittings, temperature interpolation, and temperature predicition can be finished base on the model. The experimental results show that the model fits well with the actual operation situation.(3) Prediction for current-carrying faults: Bayesian fitering, Monte Carlo methods, and particle filtering (PF) are introduced in detail, and a PF-based prediction approach is presented. This approach utilizes the operation of summing instead of integral operation to modify the model parameters, and the modified model is used to predict temperatures of contacts. Each time a sample is obtained, the parametes are optimized to forecast the temperature trend iteratively. Application results show that the prediction and the actual temperature reach a consensus gradually.(4) Numerous experiments were conducted at a real110kV distribution substation to verify the presented scheme, which include online data collection, classification, analysis, early warning, and prediction. Application results demonstate the validity of the preposed system.