Application of Prediction Accuracy Interpolation Method Based on Support Vector Machine Optimization in Transformer Oil Temperature Preprocessing

The preprocessing method of transformer oil temperature directly affects the prediction accuracy of transformer oil temperature as well as the assessment on the inner thermal state within the transformer. Based on support vector machine (SVM) optimization and the principle of linear interpolation, a method to improve the preprocessing accuracy of transformer oil temperature was proposed. Firstly, the SVM was utilized to overall distinguish the data of transformer oil temperature to determine the approximate range of the point with data exception. Secondly, a calculation model based on weighted optimized linear interpolation was established and the threshold, by which the point with abnormal data of transformer oil temperature could be determined, was set. Thirdly, an accurate discriminant to determine both positions of points with abnormal data of transformer oil temperature and the amount was given. Finally, on the basis of the principle of correlation weighting and the bell shaped function, the smoothing processing of transformer oil temperature was performed. The effectiveness and feasibility of the proposed method are verified by simulation results.