Research On GIS Substation Primary Equipment Condition Assessment Based On Support Vector Machine

With the continuous development of power system, the voltage levels of power system increase stably, power system has an increasing requirement for stability. How the power equipment runs directly relates to reliability of the entire power system. Timely evaluation on power equipment is the preparing work of discovering equipment faults, which significantly prevents accidents from expanding. As the ultimate equipment in power system, the transformer condition assessment is of great importance. Gas Insulated Switchgear (GIS) has features as small land-cover, high reliability etc, which is widely used in high voltage rating system. It is essential to ensure the work stability of GIS.This paper analyses the transformer operation states in corresponding relationship with oil dissolved gas, and propose an evaluation method which bases on the oil dissolved gas data. Select the dissolved gases H₂,CH₄,C₂H₆,C₂H₄,C₂H₂ gas content as the index of evaluation, and divided transformer condition into 5 levels. Proposed an evaluation model which was based on improved PSO optimized SVM parameters (PSO-SVM).The results of actual transformer data test show that transformer condition assessment model obtained by PSO-SVM has higher classification accuracy and assessment of accuracy.The article describes several common failures and the corresponding defects features in GIS running state. Extract 7 time domain parameters from discharge signal as evaluation characteristic quantities.On the base of PSO-SVM, propose a strategy in using of AdaBoost to enhance SVM model (AdaBoost-SVM). According to obtained ultra-high frequency (UHF) partial discharge detection signal, divided GIS operation state into Normal state and Fault state, Normal state is divided into Excellent state and Attention state, and Fault state is divided into five species.According to the actual operation of GIS data, assessment result show that, the AdaBoost-SVM model can play full potential of classification. The final classifier has stronger performance of classification, higher classification accuracy, and can be applied in GIS state assessment.