The application of multiclass support vector machine in power transformer fault diagnosis
Sun Zhipeng1, Cui Qing2, Zhang Zhilei2, Wang Tao2, Zhang Tianwei3
1. College of Electrical Engineering, Northeast Electric Power University, Jilin, Jilin 132012; 2. Hebei Electric Power Company Shijiazhuang Power Supply Company, Shijiazhuang 050051; 3. Beijing Runwei Tianhua Power Technology Co., Ltd, Beijing 102211
Abstract：The support vector machine is a new general learning method based on statistical learning theory, which better solves the learning problems of small samples, high dimensionality and nonlinearity. The performance of SVM in classification depends largely on the selection of kernel functions and kernel parameters. At present, the commonly used parameter optimization methods are grid search method, genetic algorithm and particle swarm optimization algorithm. The concentration of the five characteristic gases dissolved in transformer oil are the inputs of support vector machine, the five states of the transformer are the outputs. In the built model the radial based kernel is selected, the optimized parameters are obtained by grid search. The experiment shows that the model enables to detect transformer faults with a higher diagnosis rate under condition of small samples. The diagnosis rate for fault samples gets 83.3%, which proves its effectiveness and practicality.
孙志鹏, 崔青, 张志磊, 王涛, 张天伟. 多分类支持向量机在电力变压器故障诊断中的应用[J]. 电气技术, 2019, 20(10): 25-28.
Sun Zhipeng, Cui Qing, Zhang Zhilei, Wang Tao, Zhang Tianwei. The application of multiclass support vector machine in power transformer fault diagnosis. Electrical Engineering, 2019, 20(10): 25-28.
2019, Vol. 20 
