A novel differential particle swarm optimization for parameter selection of support vector machines for monitoring metal-oxide surge arrester conditions

Since metal-oxide surge arresters are the important over-voltage protection equipments used in power systems, their operating conditions must be monitored on a timely basis to give an alarm as soon as possible in order to increase the reliability of a power system. The paper proposes a novel differential particle swarm optimization-based (DPSO-based) support vector machine (SVM) classifier for the purpose of monitoring the surge arrester conditions. A DPSO-based technique is investigated to give better results, which optimizes the parameters of SVM classifiers. Three features are extracted as input vectors for evaluating five arrester conditions, including normal (N), pre-fault (A), tracking (T), abnormal (U) and degradation (D). Meanwhile, a comparative study of fault diagnosis is carried out by using a DPSO-based ANN classifier. The results obtained using the proposed method are compared to those obtained using genetic algorithm (GA) and particle swarm optimization (PSO). The experiments using an actual dataset will expectably show the superiority of the proposed approach in improving the performance of the classifiers.