SVM-based series arc detection algorithm for photovoltaic system

In this paper, we present a SVM-based series arc detection algorithm for photovoltaic system. SVM is a technique of extracting a hyperplane that maximizes the margin between series arc fault samples and normal state samples from training dataset and arc noise can be effectively classified using the extracted hyperplane. As input parameters for SVM, various DWT-based feature parameters are used. The hyperplane for determining the class is obtained on initial step, and this is applied to the arc detection algorithm for SVM testing. The real-time arc detection test is performed through TMS320f28335 DSP. It is verified that not only the series arc can be successfully detected with SVM-based algorithm, but also the detection time is satisfied the required detection time of the UL1699B safety standards.