Boosting performance of power quality event identification with KL Divergence measure and standard deviation

Power quality event identification is widely recognized as one of the most interesting problems in electric engineering. It consists of two sub-problems: detection and classification. In the first step, a recognition algorithm is used to detect disturbance from power quality events. The next step classifies them into some groups by a machine learning method. In order to enhance the accuracy, a detection technique is required for classifying events in timely manner. In this paper, KL Divergence and Standard deviation are used within Support Vector Machine to detect and classify events. Experimental results with 12 events suggest a specific order of harmonic present in each event. KL Divergence and Standard deviation are obtained for voltage sag of 500 values and for harmonics with swell of 800 values. After calculating KL Divergence and Standard deviation, events are detected with more accuracy. The comparison shows that the new method achieves 94.02% of accuracy which is better than 92.33% of Abdoos et al. (2016), 93.47% of Ma et al. (2017), 89.92% of Li et al. (2016) and 93.87% of Kapoor et al. (2018).