Classification of power quality events using SVM networks

This paper presents a new approach to classification power disturbances. Firstly, the growing importance of automated power quality assessment is shown. The so called support vector machines (SVM) networks are regarded as an effective tool for loosing classification problems and their architecture is described in detail. A 15 kV radial distribution network was investigated, in which one branch suffered voltage collapses due to short circuits. SVM network was implied to classify the different dips on other branches originated in the short-circuited branch. The influence of line length, type and duration of short circuit on classification rate is investigated.