Feature selection and accurate classification of single and multiple power quality events

In this paper an attempt has been made to classify the power quality disturbances more accurately. Wavelet Transform (WT) has been used to extract the useful features of the power system disturbance signal and optimal feature set is selected using Fuzzified Discrete Harmony Search (FDHS) to classify the PQ disturbances. Support Vector Machine (SVM) has been used to classify the disturbances. FDHS is used both for parameter selection of SVM and, feature dimensionality reduction to achieve high classification accuracy. Six types of PQ disturbances have been considered and simulations have been carried out which show that the combination of feature extraction by WT followed by feature dimension reduction and parameter selection of Gaussian kernel using FDHS increases the testing accuracy of SVM.