PSO BASED KERNEL PRINCIPAL COMPONENT ANALYSIS AND MULTI-CLASS SUPPORT VECTOR MACHINE FOR POWER QUALITY PROBLEM CLASSIFICATION

Electric power quality (PQ) problems are very important aspects due to the increase in the number of loads which are sensitive to power disturbances. One of the important issues in the PQ problems is to detect and classify disturbance waveforms automatically in an efficient approach, because the possible solutions can be determined after the disturbance types are detected. This paper proposes a particle swarm optimization (PSO) based kernel principal component analysis (KPCA) and support vector machine (SVM) for PQ problem classification. Wavelet based multiresolution analysis (MRA) is utilized to extract features for various PQ disturbances. Dimension of these features are then reduced by KPCA so that the noise has less impact on the classification results. The multi-class SVM is used to classify the PQ problem using the dominant KPCA. The PSO is applied to optimize the KPCA and SVM parameters in order to improve the classification performance. The classification process implemented with various PQ events shows that the proposed technique provides more accuracy than the conventional technique under both noisy and noiseless environments.