Power quality disturbance analysis using data-driven EMD-SVM hybrid approach

In the modern electrical power system network (EPSN), the power quality disturbances (PSDs) are the serious issue for the power engineer to maintain the uninterrupted and reliable power supply. Generally, PQDs are generated due to non-linear loading condition, perturb loading and other occurrences such as transient, harmonics, sag, swell and interruptions. These problems of PQDs effect the power demand mapping problem, which effect the reliability and stability of the EPSN operating condition. In this study, a novel approach for PQDs diagnosis (PQDD) is proposed, which includes real-time data generation, data pre-processing, feature extraction, feature selection, intelligent model development for PQDD. Data decomposition approach of EMD is utilized to generate the feature vector of IMFs. These features are utilized as an input variable to the intelligent classifiers. In this study PQDD is analyzed based on SVM method and obtained results are compared with conventional AI method of LVQ-NN. The results represent the higher acceptability of the proposed approach with diagnosis accuracy of 99.98% (training phase), 93.11% (testing phase) for SVM and 92.56% (training phase) and 91.0% (testing phase) for LVQ-NN based PQDD method.