Power quality diagnosis technique using continuous s-transform based analysis of variance and support vector machine / Kamarulazhar Daud

Power quality (PQ) has been an important issue in power systems in recent years. The demand for clean power has been increasing in the past several years. The reason is mainly due to the increasing usage of microelectronic processors in various types of equipment such as computer terminals, programmable logic controller, diagnostic systems, etc. Poor power quality may cause many problems for affected loads such as malfunction, instabilities, short lifetime, and so on. Poor quality is attributed due to the various power line disturbances like voltage sag, swell, impulse, and oscillatory transients, multiple notches, momentary interruptions, harmonics, and voltage flicker, etc. The objective of this work is to detect and classify the power quality disturbances (PQDs) that use the advances in signal processing and classification intelligence. In this research, a new application method of feature extraction, detection and classification the pattern or waveform shape of PQDs has been introduced. In the first stage, voltage waveforms with multiple power disturbances were simulated using the PSCAD and MATLAB software. Multi disturbance types generated from Power System Aided Design (PSCAD) and MATLAB software simulation were utilized for feature extraction purpose by deploying the Continuous S-Transform (CST). The feature selection and detection approach of feature extraction for these disturbances were employed by one and half-cycle Windowing Technique (WT) and Analysis of Variance (ANOVA). Then, the classification method for a different type of power disturbances was developed using the Neural Network (NN) including Probabilistic Neural Network (PNN), and Support Vector Machine (SVM). Consequently, the One-Cycle Windowing Technique (OCWT) was used to provide the smooth detection on PQ disturbances compared to Half-Cycle Windowing Technique (HCWT) in order to get better accuracy of classification. The significance results from OCWT produced good recognition rate as compared to HCWT. Finally, a comparison of classification using PNN and SVM were compared for the PQ diagnosis system to verify the accuracy rate of classification performance for voltage sag, swell, and transient. In overall, the significance result of SVM classifier was found more efficient as compared to PNN.