An Efficient DWT and EWT Feature Extraction Methods for Classification of Real Data PQ Disturbances

Authors : Mehmet İsmail GURSOY, Seydi Vakkas USTUN, Ahmet Serdar YİLMAZ

Pages : 158-171

Doi:10.29137/umagd.350231

View : 29 | Download : 7

Publication Date : 2017-01-29

Article Type : Research Paper

Abstract :Determination and investigation of incidents affecting Power Quality insert ignore into journalissuearticles values(PQ); is very important for consumers. In this study, estimation of PQ events is obtained to determine the disturbances of PQ by using Empirical Wavelet Transform insert ignore into journalissuearticles values(EWT); and Discrete Wavelet Transform insert ignore into journalissuearticles values(DWT); methods and with this estimated parameters. PQ disturbances were examined with Support Vector Machine insert ignore into journalissuearticles values(SVM);, Artificial Neural Network insert ignore into journalissuearticles values(ANN); and Adaptive Neuro-Fuzzy Inference System insert ignore into journalissuearticles values(ANFIS); classification methods. Voltage signals insert ignore into journalissuearticles values(sag, swell, interruption, transient and normal); used in the classification of PQ disturbances were recorded from grid with the aid of a microcontroller based on device designed with a sampling frequency of 6.4 kHz. Classification consequences using Machine Learning Methods show that DWT outperforms over EWT for feature extraction processing and the classification accuracy is tabled.  Classification by ANN and ANFIS through the use of conjecture parameters in PQ disturbances based on DWT Method has been recommended.
Keywords : Power Quality, Discrete Wavelet Transform, Empirical Wavelet Transform, Support Vector Machine, Artificial Neural Networks, Adaptive Neuro Fuzzy Inference System