ABSTRACT

The electrical power quality (EPQ) analysis has become an exponentially increasing field of interest, gaining the attention of researchers, electrical utilities and customers particularly in the past few decades. It is well known that the variation in electric power quality disturbances usually in a wide range of joint time and frequency, therefore automatic detection of PQ problems is highly desirable, and it is very difficult and elusive to be diagnosing these disturbances with conventional approaches of signal processing techniques.
The electrical power quality EPQDs) are detected and classified by
using joint time-frequency analysis technique of discrete wavelet transform (DWT) and artificial neural network (ANN) is proposed in this work. The distorted waveforms of PQD signals are generated using Matlab, based on the parametric equations and controlling parameters described by IEEE 1159-2009 standards. The distorted signals of electrical PQ are decomposed by using discrete wavelet transform (DWT) and to select its useful information as feature extraction (FE). For the classification of PQDs appropriate feature vectors are selected and applied to train the probabilistic neural network (PNN) as classifier. Comparison of test results of proposed technique with those generated by other existing methods, show enhanced performance and accuracy.


Keywords: Power Quality Disturbances (PQDs), Discrete Wavelet Transform (DWT), Energy distribution (ED), and Probabilistic Neural Network (PNN).