Identification of Voltage Transients and Notches in a Distribution System Using True RMS and Naïve Bayes Classifier

Frequent utilization of non-linear loads and power electronics devices (semiconductor devices) cause adverse power quality (PQ) disturbances due to this both utility and end users struggle for better power quality. Earlier diagnosis of PQ was time-consuming, currently, advanced monitoring devices and techniques are involved. In this paper, the Naïve Bayes Gaussian classifier and true RMS are discussed for the accurate identification of transients and notches. True RMS calculates the square of the instantaneousness value of input and then takes the averaging of signal and displays the square root of instantaneous value. True RMS perfectly monitors sinusoidal and non-sinusoidal waveforms. Using true RMS waveforms, feature extraction has been made for single-stage disturbances. Provided features further process with Naïve Bays Gaussian classifier. The Naïve Bayes algorithm assumes that one feature is independent of the occurrence of another by evaluating the mean and standard deviation for the training period of the set. The training has been performed on 40 data sets of each disturbance. The proposed artificial intelligence (AI) classifier provides $99.2 \%$ of testing accuracy in the identification of aforementioned disturbances. Thus, Naïve Bayes algorithm recommends adequately classifying single-stage notches and transients.