Certain investigations on detection and mitigation of power quality disturbances in distribution system

Abstract

Recent advancements in the field of power electronics and microcontrollers in automated systems lead to Power Quality (PQ) problems. In rder to improve the quality of supply, FACTS and Custom Power Devices (CPD) are involved for compensation and mitigation of disturbance signal. Before doing mitigation, it is essential to identify the type of disturbance. Hence a two-dimensional approach is proposed to detect and mitigate the PQ disturbances through a single system. The proposed system is well enough to analyze the signals in connection with industrial loads under the distribution system. It is made possible using a low cost and efficient power quality analyzer designed using Arduino-based microcontroller arrangement. The captured real-time signals across industrial loads are subjected to feature extraction stage. Basically, Wavelet Transform (WT) with Daubechies (db4) as the mother wavelet is chosen for feature extraction. Further, the new method called Fusion of Time Domain Descriptors (FTDD) is implemented, that provides informative features required for classification. In the next stage, the extracted features are given as the input to different classifiers like multi-Support Vector Machine (MSVM) and Naïve Bayes (NB). The performance is validated by adding noise distortions with the real-time signals. Both classifiers show a better recognition rate, among which, NB is capable of achieving better result under different Signal to Noise Ratios (SNR). Moreover, a Shunt Active Power Filter (SAPF) is switched ON, if the harmonic spectrum of the detected signal shows a Total Harmonic Distortion (THD) value above the limit. The PI controller along with the Ant Colony Optimization (ACO) technique reduces the THD to a very lower value. Thus, the distorted waveform is reconstructed into a pure normal signal. newline