Power Quality Disturbance Detection, Classification and Correction

Abstract

For the purpose of Denoising the Power signals, the accurate estimation of the noise disturbances and the time of occurrence of the noise is needed. Once the time of occurrence of the noise is detected, it is vital to classify the type of noise so that the corrective action based on the same is done. By gauging the energy of the distorted signals at different resolutions by the virtue of the Energy Difference Multi Resolution Analysis, (EDMRA) the disturbance is identified. At different levels of resolution the distorted signal's energy distribution is found. The db4 and Morlet mother wavelet is used for resolving the noise signal in both time and frequency. The power disturbances in the signal are identified based on the difference in energy for each noise type taking the pure sinusoidal signal of 50Hz as the reference signal. The generated feature vector is fed into the input layer of a pre-trained neural network, which classifies power quality abnormalities. The adaptive filter employs an adaptive linear network to produce compensatory action for the noise signal (adaline).