Automated Power Quality Assessment Using IEVDHM Technique

This paper aims to develop a new framework for the classification of power quality (PQ) disturbances based on improved eigenvalue decomposition of Hankel matrix (IEVDHM) and Hilbert transform separation algorithm (HTSA). IEVDHM decomposes a disturbance signal into a set of mono-component signals. Further, HTSA technique computes the amplitude envelope (AE), instantaneous phase (IP), and instantaneous frequency of the decomposed mono-component signals. The 5 time-dependent spectral features (TDSFs) and 6 HTSA-based features are extracted from each decomposed mono-component signals. The TDSFs are logarithmic zero order spectral moment, logarithmic difference of zero order and second order spectral moment, logarithmic difference of zero order and fourth order spectral moment, spectral sparseness and irregularity factor. Whereas, the HTSA-based features are mean of AE, energy of mono-component signal, standard deviation of AE, standard deviation of IP, envelope entropy, and mean frequency. Then, the bagged decision tree, optimizable neural network, and linear support vector machine classify these extracted features into 19 different types of disturbance classes. The bagged decision tree-based classifier has achieved an accuracy of 92.48 %, 91.07 %, 88.98 %, 88.01 % for clean PQ signals, noisy (additive white Gaussian noise) PQ signals with signal to noise ratios 60 dB, 40 dB, and 20 dB, respectively. The performance of the proposed framework is compared with the performance of empirical mode decomposition (EMD), ensemble EMD, and variational mode decomposition-based frameworks for PQ detection and is found to be superior.