Adaptive Filter Bank Based on Independent Component Analysis for Harmonic, Inter-Harmonic and Sub-Harmonic Extraction

In the present work, an adaptive filter bank based on independent component analysis (ICA) for harmonic, inter-harmonic and sub-harmonic extraction from voltage signals is proposed. The first stage of the proposed method comprises in generating a multi-channel data representation by using a finite impulse response (FIR) filter bank. In the second stage, this multi-channel representation is presented to the ICA algorithm, which extracts the harmonic, sub-harmonic and inter-harmonic components from the monitored signal. The method was evaluated with three simulated cases, considering harmonics, inter-harmonics, sub-harmonics and voltage sag, and a real signal acquired from a 13.8 kV arc furnace corrupted by harmonics, supra-harmonics, flicker and sag. The influence of noise and disturbances, the size of the processing signal window and the computational complexity of the proposed method were investigated. The results showed that the method was able to extract the disturbance components present in the signal accurately and with lower computationally complexity in comparison with the classical ICA-based method.