Fast recursive Gauss-Newton adaptive filter for the estimation of power system frequency and harmonics in a noisy environment

This study presents a new recursive Gauss-Newton method-based adaptive filter for estimating the time-varying amplitude, phase, and frequency of fundamental and harmonic components of power system voltage or current corrupted by noise. The coefficients of the adaptive filter are obtained from unconstrained optimisation of an error function weighted by a variable forgetting factor, and using suitable approximation of the recursive Gauss-Newton algorithm for reducing computational complexity. To improve the tracking speed and accuracy of the filter, the forgetting factor is tuned iteratively. This approach is very much suitable for a large variety of applications in the distributed generation control and protection, synchronised phasor measurement of voltage and current signals using phasor measurement units, power electronic converters etc. Several computational issues like the time-varying nature of the power system voltage and frequency parameters, noise and distortion, presence of decaying dc and inter-harmonics are considered to evaluate the fast operation and robustness of the proposed approach.