Comparison of adaptive filter performance in estimating the noise statistics for harmonic models

Electric power system harmonics and interharmonics having random variations in magnitude are modeled as state variables of a discrete-time stochastic system. The problem of adaptively estimating the statistical characteristics of the process and measurement noises is addressed. Incorporating the unknown noise covariance in the system leads to a state-multiplicative-noise model resulting in a nonlinear estimation problem. The main contribution of the paper is the performance comparison, by Monte Carlo simulation, between the linear minimum-variance filter and the extended Kalman filter. Conclusions based on these simulation results are presented.