An improved Adaline algorithm for online tracking of harmonic components

This investigation presents an enhanced Adaline algorithm to track the harmonic components of a power system. By using a revised objective function and a variable learning parameter technique, the proposed algorithm can yield markedly better results than most conventional adaptive algorithms for semistationary signals. The objective function used in the algorithm proposed herein combines a forgetting function and, hence, the mean error energy can force the system to adjust better to nonstationarity. To alleviate the shortcomings embedded in conventional constant-learning parameter algorithms, which generally require a tradeoff between misadjustment and the converge rate, the variable learning parameter technique uses information more efficiently and thereby yields a better overall performance. Comparative studies on numerical experiments demonstrate the effectiveness of the proposed algorithm.