Fast frequency and harmonic estimation in power systems using a new optimized adaptive filter

This paper presents an adaptive filter for fast estimation of frequency and harmonic components of a power system voltage or current signal corrupted by noise with low signal to noise ratio (SNR). Unlike the conventional linear combiner (Adaline) approach, the new algorithm is based on an objective function often used in independent component analysis for robust tracking under impulse noise conditions. However, the accuracy and speed of convergence of this algorithm depend on the choice of step size of the filter and its adaptation. Instead of choosing the step size η and the parameter β of the cost function by trial and error, an adaptive particle swarm optimization technique is used alternatively to obtain both η and β to reduce the error between the observed voltage or current samples and the estimated ones. Using the optimized values, the amplitude and phase of the fundamental and harmonic components are estimated. Further, the extracted fundamental component is used to estimate any frequency drift of the power system recursively using an optimized error function obtained from three consecutive voltage samples. To test the effectiveness of the algorithm, several time-varying power system signals are simulated with harmonics, interharmonics, and decaying dc components buried in noise with low signal-to-noise ratio (SNR) and are used to estimate the frequency and harmonic components. This approach will be useful in islanding detection of a distributed generating system.