Estimation and elimination of harmonics in power system using modified FFT with variable learning of Adaline

This paper presents a new technique for harmonic detection, estimation and elimination in power system. The approach expresses the input signal in the form of Fourier Transformation and adjusts the Fourier coefficients using Linear Adaptive Filters (Adaline). Two of the existing approaches using conventional Fast Fourier transformation (FFT) and FFT with modified W-H learning have been discussed with their merits and demerits. A new approach has been proposed using a network comprising three different Adalines and involving variable learning rate. This approach is able to mitigate the shortcomings of the existing techniques and is time efficient as well. The detailed architecture for the proposed approach has been discussed and the algorithm has then been tested on different test signals. The approach has been compared with the existing techniques and it's superiority over these has thus been established.