Adaptive and delayless filtering system for sinusoids with varying frequency

Accurate current reference is important for an active power filter to efficiently suppress harmonics caused by power electronics devices. The reference signal can be generated by applying an adaptive predictive filter to extract the sinusoidal primary signal from a noisy source. In this paper, the authors propose an FIR (finite impulse response) filter whose coefficients are updated by the LMS (least-mean-square) algorithm to construct the adaptive filtering system. A feedback is added to the adaptive filter structure to provide efficient operation with reduced computational complexity. Their adaptive filter can predict the primary sinusoid while considerably attenuating the harmonic components. Since adaptive prediction is applied, the prediction step remains accurate even if the frequency changes in time.