Predictive Filtering Methods for Motor Drive Instrumentation

摘要

In this work, the existing theory of signal prediction is applied and extended tothe sensor signal processing in motor drives. For sinusoidal signals, 'delayless' filtering schemes are described, consisting of an adaptive FIR (Finite Impulse Response) filter employing the Lease-Mean-Square (LMS) algorithm for filter coefficient adaptation. Various prefiltering, interpolation, and sampling rate normalization schemes, in turn, make it possible for the filtering systems to adapt to frequency and amplitude alternations even in the presence of considerable noise. The filtering systems are applied to zero-crossing detection of line frequency signals to facilitate the synchronization of gating pulses in thyristor power converters, zero-crossing detection of cycloconverter output currents for thyristor bridge selection task, and extraction of sinusoidal active current to be used as a basis for current reference in active power filters. The analysis of applications and simulations results illustrate that the described filtering schemes can truly improve the accurancy of estimation of signals encountered in proposed applications.