An adaptive neural online estimation approach of harmonic components

摘要

An asymptotic online estimation approach of significant harmonics and DC offset component of multi-frequency oscillating electric signals based on B-spline artificial neural networks is proposed. Harmonics are considered as unknown outputs or vibration modes of a vibrating system. Thus, a vibrating signal model is structurally used for real-time estimation design purposes of unknown terms constituting multiple frequency electric oscillations. In this fashion, harmonic outputs and DC offset are estimated using measurements of some available electric signal. B-spline neural networks are properly employed to compute online an adaptive unique estimator gain in presence of harmonic uncertainty. Analytical and numerical results prove the effectiveness of the artificial neural estimation for constitutive terms of electric oscillations.