Harmonic Estimation and Suppression for PMSM Drives via a Hybrid RBF-ADALINE Framework

Observer-based deadbeat predictive current control (DPCC) effectively handles lumped disturbances in permanent magnet synchronous motor (PMSM) drives, but suffers from periodic disturbances caused by inverter nonlinearities. While conventional harmonic rejection methods, such as resonant controllers and ADALINE-based methods, often exhibit instability or significant performance degradation at high speeds, due to discretization-induced phase lag and the absence of rotor electrical angle compensation. To overcome these limitations, this paper proposes a novel framework for harmonic estimation and suppression. First, by analyzing the failure mechanism of conventional radial basis function neural network observer (RBFO) in periodic disturbance rejection, a modified PMSM dynamic model with comprehensive disturbance characterization is obtained. Then, a radial basis function neural network disturbance observer (RBFDO) is proposed to simultaneously handle both low-frequency disturbances caused by parameter mismatch and a portion of high-frequency harmonic disturbances. Third, an adaptive linear neural network compensator (ALNC) is developed to achievedq-axis harmonic decoupling with rigorously proven stability, further improving harmonic compensation accuracy. Fourth, a parallel hybrid RBF-ADALINE framework (HRADF) is developed. Theoretical analysis shows that it reduces the observer’s relative order to suppress high-frequency phase lag, thus enhancing harmonic estimation and stability at high speeds. Finally, various experiments are conducted on a 64W PMSM experimental platform to validate the effectiveness of the proposed method.