Decoupled Virtual Vector-Based Predictive Current Control of Dual Three-Phase PMSMs With Model-Free Harmonic Current Suppression

Finite control set model predictive control (FCSMPC) based on virtual vectors for dual three-phase permanent magnet synchronous machines inherently lacks control capability in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x–y</i> subspace, resulting in an inability to suppress unmodeled disturbances decoupled in this subspace and leading to severe current distortion. Existing studies have incorporated <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x–y</i> current prediction to compensate for disturbances in a closedloop manner but require prior knowledge of the stator leakage inductance. Mismatches in this parameter degrade harmonic suppression performance. In this work, we introduce a modelfree solution that relies solely on measured input/output data, eliminating dependence on <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x–y</i> subspace model parameters. To enable the linear mapping between the input/output data to describe the rich periodic disturbances in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x–y</i> subspace, sine and cosine feature excitations are added to the regressor. The enhanced data model better approximates the physical model for predicting <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x–y</i> current dynamics. A low-complexity normalized least mean square (NLMS) algorithm is employed to optimize the regression coefficients, allowing the proposed method to be implemented on a digital signal processor with 10kHz sampling frequency. Experimental results validate the superiority of the proposed strategy over other existing methods.