Data-Driven Recursive Least Squares Estimation for Model Predictive Current Control of Permanent Magnet Synchronous Motors

The performance of model predictive controllers (MPC) strongly depends on the quality their models. In field electric drive control, white-box (WB) modeling approaches derived from first-order physical principles are most common. This procedure typically does not cover parasitic effects and often comes with parameter deviations. These issues particularly crucial in domain self-commissioning drives where a hand-tailored, accurate WB plant is available. order to compensate for such errors and, consequently, improve control during transients steady state, this article proposes data-driven, real-time capable recursive least squares estimation method current permanent magnet synchronous motor. Following machine learning approach, effect flux linkage voltage harmonics due winding scheme can also be taken into account through suitable feature engineering. Moreover, compensating interlocking time inverter proposed. resulting algorithm investigated using well-known finite-control-set MPC (FCS-MPC) rotor-oriented coordinate system. extensive experimental results show superior presented compared FCS-MPC-based state-of-the-art motor look-up tables addressing (cross-)saturation.