Long-Term Memory Recursive Least Squares Online Identification of Highly Utilized Permanent Magnet Synchronous Motors for Finite-Control-Set Model Predictive Control

This is the authors' abstract. We don't add key points for this paper.

Highly utilized permanent magnet synchronous motors (PMSM) characterized by their nonlinear magnetization due to (cross-)saturation effects are the common choice when highest power density is required. For precise torque and current control, these motors are usually characterized by extensive offline measurements on a test bench, finally resulting in look-up tables of the relations between torque, flux, and current. In contrast, this article proposes a long-term memory recursive least squares (LTM-RLS) current estimator optimized for finite-control-set (FCS) model predictive controllers (MPC). This approach is able to identify the differential inductance and flux linkage maps for online self-commissioning without additional signal injection in only few seconds. This is achieved by extending the local RLS identification with an additional long-term memory. By continuously adapting the flux linkage maps, a precise open-loop torque control can be realized without the knowledge of exact motor parameters except the stator resistance as datasheet parameter. Extensive experimental investigations demonstrate accurate predictions of the identified model and, thus, highest control performance of the FCS-MPC during transient and steady-state operation and small torque estimation errors of less than$1.2 \,%$for speeds greater than$50 \,%$of the PMSM's nominal speed. Even for speeds of only$5 \,%$of the nominal speed, the estimation error is less than$7 \,%$.