This article presents a computationally efficient and high performing approximate long-horizon model predictive control (MPC) for permanent magnet synchronous motors (PMSMs). Two continuous control set MPC (CCS-MPC) formulations are considered: the classical current tracking delta MPC (Del-MPC) and the torque tracking economic MPC (EMPC). To achieve offset-free torque tracking under model uncertainties and in all regions of operation, a disturbance observer and a dq -current reference generator are used. To enable real-time implementation of the long-horizon CCS-MPC, the development of a real-time capable solver is not required, since MPC approximation based on deep neural networks (DNNs) is considered and utilized for controller’s evaluation at run time. The approximation is done by training the DNN to learn the MPC functionality based on offline-generated training data and in an open-loop manner. The robust and offset-free tracking performance of the proposed DNN-based approximate long-horizon Del-MPC and EMPC has been validated through simulation and real-time implementation at test bench and compared to the state-of-the-art field oriented control (FOC) using internal model controller with field-weakening (FW) part and to the exact short-horizon MPC based on the fast gradient method (FGM-MPC). Results show that the long-horizon MPC can achieve significantly faster torque transient responses in comparison with the short-horizon FGM-MPC and the conventional FOC, especially in FW region.
This paper deals with the design of a predictive safety filter which enables safe learning of optimal operating points for permanent magnet synchronous motors. It is shown that the proposed safety filter can be implemented in real-time on a rapid prototyping hardware using a tailored fast alternating minimization algorithm implementation. The effectiveness of the safety filter is shown when learning optimal operating points for a permanent magnet synchronous motor at the test bench using Bayesian optimization.
12 May 2026 in IEEE Open Journal of the Industrial Electronics Society
Optimal modulation techniques such as optimized pulse patterns (OPPs) are essential for minimizing power losses and hence achieving efficient operation of electrical drive systems. Conventional OPPs are designed offline to minimize different steady-state performance criteria and require switching angles-based analytical models suitable for gradient-based optimization. Such models are however not always available and can be complex and difficult to validate. Hence, a data-driven design approach that can utilize any form of models, or even be used for direct OPP learning, e.g., from a real-world drive, without requiring any modeling effort, is of significant interest. In this paper, we present a first proof of concept for a data-driven design of steady-state OPPs using reinforcement learning (RL). We explore the feasibility of training an end-to-end neural network policy that maps desired working points (WPs) to reference OPPs by interacting with an electrical drive environment. We introduce a novel parametrization of switching angles that inherently incorporates their constraints, and we utilize an efficient simulation set-up to simplify the training for this proof-of-concept investigation. Through extensive simulation and real-world experiments, we demonstrate the potential and trade offs of the proposed RL-based approach for designing loss-optimal pulse patterns for a traction permanent magnet synchronous motor drive. The performance is benchmarked against a conventional, indirectly loss optimized, model-based OPP design. Test bench results show a potential for power loss reduction in learned OPPs compared to conventional ones, despite the simulation model bias, and motivate learning OPPs directly from a real-world drive environment.
1 January 2026 in IEEE Transactions on Control Systems Technology