Model Predictive Direct Switching Control for Multi-Phase Permanent-Magnet Synchronous Motors

Abstract

In model predictive direct switching control (MPDSC) approaches, the finite set of inverter switch positions for the control of electrical drives is taken as control variables within the MPC underlying optimization problem. The cost function can be designed to track a torque or current reference while also considering additional control goals like minimizing switching and conduction losses, as well as physical constraints. The challenge is to solve the resulting integer quadratically constrained quadratic program (IQCQP) with high sampling frequencies online in order to obtain the optimal switch positions at each fixed time step. This paper presents two new MPDSC algorithms, namely relaxed barrier functions iteration scheme (RBF) and alternating direction method of multipliers heuristic (ADMM) and compares them with two state-of-the-art algorithms, namely full enumeration (FE) and multistep with sphere decoding (MSD) for the example of a multi-phase permanent-magnet synchronous motor (PMSM) in simulations.