Real-Time Capable Model Predictive Control of Permanent Magnet Synchronous Motors Using Particle Swarm Optimisation

Permanent magnet synchronous motors (PMSM) are widely used in many industrial applications due to their attractive power and torque densities. To fully exploit their torque dynamics, non-linear model predictive control (MPC) approaches have been investigated in the last years. Besides the requirement of an accurate motor model solving a suitable cost function under real-time conditions is a challenging task, since highly-utilised PMSM exhibit a strongly non-linear behaviour due to (cross-)saturation effects. In this contribution a particle swarm optimisation (PSO) based MPC using multiple sub-swarms which are implemented on a multi-core processor platform is investigated. Thanks to parallel program execution, a suitable turnaround time (<100 mus) is achieved. In contrast to conventional MPC approaches the proposed PSO-based technique is not limited to convex cost functions or linear plant models and therefore extends the MPC application scope. Measurement results based on a 60 kW prototype motor fed by a 2-level-IGBT inverter demonstrate the general proof of concept. In comparison to a state-ofthe-art linear feedback controller (PI-type) a slightly improved control performance is observed.