Model predictive control of electric machines in commercial vehicles
Examinator: Torbjörn Thiringer, Inst för elektroteknik
Handledare: Jonas Ottosson och Faisal Altaf, Volvo Truck
Opponent: Yiwen Xu
This thesis focuses on the realization of a Finite Control Set Model Predictive Control (FCS-MPC) scheme for a benchmark electric drive system which is supported by Volvo Group. A Permanent Magnet Synchronous Machine (PMSM) is used in this electric drive system, the non-linear model of the PMSM is derived after taking the magnetic saturation and the cross-coupling effects into consideration. Certain characteristic trajectory functions of the PMSM control strategies such as Maximum Torque Per Ampere (MTPA), Field Weakening (FW) and Maximum Torque Per Volt (MTPV) are formulated based on the non-linear PMSM model. A two-level Voltage Source Inverters (VSI) without modulation scheme has 8 switching states in total to generate the input voltage for the PMSM. In order to reduce the switching losses, assumption has been made that at most one of the three inverter phase legs is allowed to change its switching state between each two sampling instances. The future system states can be predicted with the finite set of the possible input voltages and the discretized PMSM model. A multi-objective cost function is introduced to derive the optimal voltage input at the next time step. The cost function deals with three targets: torque tracking, attraction region and limitation. The torque tracking part enables the PMSM to generate the reference torque. The attraction region part forces the current states to evolve on the desired operation trajectory which is depend on the torque and speed level. The limitation part aims to avoid the current states reaching undesired operation regions. A PI controller has been designed for the benchmark electric drive system, the operation performances are compared between the FCS-MPC and PI based on the aspects of step response, tracking correctness, switching losses, current ripple and torque ripple. Though the ripple performance of the FCS-MPC is not as well as that of the PI, FCS-MPC helps to realize a faster rise time and lower switching losses. Evaluations of the FCS-MPC are also included with respect to different sampling frequencies and prediction horizons, which mainly affect the performance of tracking correctness along different trajectories.
Fredrik Lamms rum, Hörsalsvägen 11, EDIT trappa A och B