Model Predictive Control of Switched Reluctance Machine Drives

Abstract

Model predictive control (MPC) for switched reluctance machine (SRM) drives is studied in this thesis. The objective is to highlight the benefits of implementing MPC to overcome the main drawbacks of SRMs and position them as an attractive alternative among electrical drives. A comprehensive literature review of MPC for SRM is presented, detailing its current trends as an application still at an early stage. The different features of MPC are highlighted and paired with the most challenging and promising control objectives of SRMs. A vision of future research trends and applications of MPC-driven SRMs is proposed, thus drawing a road-map of future projects, barriers to overcome and potential developments. Several important applications can take advantage of the improved features that SRM can get with MPC, especially from the possibility of defining a unified control technique with the flexibility to adapt to different system requirements. The most important cluster for SRM drives is the high- and ultrahigh-speed operative regions where conventional machines cannot work efficiently. SRMs with MPC can complement then the existing demand for electrical drives with high performance under challenging conditions.

Three techniques based on the finite control set model predictive control (FCS-MPC) approach are developed out of the proposed road-map. The first one defines a virtual-flux current tracking technique that improves the existing ones in operating at different speeds and more than one quadrant operation. The method is validated for low- and high- power SRMs in simulations and diverse types of current waveform, making it easy to adapt to existing current shaping techniques. It is also validated experimentally for different operating conditions and robustness against parameter variation. The second technique proposed a predictive torque control that bases its model on static-maps, thus avoiding complex analytical expressions. It improves its estimation through a Kalman filter. The third technique uses a virtual-flux predictive torque control, similar to the first technique for current tracking. The techniques are validated at a wide speed range, thus evidencing superiority in performance without modification on the control structure.