Computationally Efficient Current Sensorless Predictive Control for PMSM Drive fed by a Matrix Converter with CMV-Free Operation

Model predictive control (MPC) for matrix converters (MCs) is typically computationally intensive due to the need to evaluate all possible switching states, limiting its industrial application. Recent advances have reduced this burden by cutting the required predictions to just two, improving computation speed and lowering costs. However, the need for multiple sensors to measure various currents and voltages still drives up overall expenses. This paper introduces two look-up tables that reduce candidate switching states from 27 to 5 to address these issues. In addition, It incorporates a Luenberger observer to eliminate current sensors while maintaining system robustness. To further enhance the overall performance of the MC, a common mode voltage (CMV) elimination technique utilizing only 6 rotating vectors is integrated into the current sensorless control scheme. Experimental results on a PMSM drive system fed by an MC demonstrate significant reductions in computational load and hardware costs, enabling high-frequency operation with enhanced performance.