MPSoC-Based FCS-MPC with Space Vector Preselection and Online Parameter Identification for Matrix Converter

The conventional finite-control-set model predictive control (FCS-MPC) method for matrix converters faces significant challenges of high computational demands and parameter uncertainty. To address these issues, an FCS-MPC approach that combines space vector preselection with online parameter identification is proposed. This method greatly reduces the computational load associated with determining the switching state for each cycle by employing a simplified state-trajectory prediction model for the input filter and decreasing the number of candidate switching states from 27 to 5. Additionally, it allows for continuous monitoring of input filter components and load conditions, integrating this functionality into the control mechanism to enhance overall system performance. The use of multi-processor system-on-chip (MPSoC) devices provides high parallelism and flexibility, enabling the algorithm to execute within 10 $\mu$s and facilitating a high sampling frequency, which leads to improved converter performance including total harmonic distortion and power factor. Experimental results from a prototype validate the effectiveness of the proposed approach across various power conversions, including AC/AC, AC/DC, DC/AC, DC/DC, and bidirectional power transfer.