Comparative analysis of real-time OSV-MPC techniques in a multicore microcontroller-based hardware-in-the-loop

Abstract

This work presents a comparative analysis of three different implementations of the Optimal Switching Vector Model Predictive Control (OSV-MPC) strategy applied to a grid-forming inverter with an output LC filter in an islanded microgrid. The main objective is to compare the power quality of the output voltage synthesized by the standard OSV-MPC, the OSV-MPC with two-step prediction and the OSV-MPC with two-step prediction and Lagrange extrapolation on the output voltage reference signal. A multicore hardware-in-the-loop (HIL) simulation scheme is implemented, using both cores of a Texas Instruments F-28379D microcontroller: one core holds the control algorithm, while the other holds the power system mathematical model. Five different load conditions are considered on steady-state and transient analyses: no load, resistive, capacitive, inductive and unbalanced loads. The real-time simulation scheme is validated using MATLAB/Simulink offline simulation results as reference. The comparative study between the OSV-MPC variations is done based on the analysis of the total harmonic distortion (THD), mean absolute error (MAE) and total frequency spread (TFS) metrics, as well as the processing time of each OSV-MPC variation is measured. The OSV-MPC with two-step prediction and Lagrange extrapolation demonstrated better cost-benefit, providing a reduction in THD and MAE compared to the other OSV-MPC variations, with an increase of less than 2 µs in algorithm processing time.