A frequency-constrained MPC approach for safety-critical control of cascaded medium voltage MMC and DC buck converters in EV fast charging

The high-voltage DC-DC converters are considered the most important components of the upcoming MV-DC EV charger system. In MW systems, the DC-DC conversion module controls the DC voltage produced through the Modular Multilevel Converter (MMC) that connects to the power grid. Maintaining accurate voltage control in the face of high input voltages to the battery while suppressing any overshoot that might result from the large voltage differential is one of the key challenges. Linear controllers like the PI controllers may suffer from wind-up effects, thus producing unwanted transient overshoot. To overcome these limitations, a novel frequency-aware Model Predictive Control (MPC) scheme is introduced for cascaded MMC–Buck architectures. Unlike existing techniques, this MPC algorithm incorporates a dynamic reference trajectory into the optimization problem to achieve zero-overshoot performance. Furthermore, the approach utilizes a conditional frequency-penalty term to stabilize switching frequency around a 50 kHz target. A discrete-time, load-independent state observer is designed to decouple control dynamics from the continuously varying system impedance inherent to the EV battery charging process. Simulation studies demonstrate the superiority of the proposed MPC over a conventional PID algorithm tuned using the Archimedes Optimization Algorithm (AOA). While the conventional controller results in a 3.90% overshoot and a 0.104 s settling time, the proposed MPC provides a strictly safe, zero-overshoot response with a settling time of only 0.063 s. Comprehensive robustness analysis indicates that the proposed observer ensures globally stable performance, even with a 20% parameter mismatch in passive circuit elements.