Computationally Efficient MPC With Embedded Adaptive Battery Balancing for CHB Inverters

This paper proposes a Tree-Based Finite-Control-Set Model Predictive Control (TBFCS-MPC) strategy for a Cascaded H-Bridge (CHB) inverter, integrated with an adaptive online Look-Up Table (LUT) for real-time battery voltage balancing. The proposed TBFCS-MPC reduces computational complexity by limiting the number of evaluated voltage vectors to a maximum of 22, compared to 127 in conventional FCS-MPC approaches. This reduction enables real-time execution on embedded hardware, achieving a cost function evaluation time of only$\bf {4.2\ \mu s}$, compared to$\bf {24.1\ \mu s}$with state-of-the-art solutions. The LUT dynamically selects switching states based on instantaneous battery voltages, ensuring active balancing. Experimental results validate the method's fast current reference tracking, effective balancing, and significantly reduced computational burden, making it a practical solution for high-performance, battery-fed multilevel inverter applications.