A Sensor-Optimized MPC Strategy for Voltage Balancing in Medium-Voltage Grid-Tied MMCs Using State Matrix Reduction and PS-PWM

The existing model-predictive control (MPC)-based sensor reduction methods require additional estimation and sorting algorithms to achieve capacitor voltage (CV) balancing in grid-tied modular multilevel converters (GT-MMCs). These methods heavily depend on accurate system parameters, complicating the control process and limiting effective balancing actions. This article proposes a novel balancing method that integrates phase-shifted pulsewidth modulation (PS-PWM) and MPC, using fewer sensors. This method represents the switching state (SS) in a simplified matrix form, reducing it into the lower arm (l-arm) and upper arm (u-arm) submatrices. Applying PS-PWM and a reduced-order matrix eliminates the need for individual submodule (SM) voltage sensors in the u-arm. The internal dynamics control involves the generation of u-arm switching pulses through modulating signals from the grid current controller and PS-PWM, while MPC governs the l-arm. This reduction leads to a significant decrease of ( $N-1$ ) voltage sensors in the system. Additionally, this work proposes a robust fault monitoring system by integrating CV balancing control with the fault diagnosis process through direct voltage measurements obtained from sensors. Simulations and test results are presented to validate the effectiveness of the proposed method, showcasing its potential in implementing MPC and determining the failure status of SMs with a reduced sensor count.