A Computationally Efficient Current-Sensorless Three-Vector Modulated Model Predictive Control with Neutral-Point Voltage Balancing for T-type Inverters with LC filters

This paper proposes a current-sensorless model predictive control (MPC) method for T-type three-level inverters with LC filters, eliminating the need for inductor or load current measurements. A capacitor current observer and a cost-function-modulated three-vector MPC are integrated to simultaneously regulate output voltage and balance the neutral-point voltage. To reduce computational complexity, a two-stage optimization strategy prioritizes vector length selection and voltage balancing, enabling its real-time implementation at kilohertz frequencies. Experimental results validate the method's effectiveness, demonstrating comparable performance to sensor-based approaches.