Reduced-Sensors-Based Predictive Controller for LC Filtered Four-Leg Inverters

Abstract

The high number ofsensors, ten in total, required by the traditional finite control set model predictive control (FCS-MPC) strategy of a four-leg inverter with LC output filter increases the complexity and cost of the system and limits the practical application and commercialization. To simplify the implementation, reduce the cost and enhance the reliability, a novel current sensorless predictive load voltage control scheme, with only four voltage sensors and no current sensors, is proposed in this work. A new dynamic model of the LC filtered four-leg inverter is constructed in the stationary ( $\alpha \beta \gamma$ ) reference frame for the proposed load voltage and capacitor current observer. The dual-boundary-layer nonlinear feedback is combined with a linear feedback term to enhance the robustness while retaining the accuracy of the observer. Extensive experiments are performed to confirm the effectiveness of the proposed controller. The results validate that the proposed sensor-less FCS-MPC with only four voltage sensors achieves an equivalent performance to the conventional FCS-MPC with ten sensors.