An Improved Current-Sensorless Model Predictive Voltage Control for Four-Leg Voltage Source Inverters

Four-Leg voltage source inverters (VSIs) with LC output filter are commonly utilized to stabilize three-phase sinusoidal voltages under unbalanced loads conditions. However, conventional model predictive voltage control (MMPVC) for four-leg VSIs relies heavily on current sensors, leading to increased costs and suboptimal output voltage performance. To address these issues, this paper proposes an improved current-sensorless modulated model predictive voltage control (MMPVC) for four-leg VSIs to effectively reduce voltage tracking errors and total harmonic distortion (THDv). First, a readily implementable capacitor current extended state observer is constructed to replace current sensors and a corresponding discrete model is reconstructed. Then, four-vector MMPVC for four-leg VSIs is carefully designed. Especially, the ineffective voltage vectors are found, and their duration time is set as zero to optimize the output voltage performance in edge regions of sectors and over-modulation regions. Finally, experimental results validate the proposed approach, demonstrating substantial improvements in output voltage quality and tracking accuracy in over-modulation regions.