Between the two circuits of the parallel-connected inverter module with Si IGBT/SiC MOSFETs, there exists coupling. Traditional strategies struggle to achieve decoupling control. However, model predictive control can address coupling issues by adding decoupling terms during mathematical modeling and possesses superior predictive and constraint-handling capabilities. This paper proposes a variable-step-length optimization control based on the difference between the square sum of predicted current tracking errors and the absolute value of common-mode voltage rate of change. Additionally, it introduces the common-mode voltage rate of change as a balancing factor in the cost function to optimize the model predictive control. Compared to traditional model predictive control, the output current THD is only slightly different, but there is a significant reduction in electromagnetic interference generated by the inverter. The feasibility of this control method is demonstrated through the construction of MATLAB simulation models and experimental verification.