Traditional model predictive control involves a large amount of calculation and it is difficult to set the weight factor. Also, it is not suitable for a multi-port modular multilevel AC/DC hybrid converter (MP-M2ADHC). In this paper, the AC/DC coupling characteristics and internal power balance constraints of an MP-M2ADHC are analyzed using the equivalent principle of arm internal potential, and a piecewise variable step-size model predictive control (PVS-MPC) strategy is proposed. Through multiple optimization links, the optimal switching state is sought over different time scales to minimize the error between the predicted and the expected value of the system, and realize the independent control of multiple control targets. To satisfy the demand for accurate discretization of the prediction model, an improved Euler method is proposed to optimize and correct the feedback value of the system, so as to significantly improve the prediction accuracy. Finally, through the time-domain simulation of power bidirectional transmission of converters in different modes, the feasibility and effectiveness of the proposed control strategy and optimization correction strategy are verified by comparing with the traditional model predictive control in terms of output current tracking, circulation suppression and harmonic content.