Model predictive control, with its powerful performance and optimization potential, has attracted widespread attention and interest in academia. However, existing methods typically focus on the perspective of individual converter units, with limited exploration of the collaborative potential among converter clusters. In this work, we propose an enhanced predictive control technique to fill the above gap. In the proposed method, a global synchronized computation unit is introduced to enable pulse pattern coordination among the converter cluster. Subsequently, by reconstructing the cost function and incorporating an adaptive weight iteration mechanism, a refined trade-off between transient and steady-state performance is achieved within the control process. Experimental data confirm its effectiveness.