Novel bucket sort-based layered model predictive control without weight adjustment of modular multilevel converter

Purpose Model predictive control (MPC) has found extensive application for handling the control problems with the constraints and multi-objective. However, for a modular multilevel converter (MMC) system with multiple sub-modules (SMs), the computational cost of the traditional MPC would prevent from realizing the real-time control and the selection of the weights of the cost function would impact the efficiency when solving the optimal control problem. Design/methodology/approach This study presents a novel layered MPC based on the bucket sort. First, the SMs are sorted by the bucket sort method according to the capacitor voltage, and each bucket is considered an equivalent SM. The optimal configuration scheme of the equivalent SMs can be given by the first-layer MPC, and based on the result of the first-layer MPC, the optimal control scheme for the inserted SMs can be gained in the second layer. Finally, the optimal control sequence of the MMC system can be obtained by enhancing the compensation of the circulating current suppression. Findings Simulation in an MMC system demonstrates that the control performance can be improved with a reduced computational cost under the proposed method and avoid the disadvantages caused by the fixed weights. Originality/value This paper introduces a novel bucket-sort-based layered MPC framework that effectively addresses the high computational complexity and weight sensitivity issues of traditional MPC in high-dimensional MMC systems. The proposed method offers a practical and efficient solution for real-time control, providing a valuable reference for future research on complex power electronic system control.