Cooperative modular multilevel converter control based on PSO-optimized fuzzy-PI and hierarchical finite-state model predictive control

To address the limitations of empirical parameter tuning in the outer-loop PI controller and the high computational burden of the inner loop in conventional dual-loop control for modular multilevel converters, a coordinated control strategy is proposed that integrates particle swarm optimization-based fuzzy PI control with finite-state hierarchical model predictive control. The outer loop employs a two-dimensional fuzzy PI controller, in which error and error variation are selected as inputs. The quantization and proportional factors are optimally determined through an offline global search using particle swarm optimization, enabling adaptive parameter adjustment and improving the tracking accuracy of current reference signals under varying operating conditions. The inner loop adopts finite-state model predictive control, where AC-side phase voltages and total arm voltages are predicted using a reverse calculation method. Based on objective function minimization combined with integer rounding, the required number of submodules for the upper and lower arms is determined. Circulating current suppression is subsequently introduced to mitigate steady-state deviations on the AC side, and predictive submodule voltage grouping and sorting are applied to balance capacitor voltages. This approach eliminates the need for weighting factor design and effectively reduces computational complexity. Finally, simulations carried out in MATLAB/Simulink confirm the effectiveness of the proposed strategy.