Improved Model Predictive Control Strategy for Modular Multilevel Converter

Abstract

Abstract. For reducing switching frequency and decrease computation of the traditional model predictive control (MPC) for modular multilevel converter, an improved MPC strategy is proposed. The control of AC current, circulating current and capacitor voltages are realized synchronously by building multi-objective functions. Combining with capacitor voltage sorting method, MPC calculation is decreased. The switch frequency was lowered by introducing weighting factors. Capacitor voltage difference among the sub-modules is lessened through the allowable value of capacitor voltage deviation between sub-modules. The simulation results verify its correctness. IntroductionModular multilevel converter (MMC) is widely applied to high-voltage direct current, power quality control, AC drive and other high-voltage and high-power applications due to its prominent features such as modularity and cascade. But in the application of MMC, it is still urgent for MMC to solve the stability of sub-module capacitor voltages and the suppression of circulating current due to the interaction between the sub-module capacitor voltages and the circulating current. So a lot of literature has done a number of studies on it.In [1][2][3], the balance strategy of the capacitor voltages for sub-modules is proposed, but the influence of the circulating current on the capacitor voltages of sub-modules is ignored. In [4], the arm inductance is used to suppress the circulating current. The method not only increases the equipment cost of the system, but also has limited effect on the suppression of the circulating current. In order to suppress double frequency circulating current in MMC, a circulating current suppressor is proposed in [5] based on double frequency negative sequence coordinate transformation. But it is more complex because of double frequency negative sequence coordinate transformation and interphase current decoupling. Model predictive control (MPC) is used to control MMC in [6,7].However, it is necessary to predict all possible switch state combinations of the upper and lower arms for the next cycle in each control cycle to achieve multi-objective optimization control. For a 5-level MMC, the number of sub-modules per arm is 4, and the number of switch states to be considered for each control cycle is . With the increase of the number of MMC level and the number of sub-modules, the computational complexity of the switch state combinations is obviously increased, which puts forward higher requirements for the controller.For realizing AC / DC power conversion and safe and stable operation, a MPC strategy based on n+1 level MMC is proposed with the control targets of AC current, interphase circulating current and capacitor voltage of sub-modules. In this strategy, the weight accumulation factors are used to fuse the multiple control objectives and to construct the multi-objective optimization functions. The strategy can reduce the calculation of MPC by combining the capacitor voltage sorting method and lower