Non-Linear Model Predictive Control for Modular Multilevel Converters

Abstract

In this paper, non-linear model predictive control (NMPC) without an explicit modulator is applied to modular multilevel converters (MMCs) in the abc reference frame. NMPC can easily be extended for longer prediction horizons as opposed to finite control set model predictive control (FCS-MPC). However, NMPC applied to power converters in previous studies uses a modulator, which limits the transient response compared to FCS-MPC. Therefore, to avoid the modulator, two strategies are presented. In the first strategy, the continuous solution (number of inserted submodules per arm) obtained from NMPC is simply rounded off to the nearest integer for both the arms of each phase. In the second strategy, the optimal solution obtained from the NMPC is further evaluated by rounding it up and down for both arms. This requires four simulations per time step, independently from the number of SMs per arm. The evaluation of the four cases is conducted only for the initial time step within the prediction horizon. Then the solution that minimizes a pre-defined cost function is applied to MMC. The second strategy offers the fastest response and provides similar dynamic performance as indirect FCS-MPC, while both strategies offer similar steady-sate performance. Simulations are performed to validate the performance of the proposed methods compared to the FCS-MPC.