Hierarchical Model Predictive Control for Modular Multilevel Converters: A Linear Parameter Varying Approach

Abstract

Modular multilevel converters (MMCs) are novel voltage source converters (VSCs) with the capacity to achieve higher efficiency and lower total harmonics distortion, but more challenging to control. MMCs can be modelled using a hierarchical structure comprising a model predictive controller (MPC) that controls the current in the top layer and sends signals to capacitor voltage modules in the lower layer (which can run in open-loop or closed-loop with a simple local controller). Typically, the prediction model used in the MPC current controller does not take into account the capacitor voltage dynamics in the bottom layer, which reduces the efficiency of the hierarchical control scheme and MMC circuit. Therefore, in this paper, we develop an extended prediction model for current control that includes the capacitor voltage dynamics for the modules in the bottom layer. This results in a nonlinear MPC control problem of higher complexity but with improved performance. To reduce the complexity of the nonlinear MPC problem for MMCs, we make use of a linear parameter varying embedding of the nonlinear prediction model, which allows solving a sequence of quadratic programs online instead of a nonlinear program. The proposed control scheme shows better performance while decreasing the computation load. Furthermore, compared to classical hierarchical control schemes, the proposed scheme reduces the capacitor voltage ripple by 20%.