Model Predictive Control for Modular Multilevel Converters based on a Box-constrained Quadratic Problem Solver

Abstract

The model predictive control (MPC) method has become more popular and widely used to control power converters due to its fast dynamic response and easy implementation. With the MPC method, it is simple to achieve multiple control objectives and handle the system constraints and nonlinearities by using a cost function. MPC can also be employed on the modular multilevel converters (MMCs) to achieve the control of output currents, circulating currents and DC-link voltage/ current. However, the biggest obstacle to applying MPC on the MMC is the huge calculation. To achieve the optimal switching state, MPC needs to evaluate all the possible switching states. As the number of submodules (SMs) increases, the number of the switching states drastically increases, which puts a huge computational burden on the processor. To solve this problem, the MPC with a box-constrained quadratic programming solver has been proposed. In this method, a quadratic problem (QP) has been solved firstly. Based on this solution, the possible switching combinations have been determined. Instead of evaluating all the possible switching states, only a very small number of switching combinations needs to be evaluated. Besides, by using this method, the computational burden does not increase as the number of SMs increases.