Robust Direct Model Predictive Control with Reduced Computational Effort for Medium-Voltage Grid-Connected Converters with LCL Filters

Abstract

The performance benefits of long-horizon direct model predictive control (MPC) methods become more evident when high-order systems are considered. However, such applications pose a challenge implementation-wise as the increased size of the system model and adoption of long horizons can significantly increase the computational requirements of direct MPC. In addition, variations in the system parameters may deteriorate the controller operation. The presented method allows to harvest the performance benefits of long-horizon direct MPC with modest computational effort. This is achieved by adopting a split horizon formulation that enables the fulfillment of two tasks, namely, the prediction of the system behavior and evaluation of the candidate switch positions with marginal computational overhead. Moreover, to enhance the controller robustness to parameter variations, a simple estimator of the grid reactance is introduced. The effectiveness of the proposed approach is verified with a medium-voltage three-level neutral-point-clamped converter connected to the grid via an LCL filter.