Robust Multilayer Model Predictive Control for a Cascaded Full-Bridge NPC Class-D Amplifier With Low Complexity

Abstract

Model predictive control (MPC) has the advantages of good dynamic performance and the ability to handle multiple control objectives and constraints. However, it suffers from large amount of calculation and high dependence on system model when it is applied to cascaded inverters. In this article, to solve this problem, a robust multilayer MPC (RM-MPC) with strong robustness and less computation is proposed for Class-D amplifier using cascaded full-bridge NPC inverter. A Kalman filter-based disturbance observer is designed to estimate the lumped disturbance caused by the mismatches of the load parameters, which allows to implement robust control by compensating the disturbance to the predictive model during each control period. In the upper layer of the multilayer MPC, the control objectives that are linear with the output level are handled centrally, which allows the optimal level to be obtained directly without repetitive predictions and evaluations. The middle layer is used to allocate the optimal level to each submodule. And the lower layer is used to determine the switching state of each submodule. Finally, the feasibility and validity of the RM-MPC are verified on the designed experimental prototype.