A Nonlinear MPC Strategy for AC/DC-Converters Tailored to the Implementation on FPGAs

Abstract

AC/DC converters are important components in many industrial applications. Their optimal operation is thus of growing relevance in the recent years. In the state of the art, optimal (nonlinear) control of AC/DC converters frequently focuses on finite set or explicit model predictive control (MPC) strategies. The main drawback is that their computational effort exponentially increases with the length of the prediction horizon and thus, only very short horizons (2 or 3-times the sampling time) are practically feasible. This can yield a significant limitation of the achievable control accuracy. Thus, in this paper a nonlinear MPC strategy is proposed, which allows for rather long prediction horizons while keeping the computational effort low. Furthermore, the MPC strategy is formulated in a way that it is tailored to the implementation on FPGAs, utilizing its inherent parallel computation capabilities. It is shown by means of simulation results that a high control accuracy and robustness to external disturbances can be achieved. Furthermore, it is discussed that the proposed MPC allows for computation times in the lower us range on an FPGA.