Fast nonlinear model predictive control: Formulation and industrial process applications

With the widespread availability of model predictive control (MPC), nonlinear MPC provides a natural extension to include nonlinear models for trajectory tracking and dynamic optimization. NMPC can include first principle models developed for off-line dynamic studies as well as nonlinear data-driven models, but requires the application of efficient large-scale optimization strategies to avoid computational delays and to ensure stability, robustness and superior performance. This study presents the application of the recently developed advanced step NMPC (asNMPC) strategy. This approach solves the detailed optimization problem in background and applies a sensitivity-based update on-line. Two large-scale process case studies are considered: detailed distillation control and multi-stage operation for steam generation in a power plant. In both cases, efficient and robust controller performance is achieved with nonlinear dynamic optimization.