Model Predictive Control (MPC) is by far the most successful advanced control technology. Nowadays, it is applied in numerous industrial, medical and financial applications, e.g. chemical process control, power plant operations, smart energy systems, subsurface oil recovery, medical devices, engines, dynamic portfolio optimization etc.
Model Predictive Control requires repeated unsupervised online solution of dynamical optimization problems. Consequently, numerical optimization technologies and efficient algorithms tailored for MPC optimization applications are very important to apply MPC successfully. In addition the modelling and tuning are also necessary to obtain good closed loop performance.
At DTU Informatics we do research in MPC and have software libraries containing optimization algorithms tailored for MPC applications as well as systematic algorithms for tuning of MPC systems. Our library contains algorithms for both linear and nonlinear MPC. We have research projects that develop algorithms and software for MPC as well application oriented research projects, e.g. robust cement mill circuit control by MPC, oil recovery by optimal control, MPC for smart energy systems, and MPC for the artificial pancreas.
A systems diagram for an artificial pancreas for people with diabetes. The control algorithm in this system is based on MPC.