||Model Predictive Control
||Dr.ir. A.J.J. van den Boom
|Contact Hours / Week x/x/x/x:
|Expected prior knowledge:
||The model predictive control (MPC) strategy yields the optimization of a performance index with respect to some future control sequence, using predictions of the output signal based on a process model, coping with amplitude constraints on inputs, outputs and states. The course presents an overview of the most important predictive control strategies, the theoretical aspects as well as the practical implications, that makes model predictive control so successful in many areas of industry, such as petro-chemical industry and chemical process industry. Hands-on experience is obtained by MATLAB exercises with academic examples and a industrial simulation of MPC on a two-product (binary) distillation column. Contents of the course: General introduction. Differences in models and model-structures, advantages and limitations. Prediction models in state-space setting. Standard predictive control scheme. Relation standard form with GPC, LQPC and other predictive control schemes. Finite/Infinite horizon MPC. Solution of the standard predictive control problem. Stability, robustness, initial and advanced tuning. Robust design in predictive control. See also: http://www.dcsc.tudelft.nl/~sc4060
The student should be able to
1. Explain how and why MPC has emerged from industry.
2. List the five basic items of MPC and discuss their role.
3. Recognize and describe two different type of models (IO and IIO models) in MPC and explain when a type of model is suited for a specific application.
4. Show that all models can be transformed into a state-space model.
5. Understand the concept of prediction in MPC.
6. Make a prediction in the noiseless and the noisy case.
7. Explain why a standard formulation is desirable.
8. Transform an MPC problem into the standard MPC problem.
9. Derive the steady-state of a system.
10. Solve the finite and infinite horizon problem.
11. Derive the realization for the LTI-case and for the inequality constrained case.
12. Describe three ways to deal with infeasibility.
13. Discuss stability for the LTI case and in the inequality constrained case.
14. Describe four modifications of the MPC design method that will provide guaranteed stability of the closed loop
15. Give the relation of the MPC scheme and the IMC scheme.
16. Describe the concept of robustness in MPC.
17. Motivate and use the rules of robust tuning in MPC.
18. Motivate the rules-of-thumb for initial tuning and use these rules for tuning an MPC controller.
19. Derive an MPC controller using MATLAB.
|Literature and Study Materials:
||Course notes "Model Predictive Control" by Ton van den Boom (TU Delft) 2013.
||Written exam and a homework assignment
||Computer use: for the homework assignment, the use of MATLAB on PC is required. The assignment can be done either at home or at the DCSC laboratory.