MI (1380) - INGEGNERIA DELL'INFORMAZIONE / INFORMATION TECHNOLOGY
055049 - MODEL PREDICTIVE CONTROL
Programma dettagliato e risultati di apprendimento attesi
MISSION AND GOALS
The course is aimed at providing the main ideas, algorithms, and properties of Model Predictive Control (MPC) and Moving Horizon Estimation (MHE). MPC, in particular, is the most widely used and successful advanced control method in the process industry and is nowadays also applied in distribution networks, coordination of autonomous systems, automotive, and in many other fields of application.
Lectures and computer sessions will allow students to: · Understand the challenges, opportunities, and issues related to MPC algorithms. · Understand the main common technical tools used in the analysis and design of basic and advanced MPC algorithms. · Learn how to design and practically implement standard and advanced MPC-based algorithms.
PROGRAMME OF THE COURSE
In this course the MPC basic problem formulation is first introduced, along with its main properties. Also, the tools used for guaranteeing theoretical properties (i.e., recursive feasibility and convergence) are discussed in a simplified setting. Then, the most popular industrial MPC formulations will be described.
Advanced problem formulations will be described in details, i.e., - robust and stochastic MPC formulations, with special focus on analytic and scenario-based methods; - explicit MPC and its approximated implementations; - hybrid MPC, to include logic constraints and integer decision variables; - economic MPC, that guarantees plant-wide optimality; - distributed and decentralized MPC implementations.
Constrained optimization-based state estimation will be also considered and a prototype Moving Horizon Estimation algorithm will be presented.
An overview of successful application case studies of different MPC-based algorithms will be presented.
Finally, the main computational tools for MPC implementation in both the linear and nonlinear settings will be presented. A dedicated computer session will be devoted to implementation details, possibly considering selected case studies.
Note Sulla Modalità di valutazione
The exam aims to evaluate, not only the students’ knowledge and understanding of the course topics, but also their ability to apply the acquired concepts and theoretical/practical tools. In particular, a project work consisting of the design of an MPC algorithm for a control problem chosen by the student must be completed and presented in a short oral discussion.
Intervallo di svolgimento dell'attività didattica
Calendario testuale dell'attività didattica
Monday 07/09 (14:00-18:00):
- Motivations - Introduction to MPC - Formulation and properties