MPCRL for Energy Systems 🏠
This course is interested in using methods of model predictive control and reinforcement learning (MPCRL) for energy systems. In particular, we are interested in saving energy for heating systems.
General Information
- HISinOne: 11LE50P-5720 (6 ECTS Version)
- Language: English
- Email: mpcrl@cs.uni-freiburg.de
Overview
This course introduces students to the field of model predictive control and reinforcement learning (MPCRL). In this course, students first attend the MPCRL school and then have a extended project work during the semester working on energy systems more specifically heating systems.
Format
The course is based on the MPCRL school, continues with two exercises that prepare students for the project phase:
- MPCRL school: 6 Oct - 16 Oct [Link]
- Lectures and exercises, project phase can be used for starting exercise 1.
- Studienleistung: Micro exam 16 Oct
- Exercise 1: 14 Oct - 13 Nov
- Deadline is 12 Nov for submitting code
- Studienleistung: In-person written test at 13 Nov 10am to 11am @ Imbit
- Exercise 2: 14 Nov - 11 Dec
- Deadline is 10 Dec for submitting code
- Studienleistung: In-person written test at 11 Dec 10am to 11am @ Imbit
- Project Phase: 1 Dec - TBD:
- During the project phase each team of three gets assigned a tutor
- Deadline is TBD for submitting short report about project
- Prüfungsleistung: Oral group presentation on the project 20 to 30min
Learning Objectives
After successfully completing the module, students will be able to:
- explain and compare fundamental MPC and RL concepts and algorithms;
- prepare simulation data and compare algorithms using suitable benchmarks and metrics;
- integrate learned components into a closed-loop energy-system controller;
- design and implement an MPC, RL, or combined MPC-RL method;
- conduct reproducible closed-loop experiments and evaluate performance and robustness;
- identify limitations and communicate methods, results, and conclusions clearly.
Recommended Prerequisites
- Knowledge of systems and control
- Numerical optimization
- Machine learning
- and basic Python programming is recommended.
Topics
Apart from the general topic of MPCRL school: