Background

In decentralized energy systems, flexibilities are essential to prevent grid congestion and balance fluctuations in energy generation. A central challenge lies in efficiently scheduling flexible resources such as battery storage systems, heat pumps, or electric vehicles, particularly when unexpected grid demands or communication failures occur.

Reactive scheduling provides a promising approach to addressing such events. Agents make decentralized decisions based on local information and, if necessary, with limited communication. Various algorithms can be utilized to manage reactive planning, including heuristic methods or algorithms from the field of distributed artificial intelligence.

Objective

The objective of this thesis is to design and implement a distributed, reactive approach to scheduling in a multi-agent system. Two different algorithms will be developed, implemented, and compared. The evaluation will focus on the efficiency, responsiveness, and robustness of the approaches in simulated scenarios.

1. Familiarization with the topic (understanding the basics of flexibility, basics of deployment planning in decentralized energy systems, research on distributed algorithms and their application in multi-agent systems)
2. Conception (development of two concepts for distributed, reactive deployment planning algorithms)
3. Implementation (implementation of the two algorithms in a programming language (e.g. Python) in the agent framework mango)
4. Evaluation (comparison of the algorithms using simulations with realistic scenarios)

Interest in decentralized energy systems and algorithms for distributed systems

Optional:

• Basic knowledge of algorithmic design and programming
• Initial experience with multi-agent systems or simulation tools