Drijfhout, Matthias (2025) Renewable Energy Valley Management using DreamerV3: a multi-agent Implementation. Master's Thesis / Essay, Artificial Intelligence.
|
Text
MScThesiss3786196.pdf Download (4MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf Restricted to Registered users only Download (196kB) |
Abstract
This thesis investigates the application of multi-agent model-based reinforcement learning (MBRL) to the energy distribution problem (EDP). Building on DreamerV3, a state-of-the-art single-agent MBRL algorithm, the study develops MADreamer, an extension to multi-agent settings. The algo- rithm is evaluated in MARLOES, a simulation environment representing Renewable Energy Valleys (REVs). We present MARLOES as a flexible environment with adaptable objective functions and a playground to test RL algorithms. MARLOES models multiple assets, such as solar, wind, and stor- age, that must coordinate to balance demand and supply locally. The research highlights the theoreti- cal advantages of MBRL and empirically exposes its limitations in an EDP modelled in MARLOES. Performance is compared to a heuristic baseline, PrioFlow, which optimizes self-sufficiency. The re- sults reveal the weaknesses of MBRL and display the issues by visualizing the intended actions. This research discusses the causes of model bias, a common phenomenon in MBRL, and underlines the challenges and weaknesses of applying a model-based method to a multi-agent environment.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Grossi, D. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 30 Sep 2025 06:58 |
| Last Modified: | 30 Sep 2025 06:58 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/37066 |
Actions (login required)
 |
View Item |