This is the accompanying repository to the publication Approximating Energy Market Clearing and Bidding With Model-Based Reinforcement Learning by Thomas Wolgast and Astrid Nieße.
Run pip install -r requirements.txt for installation. Tested with python 3.8.
If automatic torch installation does not work, try manual installation of torch first.
Note that this repository contains outdated versions of the drl repo for and mlopf repository (warning: mlopf was re-named to opfgym). The up-to-date repositories can be found at https://gitlab.com/thomaswolgast/drl and https://github.com/Digitalized-Energy-Systems/opfgym.
Note that the focus of the mlopf library shifted from multi-agent bidding to single-agent OPF approximation environments.
Details can be found in the publication "Learning the optimal power flow: Environment design matters"
After installation, run in the src directory:
python main.py --agent "general_market_maddpg:MarketMaddpgPab" --environ "ml_opf.envs.energy_market_bidding:OpfAndBiddingEcoDispatchEnv" --hyper "{'start_train': 300, 'start_train_agents': 400}" --env-hyperparams "{'n_agents': 10}" --steps 500 --store --num 1 --test-steps 3
for the model-based MADDPG experiments.
Or
python main.py --agent "maddpg:MarketMaddpg" --environ "ml_opf.envs.energy_market_bidding:BiddingEcoDispatchEnv" --hyper "{'start_train': 300}" --env-hyperparams "{'n_agents': 10}" --steps 400 --store --num 1
for the baseline experiments.