AIRGo - I-NERGY - Training script
Docker image containing the script to execute a Grid2Op agent training while using the PowSyBl Backend on a dataset.
This Docker provides the code in order to demonstrate the capacity to execute a Grid2Op agent training which uses the PowSyBl Backend (https://github.com/powsybl/pypowsybl-grid2opbackend). This training uses the dataset provided in the AIRGo - I-NERGY - Open Dataset asset and available here: https://www.ai4europe.eu/research/ai-catalog/airgo-i-nergy-open-dataset. This dataset must be unzipped into the "src/data_set" directory.
The structure of the Python script clarifies the use of the backend in this scenario.
This asset is part of a project that has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101016508.
This Jupyter Notebook corresponds to the following user story: as an AIRGo AI scientist, I want to train and validate my agent thanks to a more realistic network simulation provided by PowSyBl, so that my model actions fit real life network conditions.
This AIRGo AI scientist interacts with the grid2Op module while launching a train. The Grid2Op module executes the train. The train has a limited number of iterations and epochs. For each iteration, the PowSyBl module will be called to make the state evaluation used by the Grid2Op module.