The electrification of the economy, required by increasing environmental concerns, seems to be harder to put in place in some sectors than others. Hydrogen could help further decarbornize such sectors, in which electrification is technically impossible or not viable
from a business stand-point. This requires the implementation of a reliable and robust hydrogen supply chain. The objective of this project is to analyze how a hydrogen supply chain could optimally be integrated inthe existing power network, that is how it could take maximum advantage of the existing electricity facilities while ensuring the provision of hydrogen to the market. Reinforcement Learning and MILP algorithms were ran on similar case studies to understand which approach gives the optimal policy to implement in the hydrogen and power supply chains. While MILP allows to model the physics of the problem more intuitively, the Reinforcement Learning approach was the only method that captured the time-related dynamics of the system.

The code for both the RL approach and the MILP one is available in code/ . For more instructions, go this section. If you want to know more about the project, go to presentation_documents/ and read our paper on the project.

Malik Boudiaf - mboudiaf@stanford.edu

Calvin McSweeny - calvinmc@stanford.edu

Zoe Ghiron - zghiron@stanford.edu