EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
712. Day-ahead lot-sizing under uncertainty: An application to green hydrogen production
Invited abstract in session WB-9: Production Optimization and Supply Chain Management of Green Hydrogen under Uncertainties, stream Energy Markets.
Wednesday, 10:30-12:00Room: 10 (building: 116)
Authors (first author is the speaker)
| 1. | Victor Spitzer
|
| LISN, Université Paris Saclay | |
| 2. | Céline Gicquel
|
| Laboratoire Interdisciplinaire des Sciences du Numérique, Université Paris Saclay | |
| 3. | Evgeny Gurevsky
|
| LS2N, University of Nantes | |
| 4. | François Sanson
|
| Lhyfe |
Abstract
Green hydrogen production via water electrolysis powered by renewable energy is essential to decarbonize key industries. However, it poses challenges such as ensuring hydrogen demand despite the uncertain availability of renewable energy sources. This work focuses on short-term electrolytic hydrogen production planning, considering a real case with a production site linked to a wind farm and the electricity grid. The wind farm provides fluctuating renewable electricity at negligible cost, while grid electricity incurs higher cost. The purchase of electricity from the grid has to be planned and declared a day ahead of production, i.e. before the exact availability of wind power is known. On-site hydrogen storage offers some flexibility for production planning, as it allows to produce hydrogen in advance.
The future availability of the wind power source is informed by forecasts subject to error, leading to a production overcost. A cohesive framework is introduced to handle this case study. First, a two-stage stochastic programming approach is presented to model this problem. Then, a probabilistic neural network is used to estimate the conditional wind power uncertainty and generate scenarios from the knowledge of past forecast errors. Finally, a time-efficient Benders decomposition approach is proposed, in which special features of our problem are exploited to speed up the resolution. A realistic simulation demonstrates the benefits of the presented approach.
Keywords
- OR in Energy
- Stochastic Optimization
- Machine Learning
Status: accepted
Back to the list of papers