Data-driven scheme for optimal day-ahead operation of a wind/hydrogen system under multiple uncertainties. (1st January 2023)

• Record Type: Journal Article
• Authors: Zheng, Yi; Wang, Jiawei; You, Shi; Li, Ximei; Bindner, Henrik W.; Münster, Marie
• Is Part Of: Applied energy. Volume 329(2023)
• Publication Date: 2023-01-01
• Subjects: Data-driven -- Robust optimization -- Day-ahead operation -- Wind/hydrogen system -- DRCCP; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2022.120201
• Languages: English
• ISSNs: 0306-2619

Abstract

Abstract: Hydrogen is believed as a promising energy carrier that contributes to deep decarbonization, especially for the sectors hard to be directly electrified. A grid-connected wind/hydrogen system is a typical configuration for hydrogen production. For such a system, a critical barrier lies in the poor cost-competitiveness of the produced hydrogen. Researchers have found that flexible operation of a wind/hydrogen system is possible thanks to the excellent dynamic properties of electrolysis. This finding implies the system owner can strategically participate in day-ahead power markets to reduce the hydrogen production cost. However, the uncertainties from imperfect prediction of the fluctuating market price and wind power reduce the effectiveness of the offering strategy in the market. In this paper, we proposed a decision-making framework, which is based on data-driven robust chance constrained programming (DRCCP). This framework also includes multi-layer perception neural network (MLPNN) for wind power and spot electricity price prediction. Such a DRCCP-based decision framework (DDF) is then applied to make the day-ahead decision for a wind/hydrogen system. It can effectively handle the uncertainties, manage the risks and reduce the operation cost. The results show that, for the daily operation in the selected 30 days, offering strategy based on the framework reduces the overall operation cost by 24.36%, compared to the strategy based on imperfect prediction. Besides, we elaborate the parameter selections of the DRCCP to reveal the best parameter combination to obtain better optimization performance. The efficacy of the DRCCP method is also highlighted by the comparison with the chance-constrained programming method. Highlights: Propose a data-driven robust chance constrained based decision framework (DDF). The DDF applied to optimal day-ahead operation of a real-life wind/hydrogen system. Multiple uncertainties from wind and price prediction are considered. Discussion on the parameter selection of the DDF.