(260c) Data-Driven Robust Hydrogen Infrastructure Planning Towards Heat Decarbonization through a New Hybrid Decomposition Method

Given the increasing number of countries committing to a “Net-zero” emissions target by 2050, hydrogen as a low-carbon alternative of natural gas will become a significant player in advancing sustainable decarbonisation pathways [1]. Nowadays, there is a pressing necessity to develop a new hydrogen infrastructure network in the UK satisfying the increasing hydrogen demand to meet the ambitious target. Even though the hydrogen infrastructure planning problems on a nationwide scale have been widely explored in the literature [2,3], they ignore the demand uncertainty inherent to underlying problem, which could bring significant economic loss and even jeopardise the system security. Hence, the uncertainty-resilient schemes are required to alleviate these disadvantages. The common two ways to hedge against the uncertainty in the studies are stochastic programming (SP) [4] and robust optimization (RO) techniques [5]. However, often it can be challenging to get the probability distribution of uncertainties required in SP. On top of that, even if the stochastic process governing the uncertainty is fully or partially recoverable, the number of scenarios needed for its accurate representation tend to computationally intractable problems even for systems of moderate size. Thus, robust optimisation as an effective alternative offers a practical trade-off between feasibility and computational tractability.

In this work, we focus on the spatially-explicit hydrogen infrastructure planning problem with 5-year steps 2035-2050 and hourly resolution. This kind of complex energy-planning model typically involves representative days to alleviate the computational complexity considering their multi-scale nature. The issue of systematically accounting for the uncertainty introduced through the deployment of representative days within the planning models remains largely unexplored. To this end, we employ a data-driven adaptive robust mixed-integer linear programming optimization framework, where the demand uncertainty is captured through the introduction of uncertain representative days [6,7] and polyhedral uncertainty sets [8]. In the two-stage adaptive robust optimization (ARO, min-max-min problem), decisions are made on the “wait and see” basis, where the operational decisions like the hydrogen transmission flowrates across regions can be adjusted after the uncertainty realizations. This problem is NP-hard and cannot not be solved directly using off-the-shelf solvers. Therefore, we propose a new hybrid decomposition algorithm based on the column and constraint generation algorithm and block coordinate descent methods to handle it, which avoids the introduction of big-Mlinearizations [9], and can achieve a higher computing efficiency. Numerical results illustrate the effectiveness of the proposed framework and method with reductions in computational time exceeding 65% compared to monolithic approaches that are largely employed in the literature. Using as case study the decarbonisation of the heat sector in Great Britain, we further compare the two-stage ARO with the single-level static RO where decisions are made “here and now” without uncertainty feedback. The results show that ARO can lead to a lower system cost, and verify the advantages on controlling conservatism of ARO whilst allowing us to elucidate the added value of different flexibility options.

References

[1] R. Lowes, B. Woodman, Disruptive and uncertain: Policy makers’ perceptions on uk heat decarbonisation, Energy policy 142 (2020) 111494.

[2] M. Moreno-Benito, P. Agnolucci, L. G. Papageorgiou, Towards a sustainable hydrogen economy: Optimisation-based framework for hydrogen infrastructure development, Computers and Chemical Engineering 102 (2017) 110–127.

[3] S. K. Kamarudin, W. R. Daud, Z. Yaakub, Z. Misron, W. Anuar, N. N. Yusuf, Synthesis and optimization of future hydrogen energy infrastructure planning in Peninsular Malaysia, International Journal of Hydrogen Energy 34 (5) (2009) 2077–2088.

[4] J. R. Birge, F. Louveaux, Introduction to stochastic programming, Springer Science & Business Media, 2011.

[5] C. Shang, X. Huang, F. You, Data-driven robust optimization based on kernel learning, Computers & Chemical Engineering 106 (2017) 464–479.

[6] V. M. Charitopoulos, M. Fajardy, C. K. Chyong, D. M. Reiner, The impact of 100% electrification of domestic heat in great britain, Iscience 26 (11) (2023) 1–12.

[7] C. L. Lara, J. D. Siirola, I. E. Grossmann, Electric power infrastructure planning under uncertainty: stochastic dual dynamic integer programming (sddip) and parallelization scheme, Optimization and Engineering 21 (2020) 1243–1281.

[8] C. Ning, F. You, Data-driven decision making under uncertainty integrating robust optimization with principal component analysis and kernel smoothing methods, Computers and Chemical Engineering 112 (2018) 190–210.

[9] R. A. Jabr, Robust Transmission Network Expansion Planning With Uncertain Renewable Generation and Loads, IEEE Transactions on Power Systems 4 (28) (2013) 4558–4567.

Acknowledgements

Financial support under the EPSRC grant EP/T022930/1 is gratefully acknowledged by the authors.