Seminar 9 July 2025 16:00 (AEST)
Title: Modelling and analysis of multi-timescale uncertainty in energy system planning
Speaker: Dr Hongyu Zhang
(University of Southampton)
Summary:
We propose the REORIENT (REnewable resOuRce Investment for the ENergy Transition) model for energy systems planning with the following novelties: (1) integrating capacity expansion, retrofit and abandonment planning, and (2) using multi-horizon stochastic mixed-integer linear programming with short-term and long-term uncertainty. We apply the model to the European energy system considering: (a) investment in new hydrogen infrastructures, (b) capacity expansion of the European power system, (c) retrofitting oil and gas infrastructures in the North Sea region for hydrogen production and distribution, and abandoning existing infrastructures, and (d) long-term and short-term uncertainty. We utilise the structure of multi-horizon stochastic programming and propose enhanced Benders decomposition methods to solve the model efficiently. We propose: (1) stabilising Adaptive Benders with the level set method and adaptively selecting the subproblems to solve per iteration for more accurate information, (2) a centre point stabilisation approach when the level set problem is hard to solve, and (3) dynamic level set management to improve the robustness of the algorithm by adjusting the level set per iteration.
We first conduct a sensitivity analysis on retrofitting costs of oil and gas infrastructures. We then compare the REORIENT model with a conventional investment planning model regarding costs and investment decisions. Finally, four algorithms are implemented for solving LP instances with up to 1 billion variables and 4.5 billion constraints, and two algorithms are implemented for MILP instances with high degeneracy. The results show that: (1) when the retrofitting cost is below 20% of the cost of building new ones, retrofitting is economical for most of the existing pipelines, (2) compared with a traditional investment planning model, the REORIENT model yields 24% lower investment cost in the North Sea region, and (3) for a 1.00% convergence tolerance, the enhanced Benders is up to 6.8 times faster than the reference algorithm for MILP instances, and is up to 113.7 times faster than standard Benders and 2.14 times faster than unstabilised Adaptive Benders for LP instances. Also, for a 0.10% convergence tolerance, the enhanced Benders is up to 45.5 times faster than standard Benders for LP instances, and unstabilised Adaptive Benders cannot solve the largest instance to convergence tolerance due to severe oscillation. Finally, the dynamic level set management makes the algorithms more robust and is very helpful for solving large problems.
Biography:
Dr Hongyu Zhang is a Lecturer (tenured Assistant Professor) in optimisation for machine learning and AI within the Operational Research Group in the School of Mathematical Sciences at the University of Southampton. Previously, he held a permanent Researcher position in the Department of Industrial Economics and Technology Management at the Norwegian University of Science and Technology (NTNU). He has also been a visiting scholar at Princeton University, the University of Edinburgh, Imperial College London, Carnegie Mellon University, and The Ohio State University.
He received a BSc degree in Mathematics and Applied Mathematics from Huaqiao University in 2019, an MSc degree in Operational Research with Data Science from the University of Edinburgh in 2020, and a PhD degree in Operational Research from NTNU in 2024. His research focuses on: (1) stochastic programming with multi-timescale uncertainty, (2) decomposition algorithms for large-scale optimisation problems, and (3) long-term large-scale energy system planning under uncertainty.
JOIN ON ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
WED 9 JULY 2025 16:00-17:00 (AEST, Melbourne Time)