Uncertainty-Aware Optimisation of PEM-Integrated Hybrid Energy Storage for Data Centres: Cost, Carbon, and Resilience Trade-offs
Preprint
- 28 July 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Hybrid energy storage combining lithium-ion batteries with proton exchange membrane (PEM) electrolysers can support low-carbon data-centre power, but its design is sensitive to coupled weather, demand and market uncertainty. We develop an uncertainty-aware, scenario-based mixed-integer linear programming (MILP) framework for a PEM-electrolyser-integrated hybrid energy storage system (HESS) serving mission-critical loads. The model enforces one non-anticipative sizing decision across 100 scenarios, including 95 representative scenarios and 5 stress scenarios, under three objectives: capital expenditure (CAPEX), total annualised cost (TAC) and carbon dioxide (CO₂) emissions.The results show three main findings. First, the accuracy cost of temporal aggregation is objective-dependent: the design-relevant error is objective-value deviation, not load-reconstruction root mean square error (RMSE). The emissions objective is most sensitive, with deviation from the full 8760-hour reference falling from approximately 8% at 24 typical days to approximately 2% at 36 typical days, while CAPEX deviation remains approximately 1%.Second, cost-optimal infrastructure shifts through regimes rather than scaling smoothly. Moderate CO₂ reduction is achieved by re-dispatch, intermediate reduction by expanding photovoltaic (PV) generation and battery capacity, and deep reduction by overbuilding PV, reducing grid imports and activating the hydrogen chain.Third, the optimisation objective determines the resilience profile. CAPEX-minimising designs are most exposed to grid curtailment, while emissions-minimising designs are most exposed to PV loss. The framework therefore reveals a cost–carbon–resilience trade-off hidden by single-scenario sizing.Keywords
This publication has 35 references indexed in Scilit:
- Design of distributed energy systems under uncertainty: A two-stage stochastic programming approachApplied Energy, 2018
- Electrochemical conversion technologies for optimal design of decentralized multi-energy systems: Modeling framework and technology assessmentApplied Energy, 2018
- Optimal design of multi-energy systems with seasonal storageApplied Energy, 2018
- Robust optimal design of renewable energy system in nearly/net zero energy buildings under uncertaintiesApplied Energy, 2017
- Multi-objective design and operation of Solid Oxide Fuel Cell (SOFC) Triple Combined-cycle Power Generation systems: Integrating energy efficiency and operational safetyApplied Energy, 2017
- Medium-term energy hub management subject to electricity price and wind uncertaintyApplied Energy, 2016
- The life cycle assessment of a UK data centreThe International Journal of Life Cycle Assessment, 2015
- A two-stage stochastic programming model for the optimal design of distributed energy systemsApplied Energy, 2013
- Adjustable robust solutions of uncertain linear programsMathematical Programming, 2004
- Robust solutions of Linear Programming problems contaminated with uncertain dataMathematical Programming, 2000