The impact of EURO-CORDEX climate projections on the robustness of multi-energy system design

The optimization of a multi-energy system consists of selecting the sizes of the different units in the system and determining their operation, to fulfill the energy demand while minimizing, in this case, the total cost of the system. This process is carried out on historical weather data, under the implicit assumption that past conditions still describe the ones the system meets over its lifetime. This work examines whether replacing historical data with climate projections in the sizing phase produces more robust and better performing configurations, where robustness is the ability to satisfy the energy demand without relying on external energy. It further establishes how much weight the environmental variability carries relative to the economic one, and whether accounting for variability at all improves on a deterministic benchmark. Climate projections drawn from EURO-CORDEX are used, alongside historical reanalysis data, to feed a mixed integer linear model of a multi-energy system serving an urban demand, which is sized both deterministically and stochastically. The resulting configurations are tested in operation against full-year datasets from three time windows, covering two locations, two emission reduction targets and two variability assumptions, the first isolating the climate signal and the second also including the economic one. The resulting configurations are evaluated in terms of robustness and total cost. Matching the sizing decade to the operational one proves beneficial under both variability assumptions, though the benefit shows up on different fronts. With environmental variability alone, it improves the ability to run off-grid by about one year out of the fifteen tested by each configuration, without any ordering by how distant the sizing window is, while total cost remains stable across configurations. With economic and environmental variability, the higher expected fuel prices push the designs towards renewable capacity, storage, and heat pumps. This shift raises the robustness of the systems, with the share of years operated without import growing from 66% to 82%, and it also improves their economic performance when the sizing and the operational decade are matched, by up to 34%. The deterministic design is unable to operate without external energy in 28 of the 30 cases tested, and imports and emits more than any stochastic one. Climate projections alone thus bring a limited improvement in robustness and cost, while economic variability drives a far larger gain on both fronts, and stochastic programming outperforms deterministic modeling throughout, most markedly in terms of robustness.