Energy System Optimization and Carbon Analysis of Solid Oxide Fuel Cells in AI Data Centers

Artificial intelligence has turned data centers into one of the fastest-growing loads on the grid. The International Energy Agency projects data-center electricity demand growing around 15% a year, approaching 1,000 TWh — roughly 3% of global electricity — by 2030, four times the growth rate of the wider economy. These facilities need power that is not only low-carbon but continuously reliable, and increasingly independent of a strained central grid. This new NTU-Delta Electronics Innovation Research Funding project (2026–2027), led by the E3 Center in collaboration with Delta Electronics, asks how best to meet that demand.

The project centers on the solid oxide fuel cell (SOFC). Unlike the proton-exchange-membrane cells usually studied as short-duration backup, an SOFC combines high electrical efficiency with steady, long-duration operation and high-temperature waste heat that can be recovered for cooling — making it well suited as baseload power for a data center. It is also fuel-flexible, able to run on natural gas, hydrogen, or a blend, which makes it deployable today while the hydrogen economy matures. Most existing work looks at a single technology or a pure-hydrogen PV route; this project instead treats the data center as a multi-energy system, co-optimizing SOFC, solar PV, storage, and the grid, with hydrogen/natural-gas blending, combined-heat-and-power cooling via absorption chillers, and a comparison of two renewable pathways — PV used directly versus PV used to electrolyze hydrogen for the fuel cell.

The work pursues three objectives. First, to build a data-center multi-energy configuration optimization model integrating SOFC, PV, storage, and grid, with generation cost, carbon targets, and supply-reliability constraints — solved with a genetic algorithm (NSGA-II) to trace the cost-versus-carbon trade-off and find the optimal natural-gas/hydrogen blend. Second, to analyze how renewable configuration and the SOFC supply ratio shape system benefits — running full-year hourly simulations, comparing the PV-direct and PV-to-hydrogen paths, and quantifying how much of the cooling load SOFC waste heat can serve. Third, to assess how carbon-pricing mechanisms reshape the transition pathway — evaluating lifecycle cost, levelized cost of energy, net present value, and payback under different carbon-fee and electricity-price scenarios.

Together these produce a decision-support framework — spanning capacity sizing, hourly dispatch, and techno-economic evaluation — that data-center operators can use to plan a phased, low-carbon transition. As a cross-appointment between civil engineering (computer-aided engineering) and chemical engineering, and grounded in industry practice through Delta Electronics, the project is positioned to strengthen Taiwan's capacity in low-carbon data centers, fuel-cell and hydrogen deployment, and smart-energy-system integration as AI-driven electricity demand continues to climb.