Large-Scale Spatial Optimization for Industrial Decarbonization

Spatiotemporal optimization, 2024 to 2026. Manuscript under second-round review at Nature Energy.

The problem

Steel, ammonia, and methanol are the industries usually described as hard to abate, and the usual analysis treats them as a sector: a national average cost of clean hydrogen, a national date at which it becomes competitive. Plant operators do not make national decisions. They make one decision, at one location, against the resource quality and grid prices they actually face, and the answer varies enormously between two plants a few hundred kilometers apart.

What I built

A plant-level optimization framework covering 548 facilities, 203 steel, 158 ammonia, and 187 methanol plants that together account for 2,091 Mt of CO2. Each plant is matched against hourly renewable resources at 25 by 25 km resolution, and the model co-optimizes energy sourcing, electrolysis capacity, storage, and grid purchases for that plant. On top of the supply optimization I screened technology pathways plant by plant, comparing clean hydrogen against carbon capture and against electric arc furnace routes, so the output is a least-cost technology choice for each facility rather than a sector-wide recommendation.

What I found

Two thirds of these plants, 66%, can run on fully off-grid renewables sited within 200 kilometers. Allowing a small grid connection changes the picture more than adding more renewables does: with roughly 8% of electricity drawn from the grid, the share rises to 79%, and average hydrogen cost falls by 29%. The grid connection is doing work that storage would otherwise have to do, and it is doing it far more cheaply.

The decision this supports is which plants to target first. The least-cost pathway is not uniform across a sector, and a policy that subsidizes one technology nationally will overpay at some plants while leaving others short.

Research figure

The underlying plant-level dataset distinguishes the location, scale, age, and emissions profile of individual facilities.

Limits

The analysis optimizes supply to each plant and screens technology pathways for it. It does not model the wider power system’s response to that new demand, hydrogen transport between regions, or the capital replacement cycle that determines when a given plant is actually open to rebuilding.