A hydrogen production facility optimizes its operations against local electricity prices, local hydrogen demand, and its own power purchase agreement. Individually, each site makes reasonable decisions. Collectively, those individually reasonable decisions leave money on the table.
Centralized portfolio-level control across geographically distributed hydrogen assets yields a 2.42-fold increase in hydrogen production and a 9.4% reduction in daily operational costs compared to site-level optimization. The improvement comes from temporal and spatial arbitrage: one site's cheap electricity hours overlap with another site's expensive ones. One site's excess hydrogen can satisfy another site's demand shortfall. The portfolio sees opportunities that no single site can.
Note (2026-11-15; read against the paper's full text): the 2.42-fold and 9.4 percent figures are comparisons among three set-ups for one five-site company (H2FLEX), not a comparison of pure site-level optimization with portfolio optimization of the same problem. In Case 1 each electrolyzer operates independently with its own power purchase agreement and an individual, uniform green hydrogen target; in Case 2 the PPAs are centrally dispatched and one policy constraint is applied at the portfolio level (daily net costs 7.9 percent lower than Case 1); in Case 3 a second constraint also moves to the portfolio level (9.4 percent lower than Case 1). The paper attributes the cost reduction in Case 2 mainly to centralized management of the PPAs. I found no discussion of temporal or spatial arbitrage between sites (no hit for arbitrage), so that mechanism is the essay's own, and the essay's closing claim that wider coordination is 'always' the fix is a generalization beyond a one-day, five-site case study.
The flexibility is already in the system. Every site has electrolyzers that can ramp up or down, storage that can buffer production, and grid connections that can import or export. The constraints are the same. The difference is who makes the decisions. When each site optimizes independently, it uses its flexibility to minimize its own costs. When the portfolio optimizes jointly, it uses one site's flexibility to serve another site's needs — and the total benefit exceeds the sum of individual optima.
This is a coordination premium, not a technology premium. The hardware doesn't change. The physics doesn't change. What changes is the scope of the optimization — from plant-level to portfolio-level — and that scope change reveals value that was invisible to the individual optimizer.
The general pattern: distributed assets with independent objectives underutilize their collective flexibility. The fix is always wider coordination, never better local optimization.