The Potential Welfare Gains from Curtailment Trading Under Non-Firm Interconnection
Data centers facing mandatory disconnection can now trade curtailment rights—boosting total served load value by up to 83%.
Rapid data center growth is straining grid capacity, pushing operators to accept non-firm interconnection agreements for faster access. These agreements subject them to mandatory disconnection during supply shortfalls, but existing curtailment rules ignore the wide variation in value of lost load (VOLL). In response, Richard Mahuze and colleagues from Cornell propose a network-constrained Curtailment Credit Market (CCM) where agents bid for bilateral credit flows, subject to transmission constraints. They prove the CCM can achieve any allocation an omniscient central planner could—essentially no loss of allocative capability. The bilevel clearing problem is reformulated as a single-level MILP, solved in 0.01 to 83 seconds depending on scale.
Numerical experiments on three test systems (3-bus toy, IEEE 24-bus, reduced New York grid) demonstrate that the CCM increases total value of served load by 1.24 to 1.83 times compared to pro-rata curtailment. Importantly, under incentive-compatible benchmark payments, no participant is worse off than under the administrative baseline. This mechanism directly addresses the inefficiency of blunt disconnection rules, enabling data centers to value their load appropriately and trade curtailment risk, potentially unlocking billions in welfare gains as AI and cloud demand surges.
- CCM matches the allocative efficiency of an omniscient central planner (feasible-set equivalence).
- Clearing solved as a MILP in 0.01–83 seconds across test systems.
- Total value of served load increases by 1.24x–1.83x vs. pro-rata curtailment; no participant loses.
Why It Matters
A market-based solution for data center curtailment could save billions in lost compute value and accelerate grid interconnection.