Flexible computational loads as demand-side resources for surplus renewable energy: A review of siting criteria, thermal management, and operational strategies

초록

Renewable curtailment often occurs behind local transmission constraints, yet studies rarely examine whether flexible loads are located where surplus arises. This review assesses when Bitcoin mining, deferrable artificial-intelligence and machine-learning (AI/ML), and high-performance-computing (HPC) workloads can use otherwise-curtailed electricity. A PRISMA-informed search, citation tracking, targeted searches, and institutional records produced a 99-record source set. The synthesis follows five steps: verify surplus, match workload flexibility, screen site and thermal feasibility, select an operating pathway, and bound claims. The positive-part Normalized Surplus Index is proposed as a screening concept. A California Independent System Operator audit shows that balancing-area records cannot support nodal NSI without pre-curtailment availability and boundary-consistent absorption and export allocations. Siting criteria include electrical access, communications, cooling and water, land and environmental constraints, and heat sinks; local stakeholder elicitation is required for weighting. Reviewed models support rapid downward interruption of controllable Bitcoin-mining loads, whereas AI/ML and HPC flexibility is slower and schedule-dependent. Surplus absorption requires separately verified upward activation in the same location and time window. No operator-grade nodal validation, site ranking, field deployment, or grid-service qualification is demonstrated. Flexible compute is therefore a conditional demand-side resource, not a general curtailment solution. © 2026

키워드