The rapid growth of data centers and electrified industrial facilities presents unprecedented challenges for power grid planners. Both demand types consume enormous quantities of electricity, yet their optimal placement remains poorly understood. Siting decisions made today will lock in infrastructure investments and operational constraints for decades.
Researchers have developed a sophisticated bilevel optimization framework to address this challenge. The approach treats load allocation as a strategic planning problem where placement decisions cascade through grid expansion requirements and real-time operational constraints. Rather than treating infrastructure as static, the model captures how placement choices influence the need for transmission upgrades and affect system reliability.
The methodology employs Bayesian optimization to efficiently navigate this complex problem space. By treating detailed grid models as a black box, the approach avoids the computational intractability of traditional nested optimization methods while maintaining physical accuracy of grid operations.
Testing on a synthetic grid resembling the Texas ERCOT system revealed important dynamics. Large industrial loads like electrified refineries tend to concentrate in specific regions close to existing industry clusters. Data centers, conversely, should be distributed across the grid to avoid compounding congestion in areas already stressed by industrial demand. Under high-load scenarios, this spatial separation becomes even more pronounced, suggesting that blindly co-locating these load types could trigger expensive transmission expansion or reliability issues.
The framework explicitly balances strategic planning horizons against operational realities—a critical insight for regional transmission organizations managing both long-term network development and day-to-day grid operations. As electrification accelerates and data center capacity continues expanding, data-driven siting methodologies become essential infrastructure planning tools.
These results have immediate relevance for utilities and planners across North America and globally. The approach provides a quantitative basis for siting negotiations, tariff design, and transmission investment prioritization.
