Spatio-Temporal Computing–Power–Carbon Co-Optimization of Shared Data Center Clusters in High-Renewable Nordic Microgrids

Abstract

The increasing penetration of renewable energy and the growth of digital electricity demand expose high-renewable Nordic microgrids to coupled uncertainty in renewable output, electricity prices, carbon intensity, and load demand. This coupling intensifies the operational trilemma among low-carbon supply, economic efficiency, and system reliability. Conventional microgrid scheduling methods usually represent data centers as fixed electricity loads or simplified flexible loads, and therefore cannot adequately describe workload-driven electricity demand, cross-site workload transfer, or flexibility delivery under day-ahead and intraday uncertainty. Targeting the key scientific problem of how to coordinate shared data center flexibility and microgrid resources under renewable, market, carbon, load, and workload uncertainty to mitigate this operational trilemma, this thesis develops a spatio-temporal computing–power–carbon co-optimization framework for high-renewable Nordic microgrids.

First, this thesis focuses on the dispatchability of shared data center clusters. Interactive and batch workloads are modeled according to their quality-of-service requirements. Task-level workload features are mapped into electricity demand, and computing-service execution is converted into site-level electricity consumption. Temporal deferral, spatial migration, communication limits, computing-capacity constraints, and service-feasibility requirements are formulated to characterize the dispatchable boundary of shared data center flexibility.

Second, this thesis formulates a coordinated day-ahead and intraday computing–power –carbon co-optimization model. The model coordinates workload scheduling, renewable generation, storage operation, grid exchange, carbon-emission accounting, and flexibility-service provision. Scenario-based stochastic scheduling is used to form day-ahead operating commitments under renewable, market, carbon, load, and workload uncertainty. CVaR-based risk control is introduced to account for unfavorable scenario losses, and intraday rolling dispatch is used to update operating decisions as new information becomes available. In addition, a scenario-based uncertainty modeling method is constructed through conditional scenario generation, representative scenario reduction, and probability weighting.

Finally, a high-renewable Nordic microgrid case study is conducted to evaluate the proposed framework. The assessment covers operating cost, service-settlement performance, carbon emissions, renewable utilization, day-ahead/intraday deviation, flexibility support, trilemma-oriented performance, and sensitivity analysis. Relative to a case where SDC electricity demand is treated as site-specific local load without cross-site workload transfer, the integrated co-optimization case reduces the final net operating cost by 0.65 %, total carbon emissions by 1.44\%, DA–RT mismatch by 7.18 %, SOC limit-risk exposure by 21.8 %, and the share of high-carbon electricity imports by 17.99 %. These results indicate that SDCs can be represented and scheduled as a service-preserving form of demand-side flexibility. By reshaping the timing and location of computing-related electricity demand while maintaining workload-service feasibility, SDCs offer a computing-based demand-response pathway for mitigating the operational trilemma among economic efficiency, carbon-emission reduction, and operational reliability under high renewable penetration.

Popular Abstract

As renewable energy becomes more widely used, power systems need to remain clean, affordable, and reliable despite changing weather, electricity prices, carbon intensity, and demand. This thesis explores how shared data center clusters can help address this challenge in high-renewable Nordic microgrids.

Data centers consume large amounts of electricity, but some computing tasks are flexible. They can be delayed or moved between sites without affecting service quality. This means that data centers can act as demand-side flexibility resources rather than only passive electricity consumers.

This thesis develops a computing–power–carbon co-optimization framework to coordinate data center workloads with renewable generation, energy storage, grid exchange, carbon signals, and market operation. The framework considers both day-ahead planning and intraday adjustment under uncertainty.

The case study shows that shared data center flexibility can help reduce operating costs, carbon emissions, mismatches between planned and actual operation, and high-carbon electricity imports. Overall, the results suggest that shared data center clusters can support cleaner, more economical, and more reliable operation of renewable-rich power systems.

• author: Lan, Xinyao
• supervisor: Huan Li
• organization: Division for Industrial Electrical Engineering and Automation
• course: EIEM02 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: shared data center clusters, high-renewable Nordic microgrids, computing–power–carbon co-optimization, demand-side flexibility, stochastic scheduling, Conditional Value at Risk
• publication/series: CODEN:LUTEDX/TEIE
• report number: 5566
• language: English
• id: 9239268
• date added to LUP: 2026-09-10 09:32:09
• date last changed: 2026-09-10 09:32:09