Sustainable autonomous grid-constrained energy management for data centers using battery storage
Vigh, David (2026)
Aineistoon ei liity tiedostoja.
Diplomityö
Vigh, David
2026
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Kuvaus
This thesis is available in the LUT University archive. Contact: asiakirjat@lut.fi'
Tiivistelmä
There has been a growth in demand for data centers, but grid limitations due to congestion have become a roadblock to further growth. The new regulations aim to better utilize existing infrastructure. This creates an obligation for data centers to become active flexibility providers for the electricity grid without losing compute availability.
This thesis addresses that challenge by designing and evaluating an autonomous Energy Management System for a grid-constrained, sustainable data center, treating the compute load as fixed so that all flexibility comes from external assets: a battery energy storage system, on-site photovoltaics, and a backup diesel generator. The system is structured as a three-layer, self-adaptive architecture, a twice-daily day-ahead program, an intraday re-planner, and a component-control layer, evaluated through agent-based co-simulation.
The evaluation shows that battery storage capacity, rather than the objective weighting, is the decisive lever: served load and carbon savings both grow with battery size, while onsite photovoltaics mainly reduce grid-import dependency. The backup generator restores availability under a constrained connection only by converting Scope 2 grid emissions into larger Scope 1 on-site emissions, and intraday replanning gives no measurable benefit across random seeds. This configuration successfully demonstrates possible improvements in grid resilience, a reduction in operational carbon emissions, and lower wholesale electricity costs.
This thesis addresses that challenge by designing and evaluating an autonomous Energy Management System for a grid-constrained, sustainable data center, treating the compute load as fixed so that all flexibility comes from external assets: a battery energy storage system, on-site photovoltaics, and a backup diesel generator. The system is structured as a three-layer, self-adaptive architecture, a twice-daily day-ahead program, an intraday re-planner, and a component-control layer, evaluated through agent-based co-simulation.
The evaluation shows that battery storage capacity, rather than the objective weighting, is the decisive lever: served load and carbon savings both grow with battery size, while onsite photovoltaics mainly reduce grid-import dependency. The backup generator restores availability under a constrained connection only by converting Scope 2 grid emissions into larger Scope 1 on-site emissions, and intraday replanning gives no measurable benefit across random seeds. This configuration successfully demonstrates possible improvements in grid resilience, a reduction in operational carbon emissions, and lower wholesale electricity costs.