Oct – Nov 2026, NCI group project, with EirGrid and Open-Meteo data
Carbon-aware scheduling of data-centre workloads on the Irish grid
Data centres used about 23% of Ireland's metered electricity in 2025. Can weather forecasts predict clean hours a day ahead, and how much CO₂ do we save by shifting flexible jobs into them?
r = −0.75
CO₂ intensity vs wind share, over 21,241 hours
Built the dataset from EirGrid's dashboard and 24-hour-ahead wind forecasts for four wind-farm regions (Jun 2024 – Nov 2026). Trained forecasters with and without weather features, accounted for their own training energy with CodeCarbon, explained them with SHAP, and simulated shifting batch jobs into predicted low-carbon hours.
Python, pandas, SHAP, CodeCarbon, time-series forecasting
Code on GitHub





