Context
As electricity systems integrate more variable renewables, the CO₂ intensity of electricity can change significantly over the day. CO₂ intensity describes the amount of CO₂ emitted per unit of electricity generated (expressed as gCO₂eq/kWh) and helps indicate how “clean” electricity is at a given time
Challenge
Helping businesses and consumers reduce emissions requires accessible, granular and forward‑looking CO₂ intensity information so they can time electricity use (e.g., charging) when the grid is cleaner. Early forecasting approaches were limited in functionality and geographical scope, creating a need to expand forecasting tools and incorporate a broader range of data.
AI Solution
Elia Group developed eCO₂grid to measure and forecast CO₂ intensity across its grid and used Google Cloud Vertex AI (with partner Eraneos) to improve CO₂ intensity forecasting.
MLOps platform on Vertex AI integrates existing forecasting models and connects them to Elia Group’s internal data, enabling management of multiple models in parallel.
Impact & evidence
The platform enabled advanced 24‑hour CO₂ intensity forecasting models for Elia Group’s home markets (Germany and Belgium), with the stated possibility to extend to additional European electricity market areas (“bidding zones”).
The solution enables more granular and forward‑looking visibility on electricity CO₂ intensity, supporting informed decision‑making by businesses and consumers on when to consume electricity with lower associated emissions.
Implementation / Adoption
The solution is positioned for external adoption by allowing companies to access forecast data programmatically via an API, enabling integration into customer systems without detailed implementation steps or organisational change being specified.
Data
The solution combines Elia Group’s internally generated grid and forecasting data with data inputs required to model electricity CO₂ intensity, including information related to energy generation and system conditions as implied by forecasting requirements.
Future enhancements
Potential extension of models beyond Germany and Belgium to additional European electricity market areas (“bidding zones”) and wider adoption through an API for companies to integrate forecasts into their own systems.