Who is Resa?

Resa owns and operates gas and electricity distribution networks in the province of Liège, Belgium. It operates around 14,000 kilometers of electricity cables and 4,000 kilometers of gas pipelines, delivering electricity to around 433,000 customers and gas to 232,000 customers

The challenge

Resa is facing an important challenge: the electricity grid was not built to support the current energy shift.

People are switching their energy sourcing from fossil fuels to electricity: cars are getting electric, heat pumps are replacing the old gas and fuel heaters,… Additionally, citizens can become prosumers meaning they produce electricity thanks to the installation of photovoltaic panels.

Our collaboration started with a simple question: how can we better assess the impact of the energy transition on our grid based on different scenarios?

Our approach

We started by building a stochastic model that concretely show the impact of different scenarios on each substation, on each cable, on each consumption point.

Then, when Resa confirmed that this approach was delivering the right insights, we went on to work on the next step : building an interface to allow users to interact with the model, define scenarios and explore the grid. The key role of the interface is to understand the local impact of the energy transition on grid assets in order to prioritise investments.

The interface was built to allow the integration of operational data and was soon extended to display smart meters data, …. It is now also leveraged in daily operations to identify incidents like stall inverters, saturation,…, its integration into the day-to-day drives both data quality and trust in its data-driven recommendations

Tech stack

The whole solution is hosted on Amazon Web Services (AWS). The backend engine consists of Scala/Spark aggregation pipelines running on Amazon EMR. The prediction and phase detection engines are running with python (using the Prophet library). Those engines are orchestrated via Airflow on Managed Workflow for Apache Airflow (MWAA).

  • The simulation engine is built with Python and runs on AWS Lambda.

  • The resulting data are written in a PostgreSQL database hosted on Amazon RDS (Aurora).

  • The API is built using nodeJS & serverless and hosted on API Gateway and AWS Lambda.

  • The front-end application is built using ReactJS. Charts are made using a combination of d3.js and Plot by Observable libraries. The application is hosted on AWS S3 and served using Cloudfront.