Accurately forecasting gas flows for a global energy trader

Customer

A global energy / commodity trader.

Problem

The European Gas Network is made up of multiple interconnected regional systems each with its own price, physical constraints and dynamics. Predicting the flow of gas between these interconnectors informs the supply of gas and price movements. Our client wanted to develop tools and capabilities to help their traders better understand these dynamics.

Solution

We developed a machine learning model to forecast interconnector flows and integrated the models onto our client’s data and analytics architecture. The models also have built-in human insight that is typically manually adjusted by experts and enables the analysts to run multiple pricing scenarios and obtain immediate predictions.

Impact

The project proved the value of machine learning over their existing techniques, creating a step change in predictive power and usability with potential to scale across Europe.