Digitally optimizing energy assets for decarbonization and cost savings

Overview

SSE Energy Solutions, a UK/Ireland renewable-focused energy company, replaced a manual virtual-machine-based ML workflow with the Databricks Data + AI Platform on Azure to run machine learning optimization models for its combined heat and power (CHP) networks, including digital twins and demand forecasting. The move cut time spent on manual forecasting by 12x and is predicted to save £350K per year through energy optimization.

The challenge

SSE Energy Solutions initially ran its ML-powered CHP optimization models on virtual machines, which required large amounts of manual input and was a hugely inefficient use of data scientists' time as the energy generation mix grew more complex.

The solution

SSE Energy Solutions moved its ML workflow optimization model to Azure Databricks, feeding into edge optimization of four of SSE Heat Network's CHP-powered networks across the UK. The team automated the ETL of historic and current pricing regime data, created digital twins using historic demand data to simulate and predict future behavior, and used clustering and classification techniques to simplify optimization results for the edge optimization framework, with results fed via Power BI integration into dashboards for stakeholders.

Reported business value

A forecasting job that once took a day a week now takes no more than two hours, and the company predicts £350K in savings per year from energy optimization. The team now runs weekly optimization jobs across each CHP network site, able to spin up and schedule multiple runs simultaneously, and has moved from near-year-ahead optimization to week-ahead optimization.