UK Power Networks deploys a predictive machine learning model on Databricks for network load forecasting

Overview

UK Power Networks migrated from an on-premises data warehouse to the Databricks Data + AI Platform and rolled out a Network Load Forecasting and Estimation machine learning model, using Unity Catalog for access management and Delta Sharing for internal data collaboration, to predict network headroom, reduce unplanned downtime and support the UK's net-zero transition.

The challenge

UK Power Networks' on-premises solution spread data across multiple servers, made processing real-time data cumbersome and impeded timely business insights. It also brought licensing issues, limited data-center support, growing access-control complexity, siloed business units reliant on a single central data warehouse team, and external data-sharing challenges that hampered rapid experimentation.

The solution

UK Power Networks migrated essential data from its on-premises legacy data warehouse to the Databricks Data + AI Platform, prioritizing critical datasets over a use-case-specific approach. Databricks now provisions real-time smart meter alerts for power outages and voltage changes, powers a Network Load Forecasting and Estimation machine learning model, uses Unity Catalog for fine-grained access management, and uses Delta Sharing to streamline data sharing across the organization.

Reported business value

The Databricks Data + AI Platform is projected to create £45.7M in total value over the next three years, including £1M in annual value from reduced downtime. The load-forecasting model increases the quality of the Network Planning team's work, saves time and helps avoid unnecessary reinforcement, while democratized, unified data access lets the utility explore increasingly complex predictive and ML use cases.