National Grid analyzes weather 40% better on Oracle Cloud
On Oracle Cloud Infrastructure, National Grid’s machine learning to analyze renewable energy sources is 40% more accurate versus prior solutions, and reduces query time from hours to minutes.
“Oracle Cloud Infrastructure allows us to process tens of thousands of models, so we can train our algorithms very quickly. It’s one of the best platforms in the world for the type of work we do.”
Business challenges
Everyone in Great Britain gets power when they need it. The company is working to lower the nation’s carbon emissions, with the goal of being able by 2025 to operate without any carbon-based energy for at least some of the time. Doing so will require using much more renewable energy from sources such as wind and solar, the volume of which is difficult to predict because those sources depend on weather conditions.
National Grid ESO needed machine-learning models to handle the complexity of accurately predicting the renewable energy available at any given moment. For example, instead of estimating the power from just a few hundred big power plants, there are now millions of generation sources, with each solar panel and wind turbine exhibiting its own distinct behavior. National Grid ESO needed lightning-fast computing power—for very short bursts—to run these complex machine-learning models.
Oracle just works. You can trust it. It doesn't fall over, it just does its job, and it does it really, really well.
Why National Grid chose Oracle
National Grid ESO has had a long history of developing using Oracle Database and tool sets, so the team trusted Oracle for essential workloads and data. Using Oracle Cloud infrastructure, the team could create a virtual supercomputer, which it calls Mildred, capable of running the machine learning models needed to predict Great Britain’s energy supply and demand.
Results
National Grid ESO is seeing up to 40% performance improvements for models running on Oracle Cloud Infrastructure. When the team ran its first machine-learning model on Oracle Cloud Infrastructure, it was about 40% more accurate than the previous model in production at that time. The team expected the workload to take a few hours to run, but instead it ran in minutes.
Running machine learning models on Oracle Cloud Infrastructure allows the team to see patterns in data it couldn’t before—patterns not obvious to humans. And the National Grid ESO can focus on strategic projects rather than having to manage infrastructure.
Oracle Cloud Infrastructure provides an open platform for machine learning, allowing National Grid ESO to use more than 21,000 machine-learning models, just to predict and manage the solar energy supply.
Using Oracle Cloud Infrastructure to predict energy supply, Great Britain hit a historic milestone of producing 48.5% of its electricity from renewable sources for the 12 months ending in December 2019.