Google DeepMind trains machine learning system to predict the power generated by wind farms 36 hours in advance

clock • 2 min read

DeepMind-developed system can increase the value of wind energy by about 20 per cent, Google claims

Google's London-based AI development subsidiary DeepMind has developed and trained a machine learning system capable of forecasting the amount of power that can be generated by wind farms - up to 36 hours in advance.

According to Google, the newly developed system helped electricity grid managers schedule set 'deliveries' of power, which are more valuable to the grid than standard, non-time-based deliveries.

Due to the variable nature of wind, it is usually difficult to accurately  forecast the amount of power that a wind farm is likely to produce during a specific time period.

But DeepMind claims that its neural network, which has been trained on historical turbine data and weather patterns, can forecast the level of power output by its wind turbines at its Oklahoma wind farm in the US, 36 hours in advance, with a reasonable amount of accuracy.

The new AI model was applied to 700 megawatts of energy produced by more than 90 turbines in Oklahoma in central US. The model accurately recommended the best way to make hourly delivery commitments to the power grid a full day in advance, thus increasing the value of wind energy by about 20 per cent.

According to Google, the system enables the wind farm's managers to schedule when to supply a specific amount of electricity to the power grid and to have a better clue of pricing models.

The system is also helpful for engineers in scheduling maintenance for turbines.

Google has recently collaborated with several wind energy firms across the world in an effort to operate completely on green energy. The goal, which was first proposed in 2016, was eventually achieved in 2017.

The company also hopes that the machine learning approach that has been implemented in wind turbine farm in Oklahoma would drive further adoption of green energy on electric grids around the world.

"Researchers and practitioners across the energy industry are developing novel ideas for how society can make the most of variable power sources like solar and wind," Google researchers wrote in a blog post.

"We're eager to join them in exploring general availability of these cloud-based machine learning strategies," they added.

The Computing and Machine 

Learning Awards 2019 logo

The AI and Machine Learning Awards are coming! In July this year, Computing will be recognising the best work in AI and machine learning across the UK. Do you have research or a project that you think deserves wider recognition? Enter the awards today - entry is free. 

You may also like
Lenovo: Liquid cooling means 30% less energy used


Shifting is 'not as complicated as people may believe', says Lenovo UK&I ISG head Ian Jeffs

clock 20 June 2024 • 4 min read
Arm tries to block Copilot+ PC lauch


Part of an ongoing dispute with Qualcomm

clock 18 June 2024 • 2 min read
Ctrl Alt Lead: Computing's new IT leader interview podcast


A quick and efficient way to learn from the best

clock 17 June 2024 • 2 min read

Sign up to our newsletter

The best news, stories, features and photos from the day in one perfectly formed email.

More on Software

The social engineering of the self: How AI chatbots manipulate our thinking

The social engineering of the self: How AI chatbots manipulate our thinking

We need structured public feedback to better understand the risks, says red teamer Rumman Chowdhury

John Leonard
clock 27 October 2023 • 4 min read
AI doesn't care what you think

AI doesn't care what you think

Want to understand hallucinations? Look at your family

Professor Peter Cochrane
clock 26 October 2023 • 3 min read
IT Essentials: The fungal IT network

IT Essentials: The fungal IT network

Shadow IT grows best in darkness and solitude

Tom Allen
clock 16 October 2023 • 2 min read