Google subsidiary, DeepMind, has developed artificial intelligence that increases the viability of energy produced by wind farms. The program uses machine learning to predict farm wind output and schedule energy deliveries in advance, which are more valuable than non-time-based deliveries.
DeepMind’s creation has major implications for Google as the company depends heavily on wind power for its green energy initiatives. With this AI, Google can send wind-generated energy to power grids with increased effectiveness.
How DeepMind’s AI Works
Given the variability and unpredictable nature of wind power, many have found it challenging to harness its full potential. The energy that wind farms generate can change day to day, making it difficult to optimize storage and distribution.
DeepMind’s program overcomes this challenge using predictive modeling and machine learning. The application makes predictions 36 hours ahead of actual wind output, thus making wind power a more reliable source of energy.
In a co-authored statement, DeepMind product manager, Sims Witherspoon, and Carbon Free Energy program lead, Will Fadrhonc said, “We can’t eliminate the variability of the wind, but our early results suggest that we can use machine learning to make wind power sufficiently more predictable and valuable.”
The DeepMind team trained its neural network system using weather forecasts. Compared to a baseline established from non-scheduled energy deliveries, Google is seeing a 20 percent gain on the “value” of wind energy produced on its wind farms. Neither Google nor DeepMind has published data that quantifies the AI program’s actual value in economic or energy terms.
Furthering Google’s Successes with Clean Energy
It’s no secret that Google is deeply committed to both effective AI use and sustainable operations. As of last year, the tech giant uses 100 percent renewable sources that offset energy consumption each year. The company’s success is attributable to renewable energy certificates, purchase contracts, and investments in wind and solar farms.
It’s currently unclear where Google aims to deploy DeepMind’s creation. The Verge speculated the Midwest as a logical choice given that the company works with wind farms throughout the region.
Much-needed Win for DeepMind
This isn’t the first time that AI developed by DeepMind has benefited Google. In 2016, Google was able to use a DeepMind AI lab product to cut data center power costs by 15 percent. With the recent news, DeepMind gets a much-needed win within a relationship that has been marked by tension given the research lab’s high cash burn rate.
Outside of the renewable energy space, DeepMind has also deployed other newsworthy AI applications. Earlier this month, an AI program crushed professional gamers at “Starcraft II.” The achievement marked a significant breakthrough in AI’s ability to compete effectively against humans in strategy-based games.
Overall, DeepMind continues to impress with its AI programs and applications. Moving forward, the lab is hopeful that it can deploy more of its innovations in real-world scenarios.