Facebook Using AI to Enhance Renewable Energy Storage Amid Climate Crisis

Facebook teamed up with a university in using AI to find new ways of storing renewable energy.
Brad Bergan

Facebook and Carnegie Mellon University have teamed up to use artificial intelligence (AI) to discover new "electrocatalysts" capable of enhancing storage for electricity generated via renewable energy alternatives amid the climate crisis, according to a blog post from Facebook.


Facebook, Carnegie Mellon University team up, use AI to fight climate crisis

Facebook AI has partnered with Carnegie Mellon University's (CMU's) department of chemical engineering to launch a new project called Open Catalyst Project — which aims to accelerate quantum mechanical simulations on the order of 1,000 times faster, using AI. In this way, they hope to find new electrocatalysts needed to build more efficient and scalable means of storing and using renewable energy, according to Facebook AI's blog post.

Electrocatalysts are useful because they can convert excess solar and wind power into other fuels — like ethanol and hydrogen — that are not as difficult to store. But present-day electrocatalysts are expensive and rare — especially platinum. This is why finding new ones has proved difficult — since they may be forged via billions of potential elemental combinations.

Researchers working with catalysis are testing tens of thousands of possible catalysts per year, but CMU and Facebook think they may increase this number to millions — potentially even billions — of catalysts, thanks to AI assistance, reports CNBC.

Solar Power Needs CISO
A graph showing energy needs through one day versus the generation of wind and solar power. Source: Facebook AI / California Independent System Operator

'Relaxation' calculations help distinguish good catalysts

The joint Wednesday announcement came along with a few of their own software "models" capable of enhancing the search for new catalysts — so other scientists may join the global struggle for new options.

Most Popular

Facebook said its "Open Catalyst 2020" data set needs 70 million hours of computing time to produce viable results. The dataset involves "relaxation" calculations to vet millions of possible catalysts, in addition to supplemental calculations.

Relaxations are a common measurement in catalysis, which are calculated to determine which specific combinations of elements may make a good catalyst, and which won't.

Facebook AI Simulator
An AI simulation of relaxation on a catalyst's surface, magnified at 10^8 times. Source: Facebook AI

Relaxation calculations simulate atom interaction

Every relaxation calculation — which simulates the way atoms from various elements interact — takes on average eight hours for scientists to complete, but Facebook claims AI software might complete the same calculations in less than a second.

For catalysis researchers, "that's going to dramatically change how you do your work and how you do your research," said Facebook AI research scientist Larry Zitnick, during a call preceding the announcement, CNET reports.

AI crucial for enhancing renewable energy sources

Tech superpowers like Google and Facebook have tried to use AI to enhance scientific calculations and observations in a wide spectrum of fields, to emphasize speed and efficiency.

Google's company Alphabet has an AI-lab called DeepMind, for example, which developed AI software that locates tumors in mammograms more quickly and with greater accuracy than human researchers can.

The case for renewable energy sources — whether taking the form of wind, solar, or others — is growing strong amid the impending ravages of the climate crisis. And implementing AI to help enhance and speed up our search for viable alternatives to fossil fuels is crucial, and companies like Facebook and Google are on the front lines of this struggle.