Energy

AI slashes time needed to accurately predict cycle life of batteries

AI slashes time needed to accu...
Machine learning can reduce the testing time needed for new batteries by years
Machine learning can reduce the testing time needed for new batteries by years
View 1 Image
Machine learning can reduce the testing time needed for new batteries by years
1/1
Machine learning can reduce the testing time needed for new batteries by years

Over the past few decades, batteries have become much more efficient, but gauging their service life is still extremely difficult and time-consuming. To better predict how long batteries will last and target them for the appropriate devices, MIT and the Toyota Research Institute (TRI) have teamed up to employ artificial intelligence to accurately determine battery life without requiring years of testing.

While there have been a lot of advances in battery technology, this very success is making it more difficult to push that technology forward. When batteries have a short life, it's easy to do thorough testing, but as battery life becomes longer and longer, it can take years to do proper testing by charging and discharging batteries and ascertain just how long that life is.

To help speed things up, the MIT/Toyota team brought AI to bear as a way to accurately predict battery life within 9 percent with 95 percent accuracy by training it with a few hundred million data points and looking at voltage decline and other factors among the early charging cycles. According to the team, it was possible to determine if a battery has a long or short life by looking at only five charge/discharge cycles.

The new machine learning method and its publicly available dataset could be used to speed up the development of new batteries while bringing down costs for both research and production, especially in battery assembly or formation. It could reduce the time needed to validate new batteries and make it easier to sort batteries into grades and target them at the customers who need them. It would also be possible to determine if battery packs have enough life left in them for recycling.

In addition to this, the prediction method could help to optimize charging by reducing charge times to as low as 10 minutes and battery optimization time could be reduced by a factor of 10.

"For all of the time and money that gets spent on battery development, progress is still measured in decades," says Toyota's Patrick Herring. "In this work, we are reducing one of the most time-consuming steps – battery testing – by an order of magnitude."

The research was published in Nature Energy.

Source: Toyota

0 comments
There are no comments. Be the first!