In the race to bring EVs to the masses, researchers have developed a machine learning algorithm that could supercharge battery development.
A collaboration between Stanford University, MIT and the Toyota Research Institute has led to the development of a new machine learning method that aims to overcome one of the biggest issues in the development of electric vehicles (EVs).
This bottleneck involved the evaluation time of EV batteries, where new technologies must be tested for months – or even years – to see how long they will last.
However, in a paper published to Nature, the research team said its new AI slashes the amount of time it takes to test EV batteries by 98pc. While the focus of the study was on battery charge speed, it can be applied to other parts of battery development and even to non-energy technologies.
Two years to 16 days
“In battery testing, you have to try a massive number of things, because the performance you get will vary drastically,” said Stefano Ermon who co-led the team. “With AI, we’re able to quickly identify the most promising approaches and cut out a lot of unnecessary experiments.”
The team’s research set out a target of finding the best way to charge an EV battery in 10 minutes, while maximising the battery’s overall lifetime. The software quickly learned how batteries would respond to different charging approaches after a few charging cycles.
By reducing both the length and number of trials, the researchers cut the testing process from almost two years to 16 days.
Peter Attia, co-lead of the study, said: “We figured out how to greatly accelerate the testing process for extreme fast charging.
“What’s really exciting, though, is the method. We can apply this approach to many other problems that, right now, are holding back battery development for months or years.”
Scientific discovery ‘may drastically speed up’
One example of the AI in action was a substantial improvement made on existing testing methods. Typically, an EV battery is charged and recharged over and over again to see when it will eventually fail.
However, this new software determined manufacturers could better predict how long a battery would last with just 100 charging cycles. This is because the machine learning system, after being trained on a few batteries cycled to failure, could find patterns in the early data that indicated how long a battery would last.
This new AI will be made freely available for any future battery scientists looking to use it, the team said. Ermon said that the biggest hope for the team’s work is to help the process of scientific discovery itself.
“We’re asking: Can we design these methods to come up with hypotheses automatically?” he said.
“Can they help us extract knowledge that humans could not? As we get better and better algorithms, we hope the whole scientific discovery process may drastically speed up.”