Using machine learning, a team of researchers from Stanford, MIT, and Toyota has cut battery-testing times by 15fold.
When it comes to advancements in electric vehicles (EVs), nothing is as fundamental as improving battery life, which determines everything from how far you can drive to how long the car will last.
On Tuesday, battery testing for EVs surged forward, thanks to the work of a Stanford-led research team––which used AI to cut testing times by nearly 15fold.
Led by Stanford professors Stefano Ermon and William Chueh, the research team, from Stanford, MIT, and the Toyota Research Institute, wanted to come up with the best way to charge an EV battery quickly, in 10 minutes, and still maximize the overall battery life.
SEE: Cheat sheet: Artificial intelligence (free PDF) (TechRepublic)
The study, which the team published in Nature on Tuesday, shows how the patented AI program could predict the different ways that batteries would react to charging methods.
According to Stanford: “The software also decided in real-time what charging approaches to focus on or ignore. By reducing both the length and number of trials, the researchers cut the testing process from almost two years to 16 days.”
Fast-charging often strains battery life. To find the best way to prevent this, the team used AI to help sort through the various charging trials.
The machine learning system, which was trained on data of batteries that failed, was able to detect patterns to predict how long batteries would last. This way, researchers wouldn’t have to fully charge and recharge every battery until it failed.
This resulted in a new fast-charging protocol, which showed how charging with a high current midcycle could optimize battery life.
Using AI in battery testing is a new approach, according to the researchers.
“When talking to material scientists and people who work in batteries for a living, we realized that nobody was actually using more sophisticated AI in this space, so we thought it was promising,” Ermon, a professor of computer science at Stanford, said in an interview.
There are many different ways to charge a battery.
“You can apply different voltages, different currents, different intensities––they may all charge the battery in the same amount of time, but some might harm the internal components of the battery,” he said. “Depending on what kind of charging protocol you use, that can significantly affect the life of the battery.”
But with this new technology, the charging protocol has improved. “The Teslas, the Toyotas, the Volkswagens can make investments in this space and potentially use this technology to come up with better solutions,” he added.
The findings have implications for future research and could “accelerate nearly every piece of the battery development pipeline: from designing the chemistry of a battery to determining its size and shape, to finding better systems for manufacturing and storage,” according to Stanford, which could extend beyond batteries to other types of energy storage.
Machine learning-optimization such as the kind used in this research can also be applied to other areas, ranging from drug development to optimizing the performance of X-rays and lasers, according to Stanford’s release.
“I’m excited about the use of it from an AI perspective. This current challenge of using AI in the scientific discovery process is forcing us to think about some very deep problems,” Ermon said, “figuring out how to incorporate new knowledge, induction and deduction.”
“Doing science is hard,” he added, “so figuring out how to get a machine to help with the process is challenging the state of the art.”
Also see
Source: TechRepublic
