Tim Sherstyuk (L) and Nick Sherstyuk (R) founded GBatteries and are part of a growing number of Silicon Valley scientists using artificial intelligence to overcome the limitations oflithium-ion batteries. (Handout)
Tim and Nick Sherstyuk

In a draughty basement lab, engineers Tim Sherstyuk and his father Nick set out to solve a problem: how to make smartphone batteries last longer. Two years later, the team from Ottawa had come up with an answer that could reduce the speed at which batteries degrade.

Their ultrafast charging protocol aims to charge a car battery in the time it takes to fill a tank with petrol. The key is algorithms that measure conditions inside the battery and optimise the flow of energy accordingly.

The Sherstyuks, co-founders of start-up GBatteries, are part of a growing number of scientists using artificial intelligence to overcome the limitations of lithium-ion batteries — life cycle and energy density — that are holding back the global shift to electric vehicles.

Carmakers such as Volkswagen and Tesla are starting to develop AI battery research in-house as competition to supply cleaner vehicles and transport in “smarter” cities revs up.

The market in lithium-ion batteries is expected to expand by at least tenfold in the next decade, from 151 GWh in 2019 to 1,748GWh by 2030, according to BloombergNEF. Some predictions estimate that it will be worth about $25bn by 2025. The European Commission goes further, suggesting that by 2025 the global market for lithium-ion batteries for electric cars alone could reach €40bn-€55bn a year.

As well as powering the EV revolution, batteries are a crucial component in the storage of renewable energy that is needed to shift to a greener economy. “There is already a proliferation of groups working on new batteries out there but no one is looking at how energy actually enters and affects the battery. The overall charging method has remained virtually unchanged in 100 years,” says Tim. “Few are looking at the role of algorithms in all this.”

Soldering of charging board in the GBatteries hardware lab. (Handout)

Battery innovation has historically centred on tinkering with chemistries. But commercial demand for better industrial batteries is rising. “The market is wide open for another solution. Charging time is definitely a limiting factor in a lot of applications,” he says.

At the forefront of research are “dispatch controller” algorithms. These control the rate at which energy enters a fuel cell. The faster you charge a battery, the faster it degrades, so more precise regulation makes cells last longer. For example, the Chevy Bolt electric car — with a 66kWh battery and 238-mile range — currently charges 90 miles in 30 minutes, or 15 miles in 5 minutes, “because manufacturers limit the charge rate for batteries in order to preserve their life”, says Tim. Fast-charging can enable such batteries to charge in five minutes to 50 per cent of capacity, or 119 miles range, says the start-up.

“Think of it as a mesh and you’re dumping a bunch of rice on top. Your goal is to push as much rice through the mesh as you can.

“As you start pushing some of it will go through but most will accumulate on top and eventually break it. We take that mesh and we shake it around and let the rice, or in this case ions, sift through more naturally,” says Tim.

Trials so far have focused on small devices such as power tools. “We still have challenges to overcome before being commercially ready for a vehicle application but it’s what the whole team is working towards,” he says.

Another area of research is in “predictive solutions”, an advanced form of battery testing.

Stanford University and Massachusetts Institute of Technology researchers, working with Toyota Research Institute, have found AI can be used to measure the useful lifetime of lithium-ion battery-powered devices.

Findings published in Nature Energy, a scientific journal, show this can spur improved battery design and reduce production costs. “It’s a very new form of battery research. People are getting excited about it and what it means for future innovation,” says study co-author Patrick Herring, of the Toyota Research Institute in Los Altos, California.

The researchers say their model can cut down testing times for batteries by a factor of ten.

Other potential applications include “sorting” used batteries between those with long and short lifespans, which could transform recycling as weaker cells are given a “second life” as street lights or in data centres.

However, the commercialisation of these technologies is still in its infancy and much research remains closely guarded by the companies that house such incubators.

Mr Herring says mass adoption of advanced battery systems using AI is still 10 years away but it is conceivable that, within five years, car owners could see algorithmically managed batteries fitted to their vehicles allowing them to improve charging times markedly.

With the arrival of AI-optimised batteries, fast-charging could become something anyone can do in their basement.

Get alerts on Batteries when a new story is published

Copyright The Financial Times Limited 2022. All rights reserved.
Reuse this content (opens in new window) CommentsJump to comments section

Follow the topics in this article