Artificial Intelligence, Big Data Optimize EV Batteries

Using AI-backed study of battery history and performance to optimize charging and acutely predict EV battery life

Ottawa’s GBatteries presented demos of its AI-based charging protocol at CES 2019 in Las Vegas. (GBatteries)

Everywhere you turn these days, there’s news about the power of artificial intelligence (AI) to transform our lives—from diagnosing cancers to devising better pick-up lines at bars. Now AI practitioners are aiming their algorithms at electric vehicle (EV) batteries.

Screen image of Dox’s battery-analytics platform platform that can provide analytics of in-use EV battery packs down to the cell level. (Dox)

The Toyota Research Institute (TRI) made a splash earlier this year when it announced that its machine-learning models could accelerate battery research. “You usually don’t predict a battery’s lifetime. You just measure it by testing the thing until it fails,” said Brian Storey, director of TRI’s Accelerated Materials Design and Discovery program. “It’s a bottleneck because it takes a lot of time.”

Storey said that TRI can run 400 different battery experiments simultaneously. The work takes place at Stanford University and other academic labs supported by funding from TRI. “Before, every time you change a process, introduce a new design or chemistry, you’d have to start over by validating the lifetime,” he said. “AI allows you to accelerate R&D cycles.”

The use of AI for battery research also levels the playing field for startups. In 2012, Tim Sherstyuk, as a chemistry student at Carleton University, wanted to know why cell-phone batteries die so quickly. Sherstyuk, now the chief commercial officer at Ottawa’s GBatteries, teamed up with his father, Nick – an electrical engineer and now the company’s CTO – to explore the potential of “pulse” recharging as an alternative to the prevailing constant-current, constant voltage (CCCV) protocol. The goal is to reduce impedance and the damaging impact of rising battery temperatures during fast charging.

Seven years of testing produced promising results and a mountain of data. “Pulse charging is not new, but our technology is dynamic, based on decisions made by the AI in real time,” said the younger Sherstyuk.

GBatteries grew from experiments in the family’s basement into a startup with about 20 employees – and backing from Airbus Ventures, Initialized Capital, Y Combinator and others. The company doesn’t alter the battery: it instead uses an adapter-like device that could be built into the charging connector. “We filter the energy going into the battery pack according to waveforms determined by the algorithms based on real-time measurements,” said Sherstyuk. GBatteries now is working to scale its AI-powered approach from device and EV batteries to large packs in vehicles up to the size of delivery trucks.

Both TRI and GBatteries simulate loads on batteries when in vehicles on the road. Storey acknowledged that the scale complexity of the data from lab testing is “just a drop in the bucket” compared to field data.

Battery research at Carnegie Mellon University funded by Toyota Research Institute can run hundreds of simultaneous battery experiments to help reduce R&D lead times. (CMU)

The challenge of handling the high volume and complexity of EV batteries in the real world is being tackled by Dox, a three-year-old company that specializes in predictive algorithms and big-data visualizations. Its headquarters are in Chicago with research and development in Beirut, Lebanon. “We compare the battery manufacturer’s data sheet with the current operation of the battery,” said Nicolas Jamal, Dox’s chief executive officer. By performing what he calls “anomaly detection,” companies and drivers gain a deeper understanding of battery data –transmitted via telematics to the cloud – based on the vehicle’s specific location and use.

“Our AI stores historical data with the signature of a particular user and how he or she drives and fast-charges,” explained Jamal. The data is sliced and diced down to the cell level and according to location, vehicle model and type, chemistry and numerous other aspects of the battery supply chain. Knowing when it’s time to replace a battery or to keep driving is a potential game-changer. “You can squeeze the hell out of the battery to the maximum,” said Jamal.

The environmental benefits could be profound, considering the resources required to manufacture, distribute and dispose of EV batteries and to extract the raw materials. “If there are many millions of batteries around the globe and we can extend their operation by just one month, there would be a huge reduction in CO2 emissions,” said Jamal.