Covering Disruptive Technology Powering Business in The Digital Age

Home > Archives > News > AI Is Now Being Used to Speed Up the Development of Longer Lasting Batteries
AI Is Now Being Used to Speed Up the Development of Longer Lasting Batteries
October 21, 2020 News


Team of researchers from Stanford MIT and the Toyota Research Institute are working on leveraging AI to help provide insights to improve the lifecycle of a battery, in its raw state. This enables them to predict the performance of lithium-ion batteries through the data generated from numerous experiments, which would take years without AI.

Here’s the gist of how they’re able to achieve that: The researchers will subject 124 commercial lithium-ion phosphate/graphite cells under fast-charging conditions, with varying life cycles ranging from 150 to 2300 cycles. Data is then generated, showing information such as discharge voltage curves and capacity degradation, which can then be analysed and tweaks can be made to improve the numbers.

Such a feat is different from ordinary experiments, as others would collect data only after the battery had started to degrade. In this situation, the researchers’ AI collects data while the battery is still in its best condition.

By applying machine-learning tools to the batteries, the researchers are able to predict and classify cells by cycle life. Meaning, they would be able to see trends that can be used to improve battery life before it starts to degrade. Since a battery’s life is complex and non-linear, with no two batteries being the same, it would be hard to conduct experiments by subjecting it through many possibilities. This is where AI can help, as AI feeds on large amounts of data and excels at pinpointing patterns, anomalies and making predictions.

With that, the paper also talked about how the researchers developed data-driven models that accurately predict the life cycle of the batteries using early-cycle data, with no prior knowledge of degradation mechanisms.

“Machine-learning approaches are especially attractive for high-rate operating conditions, where first-principle models of degradation are often unavailable”, stated in the paper.

This means opportunities for improving upon state-of-the-art prediction models include higher accuracy, earlier prediction, great interpretability and broader application to a wide range of cycling conditions.

According to one of the authors, this development in the studies of batteries is a team effort, and the data brought forth by different organisations is vital. This research also aims to reduce the carbon usage of companies by using rechargeable electric batteries instead of non-renewable resources, and AI can greatly help in this mission.