AI Fast-Charging Strategy Extends EV Battery Life 23%

Chalmers researchers developed an AI-driven fast-charging strategy using reinforcement learning in the BMS to adapt current to each battery’s chemistry and health, extending cycle life by 23% without longer charge times.

Share This Post

Researchers at Chalmers University of Technology in Sweden have developed an AI-driven charging strategy that can extend electric vehicle battery life by nearly 23% without increasing charging time. By incorporating reinforcement learning into a vehicle’s existing battery management software, the method dynamically adjusts the fast-charging current based on each battery’s chemistry and state of health.

Fast charging is essential for taxis, heavy vehicles and long-distance passenger travel, but it accelerates degradation mechanisms such as lithium plating, which can reduce capacity and introduce safety risks. Electric vehicle batteries today typically last 8 to 15 years, depending on usage and charging habits, and consumer concerns about longevity remain a key factor in purchase decisions. Analysis indicates fast charging accounts for about 10–12% of all charging events, especially among drivers without home chargers or those covering long distances.

In the study, led by Changfu Zou of Chalmers and Meng Yuan of Victoria University of Wellington, the team simulated one of the most common lithium-ion battery chemistries. They trained an AI model to reward charging actions that balance speed with minimal electrochemical wear. Compared to standard protocols, this adaptive approach achieved a 23% increase in equivalent full cycles—how many complete charge/discharge cycles a battery endures before dropping to 80% capacity—while keeping total charging times unchanged to within seconds.

“We show that tailoring current profiles to the battery’s changing electrochemical state lets us charge nearly as fast as today’s methods but with significantly less degradation,” said Yuan. Zou noted that the strategy can be deployed via an over-the-air software update to existing battery management systems, making it both cost-effective and easy to implement.

The researchers acknowledge that each battery chemistry will require model calibration. They propose using transfer learning to adapt their AI framework quickly to different cell designs. The next step involves validating the approach on physical battery packs. If successful, this software-only upgrade could lower warranty costs, enhance resale values, and support broader electric vehicle adoption by mitigating fast-charging concerns.

Source: Chalmers University of Technology

Subscribe to Newsletter

Share This Post

Logo_Battery-Tech-Network_Thumbnail

Subscribe To Our
Weekly Newsletter​

Logo_Battery-Tech-Network_Thumbnail

Let's connect

And Find Out How We Can Work Together