Could AI extend an EV battery’s life without increasing charging time?

Researchers at Chalmers University of Technology in Sweden, in collaboration with Victoria University of Wellington in New Zealand, have developed an innovative artificial intelligence (AI) method to extend the lifespan of electric vehicle (EV) batteries without increasing charging times. Their approach uses AI to optimise rapid charging, a process traditionally known to accelerate battery degradation due to high current loads. By tailoring the charging current to the battery’s chemistry and state of health (SoH), the new method aims to mitigate the wear caused during fast charging, which is essential for passenger vehicles, commercial fleets, and industrial equipment. The study, published in IEEE Transactions on Transportation Electrification, demonstrates that the AI-based charging strategy can increase battery lifetime by nearly 23 per cent, measured in equivalent full cycles (EFCs)—the number of charge-discharge cycles before capacity falls to 80 per cent. Importantly, this improvement in longevity does not come at the expense of charging speed, with average charging times remaining virtually unchanged at just over 24 minutes. This contrasts with conventional charging methods that apply fixed voltage and current limits regardless of battery condition, potentially causing unnecessary damage over time. The AI system employs reinforcement learning, a machine learning technique where the algorithm learns optimal charging behaviours through continuous interaction with the battery environment. This allows the charging process to dynamically adapt based on how charged the battery is and its overall health, rather than relying on static parameters. Changfu Zou, a professor at Chalmers and co-developer of the system, emphasised that the true limitation in fast charging lies in the battery’s evolving electrochemical state, not merely in current restrictions. Integrating AI with a physics-based understanding paves the way for smarter, health-aware charging strategies that balance performance and durability. Looking ahead, the researchers highlight the ease and cost-effectiveness of implementing their method, which would require only a software update to existing battery management systems. However, they acknowledge that further work is needed to calibrate the AI model for different battery types and to validate the approach on physical batteries in real-world conditions. Zou noted that transfer learning could expedite adaptation to various battery chemistries, suggesting a promising path for broader adoption. This advancement could significantly enhance the sustainability and user confidence in EV technology as rapid charging becomes increasingly critical.