AI Can Extend EV Battery Life by 23% With No Extra Charging Time, Study Finds

Scientists at Chalmers University of Technology in Gothenburg, Sweden, have developed an AI-driven charging strategy that squeezes almost a quarter more usable life out of lithium-ion batteries while keeping charge times identical to today's standard protocols. The findings, published in IEEE Transactions on Transportation Electrification, arrive at a moment when battery longevity remains one of the most persistent anxieties holding back EV adoption across Europe.

Fast Charging's Hidden Cost

DC fast charging is the backbone of long-distance electric driving. Without it, cross-border trips, taxi fleets, and apartment dwellers without home charging would be far harder to electrify. But the convenience comes at a price.

When high currents are forced into battery cells over minutes rather than hours, unwanted chemical side reactions accelerate. The most damaging of these is lithium plating — a process where metallic lithium deposits unevenly on the electrode surface instead of intercalating properly into the battery's structure. "This not only reduces capacity, but in the worst case, an uneven lithium layer can cause a short circuit," the Chalmers team warns.

The problem is compounded by how today's battery management systems operate. They apply the same charging parameters regardless of whether a battery has 5,000 or 150,000 kilometres on the clock — even though the risk of lithium plating increases substantially as cells age.

Data from Geotab, which analysed over 22,700 EVs across 21 makes and models, confirms the pattern: vehicles relying heavily on DC fast charging above 100 kW degrade at up to 3.0 percent per year, roughly double the rate of those charged primarily via AC or lower-power sources. Geotab's latest study, published in January 2026, pegged the average annual degradation rate across all EVs at 2.3 percent.

An Algorithm That Learns How to Charge

Professor Changfu Zou from Chalmers' Department of Electrical Engineering and his colleague Meng Yuan from Victoria University of Wellington took a fundamentally different approach to the charging problem. Instead of designing a fixed charging profile, they built a system based on reinforcement learning — a machine learning technique where an algorithm improves through trial and reward.

The AI was trained inside a digital twin of a typical EV battery, with simulations that modelled both charging speed and the electrochemical stress factors that degrade cell health over time. Crucially, the model learned to adjust its behaviour in real time based on two variables simultaneously: the battery's current state of charge and its overall state of health.

"The AI model learned to adapt charging according to how full or empty the battery is, while also taking its overall health status into account," the authors explain. "The result is a charging strategy that maintains short charging times while minimising harmful reactions."

The outcome: battery lifespan extended by 22.8 percent compared to standard charging protocols, measured in equivalent full cycles (EFC) — the number of complete charge-discharge cycles a battery can endure before its capacity drops to 80 percent of the original value. The total charging time? Virtually unchanged, with a difference measured in seconds.

No New Hardware. Just a Software Update.

Perhaps the most striking aspect of the discovery is how little stands between the lab and real-world deployment. "In principle, this is a change that can be implemented through a simple software update to the battery management system," says Zou. No new cells. No redesigned cooling. No upgraded charging infrastructure.

This means the technology could theoretically reach EVs already on European roads — provided manufacturers choose to push the update. For the automotive industry, the implications are substantial. "A nearly 23 percent extension of battery life can mean lower warranty costs, better residual values for vehicles, and more efficient use of critical raw materials," Zou adds.

The one caveat: the method must be calibrated for different battery chemistries. The research team plans to address this using transfer learning, a technique that allows a trained model's knowledge to be adapted to new battery types far faster than training from scratch.

What It Means for Drivers

Today's EV batteries typically last between 8 and 15 years, depending on usage patterns and climate. According to the European Commission's Consumer Monitor 2025 survey, concern about limited battery lifespan remains one of the top reasons European drivers hesitate to switch to electric.

Those fears are amplified by fast charging. Although Geotab's data shows DC fast charging accounts for only about 10 to 12 percent of all charging sessions, its disproportionate effect on degradation makes it a genuine concern — especially for fleet operators and high-mileage drivers.

If AI-managed charging reaches production vehicles, a battery that would normally dip below 80 percent capacity after 160,000 kilometres under a typical warranty (such as Volvo's 8-year coverage) could retain significantly more usable range well beyond that threshold. The driver would notice no difference at the charger.

The Road Ahead

The method has so far been validated entirely in simulation. "The next step is to test it directly on physical batteries," the researchers state. Should those real-world tests succeed, the path to commercial deployment opens.

The research was supported by the European Union's Horizon Europe programme through Marie Skłodowska-Curie Actions Postdoctoral Fellowships, the Swedish Research Council, and the Swedish Foundation for International Cooperation in Research and Higher Education.

For an industry racing to convince sceptical buyers that EV batteries will outlast their combustion-engine counterparts, a 23 percent longevity boost — delivered over-the-air — would make a compelling argument.