Researchers at Uppsala University have developed a new AI model that significantly improves predictions of EV battery ageing, offering potential gains in both safety and lifetime performance.
Battery degradation has long been one of the most pressing barriers to the mass adoption of electric vehicles. While automakers have achieved steady improvements in efficiency, charging infrastructure, and vehicle range, lithium-ion batteries often age faster than other components. This results in expensive replacements, reduced resale value, and concerns over reliability.
According to industry estimates, batteries can account for up to 40% of an EV’s cost, making their early failure a major obstacle to electrification. The International Energy Agency (IEA) projects that by 2030, EVs will represent more than 60% of annual passenger car sales and drive over 85% of global lithium-ion battery demand. Extending battery life is therefore essential not only for consumer confidence but also for material efficiency, mining sustainability, and grid integration.
To address these challenges, researchers Wendi Guo and Professor Daniel Brandell at Uppsala University’s Ångström Advanced Battery Centre developed an AI-driven model that analyzes short charging segments to map the chemical reactions inside a cell.
By combining machine learning with a digital twin framework, the system connects upstream battery design parameters—such as electrode thickness, particle radius, lithium-ion concentration, and diffusion coefficients—to real-world ageing behaviors. This approach improved robustness of health and lifetime predictions by up to 69% compared to conventional models.
“Being able to learn more about the life and ageing of batteries will benefit future control systems in electric vehicles,” said Daniel Brandell, lead researcher and director of the Ångström Advanced Battery Centre, in the university’s announcement. “If we stop looking at batteries as black boxes and instead acquire a detailed picture of the processes, we can manage them so that they stay in good condition longer.”
The model offers precise insights into how batteries generate power and age over time, enabling automakers to anticipate degradation earlier and adapt usage strategies accordingly.
The study, published in Energy & Environmental Science in August 2025, was conducted in collaboration with Aalborg University in Denmark. Researchers tested thousands of charging scenarios, including fast charging at different temperatures representative of Nordic climates, to simulate real-world EV use.
Key to the innovation is the reliance on short charging segments—rather than full-cycle data—to predict state of health (SOH) and remaining useful life (RUL). This makes the approach more efficient and practical. The framework can infer sensitive design parameters within seconds, reducing prediction errors for lifetime forecasts by up to 69% and providing early warnings after just 80 equivalent full cycles, or less than 10% of an average battery’s life.
For automakers, this means a reduced need for access to proprietary or user-sensitive vehicle data. As Brandell explained, “Battery data from electric vehicles is sensitive, both for the industry and from an anonymisation point of view for users. This research shows how far you can get without needing complete datasets.”
Beyond extending battery lifespan, the model also enhances safety by identifying side reactions and design flaws that could contribute to hazardous conditions, such as lithium plating and thermal runaway. By linking design parameters directly to ageing pathways, the AI system helps engineers avoid configurations that accelerate degradation under fast charging.
The implications extend beyond individual vehicles. As EV adoption scales, longer-lasting batteries reduce demand for critical minerals like lithium, cobalt, and nickel, easing supply chain pressure and supporting circular economy strategies. More durable batteries also mean fewer replacements and less electronic waste, lowering the overall carbon footprint of the transport sector.
The study further highlights the potential of this model for vehicle-to-grid (V2G) applications, where EVs are used as mobile storage assets to supply energy back to the grid during peak demand. Frequent bidirectional charging places added stress on batteries, but accurate state-of-health predictions could mitigate these risks.
Without robust monitoring, V2G programs risk accelerating battery wear. With the new AI framework, utilities and fleet operators could more confidently integrate EVs into grid-balancing strategies, helping stabilize renewable-heavy energy systems.
This research emerges as automakers and battery manufacturers face mounting pressure to improve durability while scaling up production. Current warranty periods often guarantee EV batteries for 8–10 years or 100,000 miles, but consumers remain concerned about resale value and long-term performance.
According to Stanford University, advances in battery management software could extend usable battery life by 20–30%—a projection that aligns with the Uppsala model’s demonstrated improvements. If integrated into commercial battery management systems (BMS), the AI tool could give EV makers a competitive edge while reducing costs for drivers and fleets.
The Uppsala team notes that future work will focus on expanding the framework to include electrolyte-related parameters, such as ionic conductivity and solvent composition. This could yield even deeper insights into ageing mechanisms and support the design of next-generation cells for both mobility and stationary energy storage.
The model also has potential applications in second-life batteries, where repurposed EV cells are used in stationary storage. Accurate health prediction is essential for ensuring safety and efficiency in these systems.
For now, the results mark a significant step forward in bridging the gap between battery design and real-world performance. By combining AI with physical models, the Uppsala framework demonstrates that smarter diagnostics can accelerate electrification while reducing costs and risks.
