en.Wedoany.com Reported - Chalmers University of Technology in Sweden recently disclosed an AI fast-charging research achievement. The battery health-aware charging method developed by the researchers can extend the lifespan of lithium-ion batteries by 22.9% compared to current standard methods, without increasing charging time.
Fast charging is a key experience in the adoption of electric vehicles, but high currents accelerate internal side reactions within the battery, shortening its usable life. Chalmers University of Technology points out that current charging strategies typically use fixed current and voltage limits, with little change in the control method regardless of whether the battery is new or has been in use for many years. The problem is that batteries age over time, and their internal electrochemical state changes; if the charging system cannot recognize these changes, high-speed charging is more likely to trigger side reactions such as metallic lithium deposition, causing capacity loss, increased internal resistance, and, in severe cases, affecting safety. The researchers introduced AI into fast-charging control, with the core goal of dynamically adjusting the charging current and cut-off voltage based on the battery's current state of health while maintaining fast charging speed.
The study used equivalent full cycles to measure battery life, defined as the number of complete charge-discharge cycles achievable before the battery capacity drops to 80% of its original value. The new method improved battery life by 22.9%, with an average charging time of 24.12 minutes, compared to 24.15 minutes for the standard method—a difference of mere seconds.
The technical pathway is based on reinforcement learning. Chalmers University of Technology explains that the algorithm treats the battery as an interactive operating environment, learning how to control the charging process through long-term outcome feedback, aiming to maintain short charging times while reducing harmful degradation mechanisms. Related papers show that the researchers established a mapping between side reaction overpotential and battery state of health, and used this relationship to constrain the terminal charging voltage; the model operates within a deep reinforcement learning framework, coupled with a high-fidelity single-particle electrochemical model and the PyBaMM simulation platform, to conduct lifecycle comparisons against conventional constant-current constant-voltage methods and their variants. Chalmers University of Technology also emphasizes that the trained AI model does not require specialized laboratory sensors during the operational phase and could theoretically be imported into existing vehicles or energy storage systems via a battery management system software update.
This research is still in the method validation stage. The research team noted that the relationship between charging voltage and battery state of health is influenced by temperature and cell chemistry, thus requiring characterization for different battery types. The next steps will explore transfer learning to reduce the experimental workload needed for adapting to different chemical systems, and to test the trained AI controller directly on physical batteries.
This AI fast-charging method from Chalmers University of Technology offers a software-based pathway to extend the life of EV and energy storage system batteries. If subsequently validated successfully in real battery and full vehicle environments, the nearly 23% lifespan improvement could reduce warranty risks, enhance residual value, and improve the utilization efficiency of critical raw materials.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
