Scientists have created a model that can forecast the lifespan of high-energy-density lithium-metal batteries by utilizing machine learning techniques on battery performance data. This model successfully predicts the longevity of batteries by examining their charging, discharging, and voltage relaxation data, without needing to make any assumptions about specific degradation processes. This approach is anticipated to enhance the safety and reliability of devices that run on lithium-metal batteries.
NIMS and SoftBank Corp. have collaboratively developed a model that can forecast the lifespan of high-energy-density lithium-metal batteries by using machine learning techniques on battery performance data. This model has demonstrated the ability to accurately estimate the longevity of batteries by analyzing their charge, discharge, and voltage relaxation data without assuming any specific degradation mechanisms. This method is expected to significantly improve the safety and dependability of devices that utilize lithium-metal batteries.
Lithium-metal batteries have the potential to provide higher energy densities per unit mass compared to the lithium-ion batteries currently available. Therefore, there is great anticipation for their application across various technologies, such as drones, electric vehicles, and home energy storage systems. In 2018, NIMS and SoftBank launched the NIMS-SoftBank Advanced Technologies Development Center. Since then, they have collaborated on research regarding high-energy-density rechargeable batteries for diverse applications, including mobile phone base stations, IoT devices, and high-altitude platform stations (HAPS).
Previously, a lithium-metal battery was reported to achieve an energy density exceeding 300 Wh/kg and maintain a lifespan of over 200 charge/discharge cycles. However, safely implementing high-performance lithium-metal batteries like this will necessitate the creation of techniques that can reliably predict their lifespans. The degradation processes in lithium-metal batteries are more intricate than those in traditional lithium-ion batteries, and their full understanding is still lacking, making the development of lifespan prediction models a significant challenge.
The research team produced numerous high-energy-density lithium-metal battery cells, each featuring a lithium-metal anode and a nickel-rich cathode, through advanced fabrication methods they had developed. They then assessed the charge and discharge performance of these cells. Ultimately, they built a model capable of predicting the cycle lives of lithium-metal batteries by applying machine learning techniques to the charge and discharge data. The model succeeded in making accurate predictions by analyzing the data related to charging, discharging, and voltage relaxation without making assumptions about specific degradation mechanisms.
The team aims to enhance the accuracy of the cycle life predictions of the model further and expedite the transition of high-energy-density lithium-metal batteries into practical applications by utilizing the model in the development of new lithium-metal anode materials.
