New AI-Based Model Boosts Accuracy in Estimating Lithium-Ion Battery Aging

Lithium-ion batteries is a hot topic in spectroscopic circles. Because they are a key component in many electronic devices regularly used today, the performance of lithium-ion batteries is important into building a more sustainable future (1). Accurately predicting the point at which a battery will no longer function optimally is crucial for both economic and environmental reasons because it allows for better planning around battery replacement and recycling (1).

A recent study from Qingdao University explored this topic by developing and testing a new method to improve accuracy of estimating the aging life of lithium-ion batteries. This new approach, which leverages advanced machine learning techniques like variational autoencoders (VAE) and bidirectional gated recurrent units (BiGRU), could have profound implications for the use, maintenance, and economic optimization of energy storage systems (2).

Fully charged lithium ion battery cell. Generated with AI. | Image Credit: © Ai Inspire - stock.adobe.com

In this study, led by Kai Wang of Qingdao University and published in Energy Storage Materials, introduced a model that automates the extraction of key features from electrochemical impedance spectroscopy (EIS) data was introduced (2). Traditionally, EIS data, which provides a detailed look at the internal resistance and overall health of batteries, has been difficult to analyze because of its high dimensionality (2). The complexity of this data often requires time-consuming manual processing, limiting its potential for real-time applications (2).

Conventional methods for assessing LIB health, such as EIS, have proven to be highly accurate. However, the sheer amount of data generated by these methods creates a bottleneck in analysis, particularly when multiple variables—such as temperature and cycling conditions—must be considered simultaneously (2). To overcome this issue, Wang and his team turned to artificial intelligence (AI) (2).

AI is increasingly being used in scientific research, particularly when interpreting and analyzing complex data sets (3). Wang’s study used AI here for a similar purpose. The team employed VAE and BiGRU to tackle the data complexity challenge they were encountering (2). The VAE component of the model is responsible for reducing the dimensionality of EIS data, simplifying it while retaining critical information (2). This reduction accelerates the analysis process while minimizing the computational resources needed.

Meanwhile, BiGRU was used to handle time-series data, making it ideal for mapping the reduced impedance data to the battery’s state of health and capacity over time (2). The researchers achieved good results combining VAE and BiGRU, including a mean absolute error (MAE) of less than 1.27 milliampere-hours (mAh) and a root mean square error (RMSE) under 1.43 mAh (2).

To ensure the robustness of their method, the researchers tested it under multiple scenarios, including changing the temperatures and cell degradation states to evaluate how their method performed. The research team found that their model demonstrated consistent accuracy regardless of environmental factors (2). This conclusion was an important discovery because, in practical applications, batteries are subject to fluctuating temperatures and operational conditions.

In addition to its technical accuracy, the new method offers economic benefits. By automating the feature extraction process, it reduces the labor and time costs associated with manual data analysis, which could make battery monitoring systems more accessible and affordable for industries that rely on energy storage (2).

The findings from Qingdao University represent a significant step forward in the predictive maintenance of lithium-ion batteries. As energy storage systems become increasingly integrated into global energy infrastructure, methods that improve battery longevity and performance will be vital.

References

1. Wetzel, W. Lithium Metal Batteries and the Push for More Sustainable Electronics. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/lithium-metal-batteries-and-the-push-for-more-sustainable-electronics (accessed 2024-10-17).
2. Liu, Y.; Li, Q.; Wang, K. Revealing the degradation patterns of lithium-ion batteries from impedance spectroscopy using variational auto-encoders. Energy Stor. Mater. 2024, 69, 103394. DOI: 10.1016/j.ensm.2024.103394
3. Wetzel, W. The Latest in Data Analytics, Statistics, Machine Learning, and Artificial Intelligence. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/the-latest-in-data-analytics-statistics-machine-learning-and-artificial-intelligence (accessed 2024-10-18).