A recent study looks at how to improve the aging life of lithium-ion batteries.
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.
LIBS Illuminates the Hidden Health Risks of Indoor Welding and Soldering
April 23rd 2025A new dual-spectroscopy approach reveals real-time pollution threats in indoor workspaces. Chinese researchers have pioneered the use of laser-induced breakdown spectroscopy (LIBS) and aerosol mass spectrometry to uncover and monitor harmful heavy metal and dust emissions from soldering and welding in real-time. These complementary tools offer a fast, accurate means to evaluate air quality threats in industrial and indoor environments—where people spend most of their time.
Smarter Sensors, Cleaner Earth Using AI and IoT for Pollution Monitoring
April 22nd 2025A global research team has detailed how smart sensors, artificial intelligence (AI), machine learning, and Internet of Things (IoT) technologies are transforming the detection and management of environmental pollutants. Their comprehensive review highlights how spectroscopy and sensor networks are now key tools in real-time pollution tracking.
New AI Strategy for Mycotoxin Detection in Cereal Grains
April 21st 2025Researchers from Jiangsu University and Zhejiang University of Water Resources and Electric Power have developed a transfer learning approach that significantly enhances the accuracy and adaptability of NIR spectroscopy models for detecting mycotoxins in cereals.