Doyoon Lee of Keimyung University Publishes Lithium-Ion Battery Lifespan Prediction Model in International Journal

Listed as First Author in the Prestigious Energy Journal 'Journal of Energy Storage'
Overcoming Limitations of Conventional Deep Learning by Applying Statistics-Based 'Functional Data Analysis'

The research paper on battery degradation prediction, co-authored by Doyoon Lee, a master's program researcher in the Department of Statistics at Keimyung University (President Shin Ilhee), has been published in the international journal for the energy sector, the Journal of Energy Storage.

This journal is a prestigious publication with an impact factor of 9.8, ranking in the top 14.3% in the Journal Citation Reports (JCR).

Idoyun Lee, Master's Degree Researcher in the Department of Statistics at Keimyung University. Courtesy of Keimyung University

The paper, titled "Functional Modeling of Lithium-Ion Battery Degradation for Enhanced Capacity Prediction," proposes a statistical analysis framework designed to precisely predict the performance degradation of lithium-ion batteries.

The research team applied the Functional Data Analysis (FDA) technique, which converts voltage data generated during battery charging and discharging into continuous functional forms rather than simple numerical values for analysis.

This approach has been recognized for effectively analyzing the shape information of voltage curves and subtle degradation patterns, which conventional time-series deep learning methods have difficulty capturing.

In particular, the team validated the model's performance using public datasets, including 12 NASA batteries and 8 Oxford batteries.

The analysis showed that the proposed model achieved higher prediction accuracy and stability than existing deep learning-based prediction models.

It also enhanced technical completeness by effectively reflecting nonlinear characteristics such as capacity regeneration, where battery capacity temporarily recovers during the degradation process.

The study compared and analyzed the Functional Linear Regression (FLR) model with existing models such as Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU).

As a result, it was demonstrated that excellent interpretability and predictive performance could be achieved without complex structures or separate feature extraction processes.

This research was conducted jointly with the research team led by Professor Kyungmin An of the Department of Data Science at Seoul Women’s University.

Doyoon Lee participated as the first author, leading the overall model design and data analysis, and contributed significantly to the core research outcomes.

Doyoon Lee stated, "By interpreting battery data in the form of functions, we were able to capture degradation characteristics that were difficult to detect using conventional methods," and added, "I hope this research will enhance the reliability of battery management systems (BMS) and improve the safety of energy storage devices."

Meanwhile, this research was supported by the National Research Foundation of Korea's master's student research encouragement program and by Seoul Women’s University.

This content was produced with the assistance of AI translation services.