Journal of Korea Society of Industrial Information Systems (한국산업정보학회논문지)

Korea Society of Industrial Information Systems (한국산업정보학회)
권호 표지
DOI QR코드

DOI QR Code

Large Language Model-based SHAP Analysis for Interpretation of Remaining Useful Life Prediction of Lithium-ion Battery

거대언어모델 기반 SHAP 분석을 이용한 리튬 이온 배터리 잔존 수명 예측 기법 해석

  • 이재승 (고려대학교 전기전자공학과) ;
  • 유제혁 (덕성여자대학교 데이터사이언스학과)
  • Received : 2024.07.15
  • Accepted : 2024.09.12
  • Published : 2024.10.30
Download PDF
⟨ Previous Next ⟩

Abstract

To safely operate lithium-ion batteries that power mobile electronic devices, it is crucial to accurately predict the remaining useful life (RUL) of the battery. Recently, with the advancement of machine learning technologies, artificial intelligence (AI)-based RUL prediction models for batteries have been actively researched. However, existing models have limitations as the reasoning process within the models is not transparent, making it difficult to fully trust and utilize the predicted values derived from machine learning. To address this issue, various explainable AI techniques have been proposed, but these techniques typically visualize results in the form of graphs, requiring users to manually analyze the graphs. In this paper, we propose an explainable RUL prediction method for lithium-ion batteries that interprets the reasoning process of the prediction model in textual form using SHAP analysis based on large language models (LLMs). Experimental results using publicly available lithium-ion battery datasets demonstrated that the LLM-based SHAP analysis enabled us to concretely understand the model's prediction rationale in textual form.

이동성을 갖춘 전자 장비에 에너지원을 공급하는 리튬 이온 배터리를 안전하게 운영하기 위해서는 배터리의 잔존 수명을 정확히 예측하는 것이 중요하다. 최근, 기계학습 기술의 발달로, 인공지능 기반의 배터리 잔존 수명 예측 모델이 활발히 연구되고 있다. 하지만, 기존 모델들은 모델 내부에서 일어나는 추론 과정을 알 수 없어 기계학습을 통해 예측된 값을 완전히 신뢰하고 사용하는 데 제약이 있었다. 이를 해결하기 위해 여러 설명가능한 인공지능 기법이 제안되었지만, 이러한 기법들은 단순히 결과를 그래프 형태로 시각화하였기에 사용자가 직접 그래프를 분석해야 했다. 이에 본 논문에서는 거대언어모델에 기반한 SHAP 분석을 이용하여 예측 모델의 추론 과정을 텍스트 형태로 해석하는 설명가능한 리튬 이온 배터리 잔존 수명 예측 기법을 제안한다. 공개 리튬 이온 배터리 데이터셋을 이용한 실험 결과, 거대언어모델 기반 SHAP 분석을 통해 모델의 예측 근거를 텍스트 형태로 구체화하여 이해할 수 있었다.

Keywords

References

  1. Andrioaia, D., Gaitan, V., Cuela, G. and Banu, I. (2024). Predicting the RUL of Li-Ion Batteries in UAVs Using Machine Learning Techniques, Computers, 13(64). https://doi.org/10.3390/computers13030064
  2. Bentejac, C., Csorgo, A. and Martinez-Munoz, G. (2020). A Comparative Analysis of Gradient Boosting Algorithms, Artificial Intelligence Review, 54, 1937-1967. https://doi.org/10.1007/s10462-020-09896-5
  3. Chen, Y., Kang, Y., Zhao, Y., Wang, Li, Liu, J., Li, Y., Liang, Z., He, X., Li, X., Tavajohi, N. and Li, B. (2021). A Review of Lithium-ion Battery Safety Concerns: The Issues, Strategies, and Testing Standards, J ournal of Energy Chemistry, 59, 83-99. https://doi.org/10.1016/j.jechem.2020.10.017
  4. Devie, A., Baure, G. and Dubarry, M. (2018). Intrinsic Variability in the Degradation of a Batch of Commercial 18650 Lithium-ion Cells, Energies, 11(5), 1031. https://doi.org/10.3390/en11051031
  5. Dhabe, P. R., Paraskar, S. R. and Jadhao, S. S. (2023). Battery State of Charge Estimation Methods - A Critical Review, International J ournal of Innovative Research in Electrial, Electronics, Instrumentation and Control Engineering, 11(12). https://doi.org/10.17148/IJIREEICE.2023.111201
  6. Fang, X., Xu, W., Tan, F., Zhang, J., Hu, Z., Qi, Y., Nickleach, S., Scolinsky, D., Sengamedu, S. and Faloutsos, C. (2024). Large Language Models(LLMs) on Tabular Data: Prediction, Generation, and Understanding - A Survey, arXiv preprint, arXiv:2402.17644
  7. Gao, K., Xu, J., Li, Z., Cai, Z., Jiang, D. and Zeng, A. (2022). A Novel Remaining Useful Life Prediction Method for Capacity Diving Lithium-ion Batteries, ACS Omega, 7(30), 26701-26714. https://doi.org/10.1021/acsomega.2c03043
  8. Gohel, P., Singh, P. and Mohanty, M. (2021). Explainable AI: Current Status and Future Directions, arXiv preprint, arXiv:2107.07045
  9. Gopalakrishnan, R., Vidhya, D. S., Bharath, P., Kaviyan, P. and Sivaselvam, P. (2023). Battery State Estimation and Control System for Mobile Charging Station for Electric Vehicles, International Conference on Sustainable Computing and Data Communication Systems, Mar. 23-25, Erode, India, pp. 985-989. https://doi.org/10.1109/ICSCDS56580.2023.10104624
  10. Jiang, J., Shi, W., Zheng, J., Zuo, P., Xiao, J. A., Chen, X., Xu, W. and Zhang, J. (2013). Optimized Operating Range for Large-Format LiFePO4/Graphite Batteries, J ournal of the Electrochemical Society, 161(3), 336-341. https://doi.org/10.1149/2.052403jes
  11. Jiao, R. H., Ma, X., Li, L. and Xiao, J. P. (2022). Battery Remaining Useful Life Prediction Based on a Combination of ARMA and Degradation Model, 13th International Conference on Reliability, Maintainability, and Safety, Aug. 21-24, Hong Kong, China, pp. 115-119.
  12. Kim, H. and Yu, Y. (2023). Development of a Regulatory Q&A System for KAERI Utilizing Document Search Algorithms and Large Language Model, J ournal of Korea Society of Industrial Information Systems, 28(5), 31-39. http://doi.org/10.9723/jksiis.2023.28.5.031
  13. Kim, J., Park, J., Jin, B., Park, S., Jung, S. and Kim, J. (2022). Optimal Battery Aging Model based on Moore-Penrose Pseudo Inverse Matrix for Early Prediction of Remaining-useful-life on Lithium-ion Battery, Power Electronics Annual Conference, Jul. 5-7, Gyeongju, Korea, pp. 168-170.
  14. Kim, S., Park, S. and Kim, J. (2020). State of Health estimation based on Secondary Li-ion battery Electrochemical Modeling and Electrical Experiment, J ournal of IKEEE, 24(4), 1098-1103. https://doi.org/10.7471/ikeee.2020.24.4.1098
  15. Lee, D. and Yoon, D. (2012). Evaluating Modeling Heat Generation Behavior for Lithium-ion Battery using FEMLAB, Clean Technology, 18(3), 320-324. https://doi.org/10.7464/ksct.2012.18.3.320
  16. Lee, J. and Rew, J. (2024). Large Language Models-based Feature Extraction for Short-Term Load Forecasting, J ournal of Korea Society of Industrial Information Systems, 29(3), 51-65. http://doi.org/10.9723/jksiis.2024.29.3.051
  17. Lee, M., Park., J. and Kim, J. (2021). Prediction of Maximum Available Current of High Power Lithium-ion Battery based on Electrochemical Impedance Spectroscopy, Power Electronics Annual Conference, Jul. 6-8, Buan, Korea, pp. 348-350.
  18. Liu, R. (2022). Remaining Useful Life Prediction of Lithium-ion Batteries Using Multiple Kernel Extreme Learning Machine, Recent Advances in Computer Science and Communications, 15, 715-721. https://doi.org/10.2174/2666255813999201002152742
  19. Louppe, G. (2014). Understanding Random Forests: From Theory to Practice, Dissertation, Graduate School of Liege University, Liege, Belgium.
  20. Luca, G. D., Blasio, G. D., Gimelli, A. and Misul, D. A. (2024). Review on Battery State Estimation and Management Solutions for Next-Generation Connected Vehicles, Energies, 17(1), 202. https://doi.org/10.3390/en17010202
  21. Lundberg, S. and Lee, S. (2017). A Unified Approach to Interpreting Model Predictions, Proceedings of the Neural Information Processing Systems, Dec. 4-9, California, CA, USA, pp. 4768-4777.
  22. Maulud, D. H. and Abdulazeez, A. M. (2020). A Review on Linear Regression Comprehensive in Machine Learning, J ournal of Applied Science and Technology Trends, 1(4), 140-147. https://doi.org/10.38094/jastt1457
  23. Minaee, S., Mikolov, T., Nikzad, N., Chenaghlu, M., Socher, R., Amatrain, X. and Gao, J. (2024). Large Language Models: A Survey, arXiv preprint, arXiv:2402.06196
  24. Nitta, N., Wu, F., Lee, J. and Yushin, G. (2015). Li-ion Battery Materials: Present and Future, Materials Today, 18(5), 252-264. https://doi.org/10.1016/j.mattod.2014.10.040
  25. OpenAI. (2023). GPT-4 Technical Report, arXiv preprint, arXiv:2303.08774
  26. Park, J. (2022). A Study on Battery Remaining Useful Life Prediction based on Incremental Machine Learning with Battery Usage Patterns, Dissertation, Graduate School of Seoul National University of Science and Technology, Seoul, Korea.
  27. Park, K. and Yi, K. (2013). Voltage Balancing Circuit for Li-ion Battery System, J ournal of Korea Society of Industrial Information Systems, 18(5), 73-80. http://doi.org/10.9723/jksiis.2013.18.5.073
  28. Saeed, W. and Omlin, C. (2023). Explainable AI (XAI): A Systematic Meta-survey of Current Challenges and Future Opportunities, Knowledge-based Systems, 263, 110273. https://doi.org/10.1016/j.knosys.2023.110273
  29. Shi, Y., Yang, Y., Wen, J., Cui, F. and Wang, J. (2020). Remaining Useful Life Prediction for Lithium-ion Battery based on CEEMDAN and SVR, IEEE 18th International Conference on Industrial Informatics, Jul. 20-23, Warwick, United Kingdom, pp. 888-893.
  30. Son, Y. (2023). A Study on Accelerating Battery Aging Estimation through Machine Learning for Reduced Estimation Time, Thesis, Graduate School of Kookmin University, Seoul, Korea.
  31. Stroe, D. and Schaltz, E. (2018). SOH Estimation of LMO/NMC-based Electric Vehicle Lithium-ion Batteries Using the Incremental Capacity Analysis Technique. IEEE Energy Conversion Congress and Exposition, Sep. 23-27, Portland, OR, USA, pp. 2720-2725. https://doi.org/10.1109/ECCE.2018557998
  32. Wu, J., Kong, L., Cheng, Z., Yang, Y., and Zuo, H. (2022). RUL Prediction for Lithium Batteries Using a Novel Ensemble Learning Method, International Conference on the Energy Internet and Energy Interactive Technology, Mar. 25-27, Wuhan, China, pp. 313-326. https://doi.org/10.1016/j.egyr.2022.10.298
  33. Zhao, J., Tian, L., Cheng, L., Zhang, Y. and Zhu, C. (2022). Review on RUL Prediction Methods for Lithium-ion Battery, IEEE/IAS Industrial and Commercial Power System Asia, Jul. 8-11, Shanghai, China, pp. 1501-1505. https://doi.org/10.1109/ICPSAsia55496.2022.9949753
  34. Zheng, Z., Wei, J., Hu, X., Zhu, H. and Nevatia, R. (2024). Large Language Models are Good Prompt Learners for Low-Shot Image Classification, IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun, 17-21, Seattle, WA, USA, pp. 28453-28462.
  35. Zhou, L., Lai, X., Li, B., Yao, Y., Yuan, M., Weng, J. and Zheng, Y. (2023). State Estimation Models of Lithium-Ion Batteries for Battery Management System: Status, Challenges, and Future Trends, Batteries, 9(2), 131. https://doi.org/10.3390/batteries9020131
  36. Zhou, Y. and Huang, M. (2016). Lithium-ion Batteries Remaining Useful Life Prediction based on a Mixture of Empirical Mode Decomposition and ARIMA Model, Microelectronics Reliability, 65, 265-273. https://doi.org/10.1016/j.microrel.2016.07.151