{"title":"Prediction of thermal runaway for a lithium-ion battery through multiphysics-informed DeepONet with virtual data","authors":"Jinho Jeong , Eunji Kwak , Jun-hyeong Kim , Ki-Yong Oh","doi":"10.1016/j.etran.2024.100337","DOIUrl":null,"url":null,"abstract":"<div><p>A surrogate model that predicts thermal runaway (TR) of lithium-ion batteries (LIBs) fast and accurately is essential, yet complex phenomena of TR present significant challenges to achieving adequate performance in both aspects, particularly as traditional finite element models (FEMs) incur significant time and cost. This study proposes a multiphysics-informed deep operator network (MPI-DeepONet) with encoders to address these issues. This proposed neural network aims to predict TR under various thermal abuse conditions, offering a fast and accurate TR prediction surrogate model. In this study, MPI-DeepONet with encoders is trained with virtual data from a multiphysics FEM to overcome the scarcity of actual TR data. The architecture of DeepONet solves interpolation and extrapolation problems, allowing predictions across multiple thermal abuse conditions once trained. The neural network is further enhanced by the supervision of energy balance and chemical reaction equations, ensuring accurate and robust predictions despite limited data by effectively capturing the complex phenomena of TR. Quantitative analysis, compared against actual experiments and ablation studies, confirms the effectiveness of the proposed neural network. Notably, MPI-DeepONet achieves a mean RMSE of 13.2 °C for temperature predictions in the test set, significantly outperforming the 25.4 °C RMSE of purely data-driven DeepONet. This improvement highlights the importance of integrating multiphysics constraints into the neural network. The generality of the proposed neural network is further evidenced by accurate TR prediction in both LFP and NMC cells. The features deployed on the proposed neural network enable real-time quantification of internal temperature distribution and dimensionless concentration of the key components in LIBs, which are challenging to measure directly, achieving speeds at least 10,000 times faster than FEM. The proposed neural network provides comprehensive information for advanced battery management systems to ensure safety and reliability in LIBs, accelerating the digital twin of electric transportation systems through artificial intelligence transformation.</p></div>","PeriodicalId":36355,"journal":{"name":"Etransportation","volume":null,"pages":null},"PeriodicalIF":15.0000,"publicationDate":"2024-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Etransportation","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2590116824000274","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENERGY & FUELS","Score":null,"Total":0}
引用次数: 0
Abstract
A surrogate model that predicts thermal runaway (TR) of lithium-ion batteries (LIBs) fast and accurately is essential, yet complex phenomena of TR present significant challenges to achieving adequate performance in both aspects, particularly as traditional finite element models (FEMs) incur significant time and cost. This study proposes a multiphysics-informed deep operator network (MPI-DeepONet) with encoders to address these issues. This proposed neural network aims to predict TR under various thermal abuse conditions, offering a fast and accurate TR prediction surrogate model. In this study, MPI-DeepONet with encoders is trained with virtual data from a multiphysics FEM to overcome the scarcity of actual TR data. The architecture of DeepONet solves interpolation and extrapolation problems, allowing predictions across multiple thermal abuse conditions once trained. The neural network is further enhanced by the supervision of energy balance and chemical reaction equations, ensuring accurate and robust predictions despite limited data by effectively capturing the complex phenomena of TR. Quantitative analysis, compared against actual experiments and ablation studies, confirms the effectiveness of the proposed neural network. Notably, MPI-DeepONet achieves a mean RMSE of 13.2 °C for temperature predictions in the test set, significantly outperforming the 25.4 °C RMSE of purely data-driven DeepONet. This improvement highlights the importance of integrating multiphysics constraints into the neural network. The generality of the proposed neural network is further evidenced by accurate TR prediction in both LFP and NMC cells. The features deployed on the proposed neural network enable real-time quantification of internal temperature distribution and dimensionless concentration of the key components in LIBs, which are challenging to measure directly, achieving speeds at least 10,000 times faster than FEM. The proposed neural network provides comprehensive information for advanced battery management systems to ensure safety and reliability in LIBs, accelerating the digital twin of electric transportation systems through artificial intelligence transformation.
期刊介绍:
eTransportation is a scholarly journal that aims to advance knowledge in the field of electric transportation. It focuses on all modes of transportation that utilize electricity as their primary source of energy, including electric vehicles, trains, ships, and aircraft. The journal covers all stages of research, development, and testing of new technologies, systems, and devices related to electrical transportation.
The journal welcomes the use of simulation and analysis tools at the system, transport, or device level. Its primary emphasis is on the study of the electrical and electronic aspects of transportation systems. However, it also considers research on mechanical parts or subsystems of vehicles if there is a clear interaction with electrical or electronic equipment.
Please note that this journal excludes other aspects such as sociological, political, regulatory, or environmental factors from its scope.