Early prediction of battery remaining useful life using CNN-XGBoost model and Coati optimization algorithm


Safavi V., Mohammadi Vaniar A., Bazmohammadi N., Vasquez J. C., KEYSAN O., Guerrero J. M.

Journal of Energy Storage, vol.98, 2024 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 98
  • Publication Date: 2024
  • Doi Number: 10.1016/j.est.2024.113176
  • Journal Name: Journal of Energy Storage
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
  • Keywords: CNN, Coati Optimization, Early remaining useful life prediction, Lithium-ion batteries, Machine learning, XGBoost
  • Middle East Technical University Affiliated: Yes

Abstract

Lithium-ion (Li-ion) batteries are essential for modern power systems but suffer from performance degradation over time. Accurate prediction of the remaining useful life (RUL) of these batteries is critical for ensuring the reliability and efficient operation of the power grid. On this basis, this paper presents a novel Coati-integrated Convolutional Neural Network (CNN)-XGBoost approach for the early RUL prediction of Li-ion batteries. This method incorporates CNN architecture to automatically extract features from the discharge capacity data of the battery via image processing techniques. The extracted features from the CNN model are concatenated with another set of features extracted from the first 100 cycles of measured battery data based on the charging policy information of the battery. This combined set of features is then fed into an XGBoost model to make the early RUL prediction. Additionally, the Coati Optimization Method (COM) is utilized for CNN hyperparameter tuning, to improve the performance of the proposed RUL prediction method. Numerical results reveal the effectiveness of the proposed approach in predicting the RUL of Li-ion batteries, where values of 106 cycles and 7.5% have been obtained for the RMSE and MAPE, respectively.