Sign in Subscribe
Batteries

Enhanced State of Health Estimation for Lithium-Ion Batteries Using CNN-BIGRU Model

Strategic Insights

— 2 min read

Research on improving the accuracy of lithium-ion battery state of health (SOH) estimation has been crucial for ensuring the safe operation of electric vehicles. A new study published in the World Electric Vehicle Journal proposes an enhanced CNN-BIGRU model that addresses the limitations of current data-driven approaches. The model integrates a Hiking Optimization Algorithm (HOA) to improve early training stability and an Attention mechanism to dynamically weight features, significantly enhancing prediction accuracy. Experimental validation using the NASA dataset demonstrates the model's superior convergence speed and prediction accuracy compared to the CNN-BIGRU-Attention benchmark.

Key Takeaways:

  • The enhanced CNN-BIGRU model proposed in this study integrates a Hiking Optimization Algorithm (HOA) to improve early training stability and an Attention mechanism to dynamically weight features.
  • Experimental validation using the NASA dataset demonstrates the model's superior convergence speed and prediction accuracy, with RMSEs controlled within 0.01 and R2 kept above 0.91.
  • The HOA-CNN-BIRGU-Attention model has a higher prediction accuracy and better robustness under different conditions, with RMSEs on the University of Maryland dataset all below 0.006 and R2 kept above 0.98.
  • The model reduces RMSE by at least 0.15% across different battery groups in the NASA dataset compared to the CNN-BIGRU-ATTENTION baseline model without HOA optimization.
  • The research was funded by the National Key Research And Development Program of China, the Scientific And Technological Research Project of The Hubei Provincial Department of Education, and the Knowledge Innovation Program of Wuhan-shuguang Project.
  • The study was conducted by researchers from Hubei University of Technology, including Qianli Dong, Ziyang Liu, Hainan Wang, Lujun Wang, Rui Dong, and Lu Lv.

Statistics:

  • RMSEs controlled within 0.01 and R2 kept above 0.91 on the NASA dataset.
  • RMSEs on the University of Maryland dataset all below 0.006 and R2 kept above 0.98.
  • RMSE reduced by at least 0.15% across different battery groups in the NASA dataset compared to the CNN-BIGRU-ATTENTION baseline model.

Sources:

  • "SOH Estimation of Lithium Battery Under Improved CNN-BIGRU-Attention Model Based on Hiking Optimization Algorithm." World Electric Vehicle Journal 16(9):487. (World Electric Vehicle Journal - http://www.mdpi.com/journal/wevj).
  • MDPI AG (publisher for World Electric Vehicle Journal)
  • "New Electric Vehicles Research Has Been Reported by Researchers at Hubei University of Technology (SOH Estimation of Lithium Battery Under Improved CNN-BIGRU-Attention Model Based on Hiking Optimization Algorithm)." Journal of Transportation. October 18, 2025; p 115.

Read more