Scholarlink

In recent years, with the swift advancement of lithium battery technology and annual increases in energy and power density, ensuring their safe, stable operation and accurately forecasting their Remaining Useful Life (RUL) have become crucial. This paper introduces a method that combines Fixed-Budget Kernel Recursive Least Squares (FB-KRLS) and Unscented Kalman Filtering (UKF) for estimating the RUL of lithium batteries, termed the UKF-FB-KRLS method. The combination of UKF and FB-KRLS enhances the nonlinear modeling capability of the model and can better characterize the internal features of the prediction model. In addition, the fitting ability of the model is greatly improved. To assess the predictive capability of the model, lithium battery data sourced from NASA was used for validation and comparison with other prediction models. The results show that the model attains a Mean Absolute Percentage Error (MAPE) of 0.26%, a Root Mean Square Error (RMSE) of 0.004, and a Mean Absolute Error (MAE) of 0.0037. These metrics underscore the model’s effectiveness in improving the precision of Remaining Useful Life (RUL) estimates for lithium batteries.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of FB-KRLS Method Based on UKF to the Prediction of RUL of Lithium Batteries

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pengfei Ding,
  • Jun Li

摘要

In recent years, with the swift advancement of lithium battery technology and annual increases in energy and power density, ensuring their safe, stable operation and accurately forecasting their Remaining Useful Life (RUL) have become crucial. This paper introduces a method that combines Fixed-Budget Kernel Recursive Least Squares (FB-KRLS) and Unscented Kalman Filtering (UKF) for estimating the RUL of lithium batteries, termed the UKF-FB-KRLS method. The combination of UKF and FB-KRLS enhances the nonlinear modeling capability of the model and can better characterize the internal features of the prediction model. In addition, the fitting ability of the model is greatly improved. To assess the predictive capability of the model, lithium battery data sourced from NASA was used for validation and comparison with other prediction models. The results show that the model attains a Mean Absolute Percentage Error (MAPE) of 0.26%, a Root Mean Square Error (RMSE) of 0.004, and a Mean Absolute Error (MAE) of 0.0037. These metrics underscore the model’s effectiveness in improving the precision of Remaining Useful Life (RUL) estimates for lithium batteries.