Volume 26: Closing Carbon Cycles – A Transformation Process Involving Technology, Economy, and Society: Part I

A Random Health Indicator and Deep Learning Approach Based Capacity Estimation for Lithium-Ion Batteries with Different Fast Charging Protocols Qiao Wang, Min Ye, Meng Wei, Gaoqi Lian, Yan Li

Abstract

Different fast charging protocols will cause different battery aging rates, which will reduce the accuracy and robustness of the capacity estimation. To improve the robustness and accuracy of battery capacity estimation with different fast charging rates, a random health indicator and deep learning approach based capacity estimation framework is proposed in this paper. First, a robust health indicator is proposed to extract the relationship between battery charging data and aging rate, which is a random charging curve segment consist of voltage, current, and charging capacity. Second, a deep convolutional neural network is proposed to estimate capacity based on the robust health indicator with smaller model size, and the field model can be quickly obtained by the pre-trained model and transfer learning. Finally, the proposed framework is verified by the public datasets and experimental datasets with different fast charging protocols. The results show that even the charging protocols of test data are different from that of training data, the average error of capacity estimation is within 0.35%.

Keywords Lithium-ion batteries, capacity estimation, health indicator, deep learning, fast charging