Automatic Feature Extraction Enabled Lithium-Ion Battery Capacity Estimation Using Random Fragmented Charging Data

Ziyou Zhou,Yonggang Liu,Zhigang Zhao,Huan Xia,Zheng Chen,Yuanjian Zhang

DOI: https://doi.org/10.1109/tte.2024.3357728

IF: 6.519

2024-01-01

IEEE Transactions on Transportation Electrification

Abstract:Nowadays, health diagnosis for lithium-ion batteries is critical to ensure their normal and safe operations. However, precise estimation of battery capacity is a challenging task, especially under complex and varying operation conditions. To tackle this problem, we propose an automatic feature extraction technique that utilizes random fragmented charging data to achieve precise capacity estimation across diverse operational scenarios. The automatic feature extraction is achieved by a deep autoencoder model and can apply to other conditions without additional training, justifying its generalization performance. Through a comprehensive exploration of the capacity estimation performance across various input data segments, we introduce a novel approach to select preferable input data and develop a universal estimation model for achieving accurate capacity estimation. Additionally, the Bayesian neural network is exploited in the universal estimation model to quantify the uncertainty of the estimated results. Experimental datasets from three distinct types of batteries operating under diverse conditions are applied to examine the performance of the proposed method. The results manifest that our method yields robust and precise capacity estimation under various charging conditions.

engineering, electrical & electronic,transportation science & technology