Deep Learning-Based State-of-Charge Estimation for Lithium-Ion Batteries Across the Entire Life Cycle

Accurately estimating the state-of-chargc (SOC) of lithium-ion batteries throughout their entire life cycle is challenging due to the continuous degradation of their statc-of-hcalth (SOH) with increasing charge-discharge cycles. To address this issue, a deep learning-based SOC estimation model for lithium-ion batteries is proposed. The model utilizes sequential data consisting of current, voltage and temperature measurements from the estimated time and preceding time steps as input. It leverages one-dimensional convolutional neural networks (ID CNN) to extract features from the sequence and uses gated recurrent units (GRU) to establish the nonlinear relationship between the features and SOC. Bayesian optimization (BO) is applied to optimize the network hypcrparameters, enhancing prediction accuracy. The proposed model is validated using two publicly available datasets. Experimental results demonstrate that it achieves accurate SOC predictions within a wide range of SOH and outperforms single deep learning models in terms of prediction accuracy. Compared with CNN and BiLSTM models, the proposed model reduces the root mean square error by an average of 15. 16% and 45. 22%, respectively. When the input sequence length is set to 10 and the data sampling interval is 1 minute, the root mean square error of the predictions is below 2% for both datasets. © 2024 Xi'an Jiaotong University. All rights reserved.