Enhancing state of charge and state of energy estimation in Lithium-ion batteries based on a TimesNet model with Gaussian data augmentation and error correction

Accurately estimating the state of charge (SOC) and state of energy (SOE) of lithium-ion batteries is crucial for their safe and stable operation. This study proposes a hybrid deep learning model based on Gaussian data augmentation (GDA), the TimesNet model, error correction (EC), and an improved Bayesian algorithm called Sequential Model-based Algorithm Configuration (SMAC) for SOC and SOE estimation in lithium-ion batteries. Firstly, we compared the performance of the TimesNet model with other benchmark models. Then, GDA data with different signal-to-noise ratios were used for testing, and the model's performance was improved using GDA data with appropriate signal-to-noise ratios. Finally, an error correction method was employed to further enhance the estimation accuracy. During the experiment, SMAC was used to optimize its hyperparameters. In NN and UDDS drive cycles at temperatures of 0 °C, 10 °C, and 25 °C, the highest RMSE values for SOC and SOE estimation of the proposed model were 0.105%, 0.098%, 0.227%, and 0.213%, respectively. Experimental results demonstrate that the TimesNet model achieves good prediction performance for SOC and SOE estimation. GDA and EC effectively enhance the accuracy of the model.