Improved Hybrid Neural Network Based on Cnn-Bilstm-Attention for Co-Estimation of SOC and SOE in Lithium-Ion Batteries

Abstract

As the core of modern energy storage technology, lithium-ion batteries are widely used in fields such as electric vehicles, renewable energy storage, and portable electronic devices. Accurately estimating the state-of-charge (SOC) and state-of-energy (SOE) of lithium batteries is crucial for the safety and efficiency of battery management systems. This article proposes an improved convolutional neural network - bidirectional long short-term memory neural network-attention (CNN-BiLSTM-Attention) hybrid neural network model to estimate the SOC and SOE of lithium-ion batteries. In order to effectively capture long-term dependencies in time series, the BiLSTM is proposed based on long short-term memory neural networks. Meanwhile, by introducing an encoder-decoder attention mechanism, complex data can be processed more effectively, thereby improving the accuracy and reliability of predictions. The results indicate that CNN-BiLSTM-Attention has the smallest mean absolute error (MAE) and root mean square error (RMSE). Under the time condition of 35 °C, the model estimates the MAE and RMSE of SOC and SOE to be around 1%, with SOC estimating an MAE of 0.97%. In addition, the model exhibits robustness in data processing and effectively handles the bias of random data.

Keywords: lithium-ion battery, State of charge, State of energy, Joint estimation, Hybrid neural network