State of charge estimation method for sodium-ion batteries based on a hybrid deep learning network
TL;DR
A hybrid deep learning model combining a Temporal Convolutional Network, Bidirectional Long Short-Term Memory, and Squeeze-and-Excitation module (TCN-BiLSTM-SE) is proposed for SOC estimation, supporting the practical application of SIBs in energy storage and electric vehicle systems.
Abstract
With the rapid development of transportation electrification, power electronics, and large-scale renewable energy integration, the demand for high-performance and low-cost battery technologies continues to grow. Sodium-ion batteries (SIBs) have emerged as a promising alternative to lithium-ion batteries due to their abundant resources and cost advantages. However, accurate state-of-charge (SOC) estimation for SIBs remains challenging under varying temperatures, operating conditions, and cell types, limiting battery management system reliability. To address this issue, a hybrid deep learning model combining a Temporal Convolutional Network, Bidirectional Long Short-Term Memory, and Squeeze-and-Excitation module (TCN-BiLSTM-SE) is proposed for SOC estimation. Experiments were conducted on four cylindrical SIB cell types under three temperatures (5℃, 25℃, and 45℃) and ten operating profiles. The results demonstrate that the proposed model achieves superior SOC estimation accuracy and generalization performance compared with conventional deep learning methods across different temperatures, cell types, and operating conditions, supporting the practical application of SIBs in energy storage and electric vehicle systems.