Hybrid convolutional neural network–transformer–long short-term memory framework for health estimation of electric mobility battery

The adoption of electric mobility (e-mobility) has seen tremendous growth owing to the aim of achieving sustainability in combination with reduced environmental effects. Lithium-ion batteries drive the majority of e-mobility transportation systems, and their efficient, reliable and safe operation is vital for overall adoption of e-mobility systems. Accurate state of health (SOH) estimation of these batteries is essential for ensuring the high performance of the e-mobility transportation system. To achieve an efficient SOH estimation, this paper proposes a hybrid convolutional neural network transformer long short term memory (CNNT–LSTM) model for efficient SOH estimation. The proposed model is capable of capturing both forward and backward temporal dependencies in sequential data, which is vital for modeling battery degradation over time. This work utilizes a comprehensive dataset for developing and validating the proposed model. Results demonstrate that the proposed model significantly reduces the prediction error, achieving a low mean average error (MAE) and root mean square error (RMSE). The results show the efficacy of proposed model in SOH estimation, offering an effective and promising solution for effective health management of e-mobility batteries.