Deep learning for state of health estimation in EV batteries: A CNN-BiLSTM approach

Estimation of the State of Health (SOH) of lithium-ion batteries is essential to ensure the safe and efficient operation of electric vehicles. In this paper, a hybrid architecture combining Convolutional Neural Networks (CNNs) with Bidirectional Long Short-Term Memory (BiLSTM) is proposed in order to exploit the benefits of spatial and sequential data processing in enhancing SOH predictions. The model was tested with the NASA battery dataset (B0005, B0006, B0007, and B0018), and it performed significantly better than the other models like LSTM, BiLSTM, and CNN-LSTM. The model was proven to enhance key performance metrics like Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) by up to 26.54% and 43.33%, respectively, over previously used techniques. Furthermore, the model demonstrated superior accuracy compared to recent transformer-based SOH estimation models, achieving an average RMSE of 0.4951% and MAE of 0.3867%, with improvements of 0.98% and 29.69%, respectively, over the best-performing transformer model. The findings indicate the possibility of the proposed model as a reliable solution for estimating battery health, opening opportunities for advancements in energy storage systems.