A Two-Stage Intelligent Model for State of Health Estimation of EV Lithium-Ion Battery at Variable Temperatures

State of health (SOH) estimation of lithium-ion batteries is essential to ensure the reliability and safety of electric vehicles. Data-driven methods are considered promising solutions for battery SOH estimation due to their ability to handle big data and latent relationships. However, existing data-driven methods are afflicted by the constraints of poor extrapolation due to various operating conditions and may not be applied to zero-shot scenarios such as unseen temperature conditions for SOH estimation. To this end, this paper proposes a two-stage intelligent model based on dual cascade feedforward neural networks (CFNN) for battery SOH estimation at variable operating temperatures. Instead of developing a conventional one-stage model with intricate structure to realize adaptive SOH estimation under various conditions, we first develop a CFNN with only three hidden layers for distinguishing batteries from different operating temperature conditions. Further, a set of CFNNs are developed, and are responsible for the regression task of accurate battery SOH based on the classification result of the first CFNN. The proposed model is validated on over 100 lithium-ion batteries comprising two materials and three operating temperatures. The effectiveness of the proposed model is examined by comparison with conventional one-stage model, which presents a significant reduction up to 58.33% in RMSE in SOH estimation.