State-of-health estimation for lithium-ion batteries based on hierarchical feature construction and GWO-SCNs

Lithium-ion batteries, with their high energy density, extended lifespan, and low self-discharge, are integral to the propulsion of new energy vehicles. Accurate prediction of these batteries’ State of Health (SOH) is essential. Data-driven methodologies are extensively employed for estimating the state of health (SOH). However, challenges persist due to high feature dimensions and dependence on model parameter selection. To overcome these limitations, this study introduces an innovative feature extraction method based on charging voltage curves, streamlined by a hierarchical inflection point strategy. Further, we propose a new SOH estimation method that merges the Grey Wolf Optimizer (GWO) with Stochastic Configuration Networks (SCNs). This approach enhances predictive accuracy and generalizability by automatically tuning model parameters. The efficacy of the GWO-SCNs model was evaluated using a lithium-ion battery dataset provided by NASA. The results indicate that the model achieves a maximum Mean Absolute Percentage Error (MAPE) of 1.35% and a Root Mean Square Error (RMSE) of 1.60%. Compared with traditional LSTM and SCNs models, the GWO-SCNs model demonstrates higher accuracy and reliability in battery SOH estimation, highlighting its significant performance advantages.