Abstract: Estimating the remaining capacity of lithium-ion batteries is a pivotal technology in battery management systems, serving as a fundamental prerequisite for ensuring the safe and stable operation of these batteries. To effectively estimate the remaining capacity of lithium-ion batteries, this article proposes a method for online estimating the remaining capacity of a lithium-ion battery using Radial Basis Function-Autoregressive Exogenous (RBF-ARX) model with exogenous input. The model parameters are identified using the structured parameter optimization method. The "aging information" and "energy" are combined, and the energy features are directly extracted from the battery charging current/voltage curves as new health factors. Then, transfer entropy is used to select the new health factors to construct the model input, thus achieving effective estimation of the remaining capacity of the lithium-ion battery. Finally, based on the publicly available lithium-ion battery aging data from NASA, comprehensive analysis is conducted using different training/testing sample proportions and different models. The results indicate that the proposed state-dependent RBF-ARX model for estimating the remaining capacity of lithium-ion batteries performs well compared to commonly used data-driven methods. The error metrics, including the mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE), remain at a low level respectively, demonstrating good estimation accuracy.