A data-driven method based on recurrent neural network method for online capacity estimation of lithium-ion batteries

lithium-ion batteries are a convenient choice for various energy storage systems (ESS) such as electric and hybrid vehicles. Nevertheless, the characterization of capacity degradation is critical to ensure the proper performance of lithium-ion batteries. This paper presents a data-driven technique based on a recurrent neural network called nonlinear autoregressive exogenous neural network (NARX) to estimate the capacity degradation of lithium-ion batteries. The voltage charging curves, extracted from twelve nickel manganese cobalt oxide (NMC) cells with different aging trends are used to develop a predictive model for capacity estimation. The results demonstrate that the proposed model is able to estimate capacity with high accuracy and low complexity.