State of Charge Estimation for Lithium-Ion Battery Using Dynamic Neural Networks

The estimation of the state of charge (SOC) of the power battery is one of the key parameters of the battery management system, the accuracy estimation of state of charge can ensure the energy management efficiency and cruising range of the electric vehicles. However, high accuracy estimation of SOC is a non-linear and unstable problem. In this paper, the dynamic neural network time series method is used to estimate the SOC of the lithium-ion battery, which is improved on the basis of the classic Close-loop Nonlinear Auto-Regressive models with Exogenous Input Neural Network (NARXNN) model, and the Open-loop NARXNN model considering expected output is proposed. The classic unscented Kalman filter (UKF), Back Propagation Neural network (BPNN), Close-loop NARXNN and Open-loop NARXNN are analyzed. The experimental results verify that the Close-loop NARXNN and Open-loop NARXNN have high estimation accuracy and effectiveness in the SOC estimation.