State-of-charge estimation of the lithium-ion battery using neural network based on an improved thevenin circuit model

This paper focuses on the real-time estimation of the State of Charge (SOC) in Lithium-Ion battery. When it comes to highly complex electrochemical reaction inside the battery, the conventional first order battery model is not accurate and cannot respond to the battery’s conditions correctly because of the simplicity of the model, especially in the beginning and ending of charge/discharge stages. So, the neural network (NN) is selected to estimate the SOC dynamically due to its strong nonlinear fitting ability. The NN strategy also was used to implement the parameter identification for the battery model. In order to obtain accurate and robust SOC estimations, the correct order of the equivalent circuit model was chosen. SOC mean square error (MSE) method can determine the proper order of the battery model for the online SOC estimation. This paper compared the NN strategy with adaptive extended Kalman filter for the online SOC estimation. It shows that the proposed NN strategy can estimate the SOC of battery with high effectiveness and accuracy.