Modeling and Estimation of Lithium-ion Battery State of Charge Using Intelligent Techniques

Abstract

The Li-ion battery is an energy storage system in consumer and industrial applications. Because of their cell and pack level protection, the Li-ion battery requires a battery management system. The important function of the battery management system is to monitor the Li-ion battery state of charge (SOC), to indicate the charge level of the battery. In this research article, efficient intelligent techniques-based SOC estimation algorithm is presented. The proposed techniques are feedforward neural network and layer recurrent neural network with a Scaled Conjugate Gradient (SCG) and Levenberg-Marquardt (LM) training methods. The proposed estimators are applied on 18650 single-cell Li-ion battery to test the performance of the neural networks to estimate the SOC. A real-time experiment carried out on 18650 single-cell Li-ion battery, and experimental results and characteristics are analyzed. The battery cell voltage and current obtained from experimental results are used as the input parameter to proposed networks and battery SOC as the output. The proposed estimation is carried out in the MATLAB. The simulation results show that layer recurrent neural network with LM training method has the best performance to estimate the Li-ion battery SOC in terms of accurate measurement with actual SOC and highest convergence speed.