A Comparative study on state of charge estimation techniques for Lithium-ion Batteries

State of Charge (SOC) estimation is significantly important for the optimal utilization and protection of batteries. This paper implements and compares the performance of a neural network (NN) algorithm and Coulomb Counting method for estimating state of charge (SOC) for batteries. This algorithm is applied to a battery management system (BMS) in electric vehicles. Accurate SOC information can avoid over charging and over discharging of battery, and thus increase battery life. Also, control system uses accurate SOC information to make rational decisions to save energy in electric vehicles. The advantage of NN model over Coulomb Counting method is it can be implemented in BMS Hardware where online measurements like current, voltage and temperature are available. The feature of this neural network approach is that it optimizes two important hyper-parameters to achieve a reasonable MAPE error. The performance of the proposed method is tested using two Datasets for city driving conditions. The results reveal that both methods (NN and Coulomb counting) can predict SOC with reasonable error (<6%). However, Coulomb counting outperforms Neural network MAPE for both Datasets.