Prediction of state of charge for Li-Co batteries with fuzzy inference system based fuzzy neural networks

This research proposed a method to predict the SOC of Li-Co batteries. This proposed technology can be used in the battery management system of mobile phones, power tools, electric vehicles, or hybrid electric vehicles. For life cycle testing, 60 Li-Co batteries were used to study the characteristics of the SOC. The voltage at the current sampling time and the previous two sampled voltages, the sampling time, and the present discharging current are used as the SOC patterns. The sampling time mentioned above will be affected by the current SOC. The sampling time during the normal SOC is constant but the sampling time near the very high SOC and the very low SOC is shorter due to the faster voltage variation. The fuzzy inference system (FIS) based fuzzy neural network (FNN) with the ability of training and learning was used in this study to predict the SOC of the battery. The experimental results show that the prediction of SOC using FNN is performed better with the training data taken from 36 Li-Co battery testing. The average error is -0.4%, the standard deviation is 5.3%, and the maximum error is 17.7%, and the computation time to predict the SOC is less than 148 ms. The experimental results depict that the SOC of Li-Co battery can be predicted quite accurate and than can be used for the online prediction.