A novel technique for modelling the state of charge of lithium ion batteries using artificial neural networks

This paper present a novel design for a lithium ion battery pack state of charge estimator for cellular phones using artificial neural networks (ANNs). The state of charge of a battery is a nonlinear function of the load current, battery temperature, battery chemistry and battery history and hence cannot easily be determined. Different methods have been previously been proposed in the literature for calculating the state of charge for different battery types. However, these methods are not ideally suited for mobile communication applications since the current loads they require are pulsed and hence exhibit a different behaviour on the battery. The new method investigates the effects of pulse currents loads and uses a three-layer feedforward artificial neural network which will be trained using the back propagation algorithm. Experimental and computer results are presented to highlight the advantages of the new technique and to confirm the theoretical developments.