Adaptive voltage estimation for EV Li-ion cell based on artificial neural networks state-of-charge meter

This paper reports some results relating to adaptive cell modeling from neural network state-of-charge (SOC) estimation in a full-electric-vehicle (EV) application. The cells in question are commercialized ones, Lithium-ion Polymer based, with a nominal capacity of about 100 Ah and dedicated to energy applications. Using a recurrent neural network, we developed a SOC predictor that takes into account operational conditions. More importantly, the predictor allows very precise SOC estimation, therefore allowing the vehicle controller to confidently use the battery pack´s full operating range without problem of over- or under-charging cells. In this work, the estimated SOC values helped to estimate the parameters of an adaptive-dynamic battery model using RLS algorithm with time-dependent forgetting factor. Simulation results confirmed the accuracy of the terminal voltage estimation of the battery.