A combined state space model with adaptive neural compensator based state of charge determination method for lithium-ion batteries

To guarantee the reliability and performance of electric vehicles, it is significant to monitor the available capacity of the battery system. This paper proposes a combined state space model with adaptive neural compensator based method to predict the state of charge of battery. It adopts a combined state space model to represent cell dynamics and its state equations to access the basic cell state. Furthermore, an adaptive neural compensator is employed to access the self-regulating compensated value which changes with the prediction error of cell terminal voltage. It integrates the effectiveness of combined state space model and the operation robustness, nonlinear mapping and self-learning capabilities of neural compensator. The algorithm is validated by the data set gathered from lithium-ion batteries. Results show that the presented method predicts the state of charge of battery accurately with a rapid convergence and less overshoot while remaining simple to implement. (C) 2020 Elsevier Ltd. All rights reserved.