Robust sliding mode observer using RBF neural network for lithium-ion battery state of charge estimation in electric vehicles

Abstract

A robust sliding mode observer (SMO) based on a radial basis function (RBF) neural network (NN) is presented for battery state of charge (SOC) estimation. Comparing with an ordinary SMO for the SOC estimation, the robust SMO employs the RBF NN to learn the upper bound of system uncertainties caused by the discrepancy between a battery equivalent circuit model (BECM) and a battery. The output of the RBF NN is then used as an adaptive switching gain in the sense that the effects of the system uncertainties can be compensated so that asymptotic SOC estimation error convergence can be attained by the robust SMO. The experiments are conducted on a lithium-ion (Li-ion) battery for extracting parameters of the BECM and verifying the effectiveness of the proposed scheme for the SOC estimation.