State of health estimation of lithium-ion battery based on an adaptive tunable hybrid radial basis function network

Accurate state of health (SOH) estimation is critical for the durability and safety of Lithium-ion batteries (LIBs). It is challenging to predict the SOH of LIBs due to the complex aging mechanism. In this paper, a novel adaptive tunable hybrid radial basis function network is proposed for accurate and robust SOH estimation. Firstly, two Kullback-Leibler distances are exploited to extract the dynamic and static characteristics of LIBs during the aging process. Secondly, the hybrid network is combined with the radial basis function network and the autoregressive model. A novel hybrid network state-space model is built to simulate the aging mechanism of LIBs. To increase the adaptive ability of the hybrid network, the structural parameters of the proposed hybrid network are adaptively modulated by Brownian motion modeling and particle filter. The Brownian motion with drift and scale coefficients is introduced in the state-space model to simulate the dynamic aging behavior of LIBs. The particle filter is used to update the structural parameters of the hybrid network in real-time. Furthermore, experiments are conducted on two datasets. The experimental results demonstrate that the proposed model has a high prediction accuracy. Moreover, the batteries with Gaussian white noise and dynamic discharging profiles are adopted to prove the reliability and robustness of the proposed model.

Automated summary

This study proposes a novel adaptive hybrid radial basis function network for accurate and robust estimation of the state of health of lithium-ion batteries. The model extracts dynamic and static characteristics, builds a hybrid network state-space model to simulate aging mechanism, and utilizes Brownian motion modeling and particle filter for adaptive parameter modulation. The experimental results demonstrate high prediction accuracy and reliability of the proposed model under various conditions.