State of Health Estimation of Lithium Ion Battery with Uncertainty Quantification Based on Bayesian Deep Learning

Lithium Ion (Li-ion) batteries have been widely used in the field of electric vehicles (EVs). The safety of Li-ion battery is what people concern most. Accurately predicting the state of health (SOH) of Li-ion battery is a crucial problem. Previous studies have obtained high precision in SOH estimation. However, the prediction results are always point estimates which cannot obtain the confidence interval. SOH estimation without uncertainty quantification for Li-ion battery maintenance decision is risky. The work described in this paper is an attempt to quantify the aleatoric uncertainty and epistemic uncertainty of SOH estimation for Li-ion battery. We propose a new method for SOH estimation based on Bayesian neural network (BNN) using variational inference (VI) and Monte Carlo dropout (MC dropout) approximate inference methods. The Li-ion battery dataset published by National Aeronautics and Space Administration (NASA) is applied to validate the feasibility of the proposed method. Under the condition that the precision of SOH estimation is almost constant or even better comparing with non-Bayesian probabilistic models, we also obtain the uncertainty of the estimations, which makes the results more robust.