Probabilistic dischargeable time forecasting of power batteries via statistical characterization of future loading profiles

Guangzhong Dong,Yuyao Feng,Yifei Wang,Jingwen Wei

DOI: https://doi.org/10.1016/j.est.2022.106488

IF: 9.4

2022-12-28

Journal of Energy Storage

Abstract:Accurately end-of-discharge prediction of lithium-ion batteries is of great significance for user convenience and optimal path-planning and energy-dispatching in various applications. However, accurate state-of-charge estimation, dynamic and stochastic operating conditions, and uncertainty quantification have remained major challenges. This paper thus proposes a probabilistic scheme for forecasting dischargeable time via a robust observer design and statistical characterization of future loading profiles. First, a robust observer with stability and convergence analysis using an equivalent-circuit model and Lyapunov theory is proposed to guarantee the robustness of state-of-charge estimations. Then, a power histogram is formulated to characterize future loading profiles. Next, a Gaussian process regression and a bootstrap sampling technique are combined to perform dischargeable time forecasting and uncertainty quantification. Finally, the proposed scheme is validated using experimental data acquired from battery cells at different temperatures and loading profiles. Experimental results show that the robust state-of-charge estimation can converge to the true value within an error of 3.50% against over 10% capacity biases. It also demonstrates that the proposed forecasting method can provide dischargeable time prediction within an error of 0.66h, about 20% of the total dischargeable time.

energy & fuels