Cycling Lifetime Prediction Model for Lithium-ion Batteries Based on Artificial Neural Networks

Lithium-ion battery is introduced recently as a key solution for energy storage problems both in stationary and mobility applications. However, one main limitation of this technology is the aging, i.e., the degradation of storage capacity. This degradation happens in every condition, whether the battery is used or not, but in different proportions dependent on the usage and external conditions. Due to the complexity of aging phenomena to characterize, lifetime modeling of Li-ion cells attracts the attention of researchers in recent years. This paper develops cycling lifetime prediction models, for two different commercially available Li-ion cells, by using artificial neural networks. First, accelerated cycling tests are performed under different testing conditions, including different temperatures, state of charges, depth of discharges, and discharge current rates. Then, the test data is used to train a feedforward neural network that can predict one-step ahead state of health of the cells that are cycled under different conditions. Thereafter, a sensitivity analysis method is used to investigate the dependence of the state of health of the cells to each input parameter by calculating the partial derivative of the neural network model output with regard to each input. Finally, the sensitivity profile over the whole range of the inputs is provided and discussed.