Parameter Identification and Optimization for Lithium-Ion Battery State of Health Detection

Lithium-ion batteries (LiBs) are being extensively employed in consumer goods, electric vehicles, and spacecraft. Nevertheless, due to the ever-increasing demand for high energy density and a harsher working environment, the issue of available LiB capacity, its workable life, and inherent safety need to be addressed. Therefore, predicting the state of the LiB, i.e., its available capacity, workable life, and inherent safety with limited charge and discharge data, can be instrumental for efficient LiB management systems. In this work, PyBaMM’s coupled Doyle-Fuller-Newman (DFN) model incorporating degradation mechanisms such as solid-electrolyte interphase (SEI), lithium plating, particle cracking, and loss of active material is used to simulate complete LiB discharge voltage as a function of time while generating the optimized LiB parameters. A Bayesian optimization algorithm is used to optimize these DFN model parameters against experimental discharge voltage data. The resulting identified model parameters enable the state of the LiB identification. The authors believe that this is the first attempt at optimizing nearly all the parameters related to the degradation mechanisms incorporated in the DFN model using only discharge voltage data to predict the state of the LiB. The two novel outcomes of the work are the identification of the parameters affecting the degradation mechanisms of the coupled DFN model using non-invasive discharge voltage measurements and the use of Bayesian optimization to optimize the coupled DFN model’s parameter values to predict the state of the LiB.