A semi-empirical, electrochemistry-based model for Li-ion battery performance prediction over lifetime

Predicting the performance of Li-ion batteries over lifetime is necessary for design and optimal operation of integrated energy systems, as electric vehicles and energy grids. For prediction purposes, several models have been suggested in the literature, with different levels of complexity and predictability. In particular, electrochemical models suffer of high computational costs, while empirical models are deprived of physical meaning. In the present work, a semi-empirical model is suggested, holding the computational efficiency of empirical approaches (low number of fitting parameters, low-order algebraic equations), while providing insights on the processes occurring in the battery during operation. The proposed model is successfully validated on experimental battery cycles: specifically, in conditions of capacity fade >20%, and dynamic cycling at different temperatures. A comparable performance to up-to-date empirical models is achieved both in terms of computational time, and correlation coefficient R2. In addition, analyzing the evolution of fitting parameters as a function of cycle number allows to identify the limiting processes in the overall battery degradation for all the protocols considered. The model suggested is thus suitable for implementation in system modelling, and it can be employed as an informative tool for improved design and operational strategies.