An Analytical Model for Lithium-Ion Batteries Based on Genetic Programming Approach

In this paper, a novel approach based on a Genetic Programming (GP) algorithm is proposed to develop behavioral models for Lithium batteries. In particular, this approach is herein adopted to analytically correlate the battery terminal voltage to its State of Charge (SoC) and Charge rate (C-rate) for discharging current profiles. The GP discovers the best possible analytical models, from which the optimal one is selected by weighing several criteria and enforcing a trade-off between the accuracy and the simplicity of the obtained mathematical function. The proposed models can be considered an extension of the behavioral models that are already in use, such as those based on equivalent electrical circuits. This GP approach can overcome some current limitations, such as the high time required to perform experimental tests to estimate the parameters of an equivalent electrical model (particularly effective since it must be repeated with the battery aging) and the need for some apriory knowledge for the model estimation. In this paper, a Lithium Titanate Oxide battery has been considered as a case study, analyzing its behavior for SoC comprised between 5% and 95% and C-rate between 0.25C and 4.0C. This paper represents a preliminary study on GP-based modeling, in which the best behavioral model is identified and tested, with performances that encourage further investigation of this kind of evolutionary approaches by testing them with experimental characterization data.