Machine Learning and Genetic Programming-based behavioral modeling approaches of Li-ion Batteries
Authors: IEEE, Di Capua, Giulia, Molinara, Mario, Maffucci, Antonio, Porpora, Francesco, Femia, Nicola, Oliva, Nunzio
Abstract
This paper investigates and compares the performance of various behavioral modeling approaches, both analytical and machine learning-based, for Lithium-ion batteries. The analytical models rely exclusively on the genetic programming algorithm, while the machine learning-based models employ several well-known regression techniques, including multi-layer perceptron, support vector machine, and gradient boosting. These data-driven models are used to relate the battery's terminal voltage to its state of charge, charge/discharge rate, and temperature, using a consistent dataset for the case study. The study focuses on the transient discharge phase of a Lithium Iron Phosphate battery under realistic operating conditions, with a state of charge between 20% and 80%, discharge rates ranging from 0.25C to 1C, and temperatures between 5°C and 35°C.
BibTeX
@article{ieee2025machine,
title={Machine Learning and Genetic Programming-based behavioral modeling approaches of Li-ion Batteries},
author={IEEE and Di Capua, Giulia and Molinara, Mario and Maffucci, Antonio and Porpora, Francesco and Femia, Nicola and Oliva, Nunzio},
year={2025},
doi={10.1109/iscas56072.2025.11043719}
}