Quantifying Lithium-ion Battery Pack Thermal Behavior Based on a Metric-Driven Approach and Machine Learning

A novel approach is presented for quantifying the thermal behavior of lithium-ion battery (LIB) packs through a metric-driven classification technique. Using experimental data including state of charge (SOC), C-rate, and temperature (T), sampled from a LIB pack, the suggested approach aims to classify various test cases based on their electrothermal characteristics. This classification is essential for real-world scenario modeling, as high accuracy simulations of LIBs, that cover a wide range of operation, are vital. In particular, to address the diverse behaviors encountered in practical applications, a comprehensive set of six groups of metrics is proposed. These metrics are designed to evaluate various aspects of electrothermal performance, including the fraction of each test duration relative to the total dataset, the initial conditions and the coverage ranges of T and SOC. In addition, the remaining metrics focus on analyzing the behavior of the C-rate, including its distribution and the frequency of significant charging or discharging portions. In order to prove that these metrics facilitate the categorization of different test cases, that is essential for both model calibration and validation, the metrics have been applied on a set of case-study of 13 different experimental test cases. It is concluded that this methodology offers valuable insights for enhancing LIB modeling accuracy and optimizing battery system design and management strategies.