Abstract
Abstract
The operating conditions, either C-rate or DoD, of all the Li-ion cells in the battery pack of EVs are subject to continuous change. In addition, the cells’ ambient temperature (Tamb
) is also not constant due to geographical or day/night conditions. As the generation and transfer of heat from the cells are vital functions of C-rate, DoD, and Tamb
, choosing an appropriate heat loss coefficient for the given conditions is imperative to maintain the operating temperature of the cell below a specified Set Point Temperature (SPT). The selected heat loss coefficient must be the minimum possible such that overcooling of the cells can also be eliminated. The present study employed a machine learning based surrogate model called Gaussian Process Regression (GPR) to achieve this objective for an AMP20M1HD - A0 Li-ion pouch cell. The training and validation of the surrogate model are conducted with the samples generated using Latin Hypercube Sampling and simulated using the NTGK model available in the Ansys Fluent. The model’s accuracy is further tested for three new combinations of the operating conditions, which are not used for training or validation. Using the present model, the predicted minimum heat loss coefficient successfully regulates the cell’s maximum temperature below a user-specified SPT for the same user-given operating conditions. The developed model immensely helps in designing a cost-effective battery thermal management system with optimum cooling capacity by predicting the nature of heat loss coefficients for all plausible combinations of the operating conditions.