Multiphysics-Multiscale Reduced Order Model based Design Optimization of the Immersion Cooled Battery System

<div class="section abstract"><div class="htmlview paragraph">A vital aspect of Ultra-Fast Charging (UFC) Li-Ion battery pack is its thermal management system, which impacts safety, performance, and cell longevity. Immersion cooling technology is more effective compared to indirect cold plate as heat can dissipate much quicker and has a potential to mitigate the thermal runaway propagation, improve pack overall performance, and cell life significantly.</div><div class="htmlview paragraph">For design optimization and getting better insight, high fidelity Multiphysics-Multiscale simulations are required. Equivalent Circuit Model (ECM) based electro-thermally coupled multi-physics CFD simulations are performed to optimize the innovative busbar design, of a recently developed immersion cooled battery pack, which enables the capability to remove individual cell. Further, high fidelity 3D transient flow-thermal simulations have helped in optimizing the coolant flow direction, inlet positions, cell spacing and separator design for efficient flow distribution in the module. While high-fidelity CFD models accurately depict flow and thermal behavior, their computational demands often hinder quick optimizations.</div><div class="htmlview paragraph">Therefore, this study focuses on generating Reduced Order Models (ROM) from high fidelity CFD models, to improve prediction performance of Battery Management System (BMS) using real-time simulations. A detailed methodology for creating a linear parameter variant (LPV) ROM, and multiple linear time in-variant (LTI) matrices, for quicker parametric studies, are being studied. The ROM fully integrates electro-thermal aspects for immersion cooling systems where the dielectric liquid is in direct contact with cells and flows along axial direction of the cells. The required training data for the LPV ROM creation is generated by running transient step response thermal simulations on converged steady-state flow solution for different flow rates. The experimental module setup comprising of 144 cells immersed in the dielectric fluid is also prepared for the model validation. The model validations done against test results confirms ROM's accuracy and robustness, with a tenfold reduction in computational time and minimal loss in solution accuracy.</div></div>