DOI: 10.30939/ijastech..1966817 ISSN: 2587-0963

One Dimensional Modeling and Multi Objective Optimization of the Parallel Channel Count and Coolant Flow Rate of a Battery Module Cooling System via Machine Learning Surrogates

Çağrı Kürklü, Emre Bulut
Show PDF Cite
Thermal management of lithium ion cells is one of the main challenges in EV battery pack design. In this study, a surrogate model assisted multi objective optimization framework is presented for a liquid-cooled parallel channel battery thermal management system (BTMS). A battery module consists of 30 A123 AMP20 LFP cells is modeled in MATLAB Simscape and validated against published experimental data at 1C, 2C, 3C and 4C discharge rates. Three cooling plate geometries with 3, 4 and 5 parallel channels are investigated by the Latin Hypercube Sampling method over the mass flow rate range of 0.010 to 0.024 kg/s, and a total of 30 simulation points are generated. A contradictory relationship is observed between the channel count and the mass flow rate: the maximum cell temperature is decreased by 1.64 K when the mass flow rate is increased, while the pressure drop is increased 4.6 fold across the same range. Sixteen regression algorithms are benchmarked with leave one out cross validation, and the cubic polynomial, the Bayesian cubic polynomial and the Gaussian process regression are determined as the winning surrogate families. The nine selected surrogates are coupled with the NSGA II algorithm, and five multi criteria decision making methods with a uniform Min Max normalization are applied to the Pareto front points under seven weighting scenarios. It is determined that the 5 channel cooling plate at 0.024 kg/s is the optimum design with Tmax of 304.76 K, ΔT of 0.948 K and ΔP of 251.6 Pa, and a 0.71 K decrease in the peak cell temperature is obtained with a penalty of 70% increase in the pressure drop compared with the efficiency priority configuration.

More from our Archive