Managing heat generation in lithium-ion batteries remains a major obstacle, as excessive temperatures and non-uniform thermal distribution can significantly undermine safety, operational efficiency, and service life. To mitigate these challenges, this research introduces a structured hybrid methodology for advancing battery thermal management systems (BTMS). The approach integrates machine learning-driven forecasting, optimization techniques, and multi-criteria evaluation within a four-phase framework: preprocessing of data, predictive modeling through hybrid learning algorithms, optimization of design variables, and systematic decision analysis. Two hybrid predictive schemes were constructed, GA-MLPNN and MPA-MLPNN, to estimate critical thermal and performance indices. Comparative assessment revealed that MPA-MLPNN exhibited the highest precision in forecasting temperature differentials (ΔT, R > 0.9985), whereas GA-MLPNN delivered superior predictive reliability for peak temperature (Tmax, R > 0.9986) and energy density (ED, R > 0.999). Multi-objective optimization using the MOMVO algorithm illuminated inherent trade-offs among heat distribution, maximum temperature reduction, and energy storage capacity. For instance, adopting thicker MHPA layers (~ 3.8-3.9 mm) and taller configurations (~ 18-25 mm) effectively reduced hotspots, producing ΔT values of ~ 1.8-2.5 °C and Tmax of ~ 38-39 °C, though at the expense of lower ED (~ 139-144 Wh/kg). Conversely, thinner designs (~ 2.6-2.8 mm) elevated energy density (~ 157 Wh/kg) but impaired thermal stability. Intermediate geometries offered more balanced outcomes, with spacing analyses confirming geometry as a decisive factor in thermal regulation. Finally, multi-criteria evaluation through the CoCoSo method translated Pareto-optimal fronts into practical design recommendations, demonstrating adaptability across diverse operational priorities, including safety-focused, high-capacity, and compromise-oriented battery applications.
Ali et al. (Mon,) studied this question.