Transient thermal management of nano-enhanced phase change material jackets for lithium-ion batteries: a two-dimensional finite-volume numerical study with one-dimensional PINN verification

Abstract Passive thermal management of lithium-ion cells using phase change material (PCM) jackets is limited by the low thermal conductivity of paraffin carriers. This work develops and verifies a numerical framework for transient conduction and melting in nano-enhanced PCM (NePCM) jackets. A two-dimensional finite-volume solver models a prismatic cell with two paraffin RT42 jackets on its large faces using a fixed-grid apparent-heat-capacity formulation; non-uniform tab-weighted heat generation renders the field genuinely two-dimensional. Effective properties follow volume-fraction mixture rules, and graphene nanoplatelets (GNP), CuO and Al 2 O 3 are compared at 2, 5 and 10 wt%. A physics-informed neural network (PINN) solves the one-dimensional through-thickness sub-problem and is verified against the finite-volume solver, which itself reproduces the analytical two-phase Stefan solution to within a 0.15 mm front error and a 0.17 K root-mean-square error (RMSE); the PINN matches the one-dimensional field with RMSE below 0.32 K and maximum deviation 0.40 K. All parametric battery results are finite-volume solutions; the energy balance closes to within 0.1%. At 30 C ambient, a 5 wt% GNP jacket lowers the peak cell temperature by 0.6 C at 2 C and 0.9 C at 3 C. The extrapolative 10 wt% case gives a 1.2 C reduction at 2 C and 30 C, with a maximum reduction of 1.8 C at 3 C and 35 C ambient; GNP gives the largest peak-temperature reduction within the selected mean-field property model, a ranking that is model-dependent rather than experimentally established. For the primary 5 wt% GNP case, the peak still exceeds the 55 °C reference threshold at 2 C and 35 C ambient and under all investigated 3 C conditions, so nanoparticle enhancement reduces but does not eliminate the high-rate exceedance of that reference threshold; these reductions (0.3–0.9 C at 5 wt%) are small relative to the model uncertainties and should be read as a modest, conditional improvement rather than a decisive safety solution. Absolute temperatures remain conditional on the assumed heat-generation model, whose reversible-heat term dominates the sensitivity. The PINN is used only for the one-dimensional through-thickness verification case; the study is a numerical verification exercise in which the solvers are cross-checked against an analytical solution and against each other, all reported temperatures are model predictions.