Lithium-ion battery (LIB) incidents remain challenging to mitigate because thermal runaway (TR) cells retain substantial internal heat, which can drive re-ignition and thermal runaway propagation (TRP) if suppression cooling is delayed or under-designed. This study combines pure-water spray cooling experiments on a 51 Ah prismatic NCM cell undergoing TR with a validated three-dimensional heat-transfer model to quantify how spray duration ( \(\:\tau\:\) ), total sprayed mass ( \(\:{M}_{spray}\) ), nozzle height ( \(\:{H}_{spray}\) ), and cell properties govern cooling outcomes. Cooling performance is evaluated using the rebound-peak temperature after spray termination ( \(\:{T}_{reb}\) ) and the cumulative exposure time above 150 °C ( \(\:{t}_{>150\:^\circ\:C}\) ), where 150 °C is adopted as a conservative threshold for propagation risk. A validated three-dimensional transient heat-transfer model was established to describe post-TR spray cooling, internal heat redistribution, and temperature rebound. Under \(\:{M}_{spray}\) = 5 kg and \(\:{H}_{spray}\) = 0.25 m, τ ≈ 310 s minimizes \(\:{t}_{>150\:^\circ\:C}\) while keeping \(\:{T}_{reb}\) low. The existence of an optimal \(\:\tau\:\) reflects a trade-off: short \(\:\tau\:\) is limited by internal heat conduction and induces strong post-spray rebound, whereas long \(\:\tau\:\) reduces the impingement mass flux and slows the early-stage cooling rate. Increasing \(\:{M}_{spray}\) reduces \(\:{t}_{>150\:^\circ\:C}\) with diminishing returns, and increasing \(\:{H}_{spray}\) monotonically degrades performance. Across prismatic cells, the recommended \(\:\tau\:\) shifts with cell geometry and thermal load, consistent with the role of thermal diffusion in post-spray heat redistribution. Finally, a Gaussian process regression surrogate model was developed to predict \(\:{T}_{\text{r}\text{e}\text{b}}\) , \(\:{t}_{>150,\text{D}\text{S}}\) , \(\:{t}_{>150,\text{A}\text{S}}\) , and \(\:{t}_{>150,\text{t}\text{o}\text{t}\text{a}\text{l}}\) from spray conditions and cell-related parameters. Grouped five-fold cross-validation yielded \(\:{R}^{2}=0.989-0.997\) , supporting rapid screening of spray settings within the investigated parameter ranges.
