Nusselt number analysis from a battery pack cooled by different fluids and multiple back-propagation modelling using feed-forward networks

In this article, analysis of average Nusselt number (Nuavg), which indicates the heat removal from the battery pack cooled by flowing fluid is carried out considering coupled heat transfer condition at the pack and coolant interface. Five categories of coolant, mainly gases, common oils, thermal oils, nanofluids, and liquid metals, are selected. In each coolant category, five fluids (having different Prandtl number Pr) are selected and passed over the Li-ion battery pack. The analysis is made for different conductivity ratio (Cr), heat generation term (Qgen), Reynolds number (Re), and Pr. Pr varying in the range 0.0208–511.5 (25 coolants) and Cr for each category of coolant having its own upper and lower limit are used to analyze the heat removed from the battery pack. Using single feed-forward network and integrating two feed-forward networks having multi-layers with back-propagation is employed for artificial neural network (ANN) modelling. In this modelling, the concept of the main network and space network is devised for multiple back propagation (MBP). The numerical analysis revealed that the temperature distribution in battery and fluid is greatly affected by increasing Cr. The maximum temperature located close to the upper edge of battery is found to get reduced significantly with the increase of Cr, but upto a certain limit above which reduction is marginal. The analysis carried out reveals that Cr and Qgen have no role in improving Nuavg while Pr and Re vary it significantly in each step. Moreover, Nuavg is found to increase with Re continuously irrespective of any Cr and Qgen. While, for oils with an increase in Pr and Re, Nuavg was found to reduce significantly. Nanofluids are found to be more effective in improving heat transfer from the battery pack when cooled by flowing nano-coolants over it. The MBP networks proposed are successfully trained, and hence they can be used for prediction of Nuavg.