An inverse methodology to estimate the orthotropic thermal conductivities of AMP20M1HD-A pouch-type Li-ion battery

Abstract

Electric vehicles (EVs) have emerged as a promising solution for addressing some of the crucial sustainable development goals like affordable and clean energy. However, they are susceptible to battery thermal runaway raising serious safety concerns. To mitigate these concerns, it is imperative to design an effective battery thermal management system (BTMS). A knowledge of the thermophysical properties of the battery is a prerequisite for designing an effective BTMS. It is challenging to measure these properties as the batteries exhibit anisotropic behavior. The current work employs the powerful Metropolis Hastings–Markov Chain Monte Carlo (MH-MCMC) coupled Bayesian-inference-based inverse methodology driven by an artificial neural network (ANN) to estimate the orthotropic thermal conductivities (k_xx, k_yy, and k_zz) of an actual AMP20M1HD-A pouch-type Li-ion battery using experimentally measured surface temperatures at suitable locations for five different heat inputs. The obtained average mean estimates (k_xx = (20.14 ± 1.44) W/mK, k_yy = (2.90 ± 0.2) W/mK, and k_zz = (21.47 ± 1.39) W/mK) were found to be in close agreement with the values reported in the literature. The impact of temperature on k_xx, k_yy, and k_zz was studied, and the results show that the properties are temperature-independent. Based on the sensitivity analysis conducted, the most and the least sensitive thermocouples in the estimation of k_xx, k_yy, and k_zz were identified. The outcomes of the current work show the superiority of the proposed inverse methodology in the thermal characterization of live batteries using minimal temperature measurements and temperature measuring devices having varied levels of uncertainty.