An experimental investigation and hybrid neural network modelling of thermal management of lithium-ion batteries using a non-paraffinic organic phase change material, Myristyl alcohol

Abstract

The usage of electric vehicles is substantially increased worldwide to reduce greenhouse gas emissions and to decrease the dependence on crude oil. Li-ion battery (LiB) has a promising future due to their lightweight, high-energy density, and can be easily charged and discharged, with a relatively long lifespan. The lifespan/durability and performance of a battery depends dominantly on its operating temperature. Hence, cooling strategies for Li-ion battery is essential for better lifespan, safety, and performance. In this work, experiments are conducted using a stack of 20 Li-ion cells in a 5-series-4-parallel configuration at various discharge C-rates for natural convection air and PCM-assisted cooling methods. In comparison with the natural convection air-cooling scenario, the battery module maximum temperature is reduced by 21.66%, 25.96%, 31.71%, 31.72%, and 34.48% in the case of PCM-assisted cooling for discharge C-rates of 1C, 1.5C, 2C, 2.5C, and 3C, respectively.The use of PCM has increased the time for which the battery module is maintained below the desired operating temperature compared to the natural convection air cooling by 127.84%, 98.18%, and 115.38% for the 2C, 2.5C, and 3C discharge C-rates, respectively. The fraction of total discharge time under desired operating conditions maintained during the PCM cooling scenario is 100%, 75.69%, and 46.66% for discharge C-rates of 2C, 2.5C, and 3C, respectively. The battery module's maximum temperature is maintained under desired operating conditions by PCM cooling up to a discharge C-rate of 2C. This work shows that passive thermal management with Myristyl alcohol as a phase change material is safe and effective for high-current and high-energy-density Li-ion batteries. In addition, a prediction model using ANFIS has been created to estimate the battery module's maximum temperature values and found that a combination of gauss input and linear output membership function is the best model with a coefficient of determination of 0.99.