Investigation of maximum temperatures in lithium-ion batteries by CFD and machine learning

Abstract

Alternative fuels are becoming more popular as awareness of fossil fuel depletion, pollution, and climate change grows. Numerous industrial companies are producing electric automobiles for use worldwide. Electric vehicles’ battery packs’ cooling causes firing due to high temperatures. In this study, the surface temperatures of a single electric battery with dimensions of 160 mm × 210 mm within a battery pack were investigated using computational fluid dynamics and, subsequently, Levenberg-Marquardt machine learning as a function of ambient temperature, convective heat transfer coefficient, nominal capacity of the electric battery, and discharge rate. The transport coefficient has been calculated for a rechargeable electric battery with a nominal capacity ranging from 14.6 to 20 Ah and a discharge rate varying between 1 and 15, taking into account conditions of stagnant air at temperatures ranging from 20°C to 35°C and values between 5 and 20 W/m2.K. Insufficient or absent cooling of battery temperatures can lead to them reaching combustion temperatures of electric vehicle batteries, typically from 50°C to 80°C, depending on the operational circumstances. An artificial neural network was utilized in machine learning to forecast maximum temperatures based on operating conditions without requiring simulation. The neural network achieved an estimated mean squared error of 0.00552 and a calculated coefficient of determination of 0.99. The neural network model can predict outputs with mean and standard deviation rates below 0.237. The anticipated artificial neural network model can accurately forecast the maximum surface temperature of an electric vehicle battery.