Thermal analysis of batteries and prediction with artificial neural networks

Abstract

Purpose In this study, it is aimed to develop cooling models for the efficient use of batteries and to show how important the busbar material is. Batteries, which are indispensable energy sources of electric aircraft, automobiles and portable devices, may eventually run out. Battery life decreases over time; the most critical factor is temperature. The temperature of batteries should not exceed the safe operating temperature of 313 K and it is recommended to have a balanced temperature distribution through the battery. Design/methodology/approach In this study, the effect on the battery temperature caused by using different busbar materials to connect batteries together was investigated. Gold, copper and titanium were chosen as the different busbar material. The Air velocities used were 1 m/s and 2 m/s, the air inlet temperatures were 295 and 300 K and the discharge rates 1.0–1.5–2.0–2.5C were chosen for cooling the batteries. Findings The best busbar material was identified as copper. Because these studies are long-term studies, it is also suggested to estimate the data obtained with ANN (Artificial Neural Networks). The purpose of ANN is to enable the solution of many different complex problems by creating systems that do not require human intelligence. Four different program (BR-LM-CGP-SCG) were used to estimate the data obtained with ANN. It was found that the most reliable algorithm was BR18. The R2 size of the BR18 algorithm in the test phase was 0.999552, the CoV size was 0.007697 and the RMSE size was 0.005076. Originality/value When the literature is considered, the cooling part of the battery modules has been taken into consideration during the temperature observation of the battery modules, but busbar materials connecting the batteries have always been ignored. In this study, various busbar materials were used and it was noticed how the temperature of the battery model changed under the same working conditions. These studies are very time-consuming and costly studies. Therefore, an estimation of the data obtained with artificial neural networks (ANN) was also evaluated.