Machine-learning optimization of an innovative design of a Li-ion battery arrangement cooling system

This study employed the XGBoost, the Bayesian ridge, and the SVR to optimize a new Li-ion battery arrangement cooling system design. The battery arrangement consisted of 6 × 4 cells which were cooled by a fan. Two thousand eight hundred eleven experimental data were involved in the optimization. Measurements were done at 2 points (Temperature2 and Temperature1) and for 46 min and 51 s. One or several parameters must be calculated with other parameters or the time calculation to determine the parameters' values. The results showed that the Bayesian ridge algorithm performs perfectly for forecasting such models. Therefore it can be used in future studies primarily; this algorithm can be used to forecast temperatures. This claim is because the R-squared for forecasting Temperature1 by Temperature2 was 0.973, and the Pearson coefficient was 0.98645393. Also, the R-squared value for Temperature2 by Temperature1 was 0.975, and the Pearson coefficient was 0.98753252. The MSE for these states respectively are 0.0242, 0.1619. Employing the XGBoost, the temperatures were calculated during the time, while the R-squared value for polynomial regression was 0.80 the MSE value for XGBoost is 0.6617. Therefore, by the ARIMA algorithm can predict the future of the Temperature2 for 5622 s. By using the image processing algorithms and OpenCV library is computed the distance between Temperature1 and Temperature2 to find other points with a fixed distance. For the prediction of Temperature1 by Temperature2 and Temperature2 by Temperature1, the model is designed by the SVR with RBF kernel with the R-squared of 0.998 and the MSE of 0.0044 for Temperature2 and 0.992 the MSE of 0.0067 for Temperature1.