A data-driven performance analysis and prediction method for electric vehicle cabin thermal management system

Urban conditions and passenger behaviors primarily affect the performance of the electric vehicle (EV) cabin thermal management system (CTMS). The systematic analysis and accurate prediction of performance variations are crucial for the holistic optimization of the system. This study evaluated the effect of different ambient temperatures, solar radiations, driving patterns, and preset temperatures in Beijing, China. Energy analysis, exergy analysis, and the adaptive predictive mean vote method were employed to comprehensively analyze the performance of air conditioning (AC) system and cabin thermal comfort. A support vector machine model based on the multi-improved snake optimizer (MISO-SVM) was proposed to predict the CTMS performance. Chaotic mapping, subtraction-average-based optimizer, and lens imaging reverse learning strategy were applied to enhance the optimization capabilities of the traditional snake optimizer. A CTMS model was developed and validated to facilitate analysis and dataset acquisition. The results indicate that the energy consumption increases dramatically with elevated ambient temperatures (9.27 times increase from 26 °C to 40 °C) and solar radiation (11.42 times increase from 0 to 1200 W/m²). Conversely, high-speed driving patterns and higher preset temperatures (16 °C to 26 °C) are associated with improved AC system performance, leading to a reduction in energy consumption by up to 79.59 %. Moreover, the MISO-SVM model outperforms the traditional prediction models regarding CTMS performance prediction efficiency. The optimized MISO algorithm reduces the average convergence iterations by 20.59 % and achieves an average reduction of 18.92 % in fitness values. Across diverse prediction tasks, the MISO-SVM consistently maintains R² above 0.99 and the lowest mean squared error. This study may provide new insights into the performance analysis and prediction of CTMS in EVs.