Amid rapid global urbanization, rail transit systems have expanded significantly due to economic, environmental, and efficiency benefits. However, enclosed transfer spaces within stations often suffer from poor thermal comfort due to high passenger density. This study examines Sigongli Station on Chongqing Rail Transit Line 3, analyzing thermal conditions in semi-open transfer spaces. Field data from 372 valid questionnaires and 356 environmental measurements at two test points with varying degrees of openness. The study's findings, derived from regression analysis of the collected data, yielded thermo-neutral temperatures of 23.82 °C and 22.63 °C at the two points. A comparison of the thermal satisfaction votes (TSV) regression model with the predicted mean vote (PMV) regression model revealed that the PMV model was challenging to calibrate with precision under extreme thermal conditions during summer. However, the application of the extended predictive mean vote (ePMV) modified model led to a substantial enhancement in PMV correction, with an improvement of 66.1% in prediction accuracy. This improvement was particularly notable in naturally ventilated high-temperature environments. However, under high-temperature, high-humidity, and windless conditions, the corrective effect of ePMV remains unsatisfactory. Consequently, a constant based on ePMV was introduced, and a novel PMV modified model for predicting and evaluating thermal sensation in high-temperature, high-humidity environments during summer was established: PMV’ = 0.5816PMV + 0.5806. This significantly improves the prediction accuracy over the traditional PMV model. Finally, improvement strategies are proposed and reflections on this study are made in order to provide lessons and references for subsequent studies on thermal comfort in rail transportation.