Abstract: Based on the statistical analysis of meteorological temperature records from 1970 to 2019 in six representative cities in China, the probability distribution of reference air temperature was determined. The value law of closure temperature was investigated, and the allowable fluctuation range of closure temperature was determined. According to the probability distribution function of reference air temperature and closure temperature, the positive and negative temperature difference were determined, and then the partial factor and combination value factor of thermal action were calculated by mathematical statistics method. For various structural members under different loading actions, the reliability indices of partial factor and combined value factor of thermal action were calculated by first-order second-moment method, and the rationality of partial factor and combined value factor of thermal action was obtained. The analysis results show that the monthly average maximum temperature of the highest temperature month and the monthly average minimum temperature of the lowest temperature month in China both obey extreme I distribution. The closure temperature should be close to the annual average temperature, and approximately follow uniform distribution within the allowable temperature range. The partial factor of thermal action increases with the increase of closure temperature range, and the combined value factor of thermal action decreases with the increase of closure temperature range. With the increase of thermal action, the reliability index of partial factor first increases and then decreases, and the reliability index of combined value factor gradually decreases. When the partial factor is 1.5 and the combined value factor is 0.6, the requirements on the corresponding reliability index can be satisfied.