The Predicted Heat Strain (PHS) model predicts physiological responses of European people to heat stress; while its performance for Chinese population remains underexplored. The study conducted a heat exposure experiment (temperature: 35 °C/38 °C/40 °C, relative humidity (RH): 25%/40%/60%) in a well-controlled climate chamber. 10 male Chinese workers were recruited to perform exercises on the treadmill at a speed of 0.5m/s for 120min, to simulate moderate metabolic rate. Rectal temperature(Tre), skin temperature(Tsk), sweat rate(SR) and heart rate(HR) were monitored continuously; the corresponding predicted values were obtained by the PHS model. The results showed that the measured Tre, Tsk, SR increased significantly with increased temperature and humidity. The PHS model overestimated the maximum allowable exposure time of subjects but underestimated the final Tre and Tsk. Bland-Altman analysis showed that the differences and 95%CI between the observed and predicted values increased with increasing temperature and RH and exposure time, indicating the significant prediction deviation of the PHS model. Through adjusting the initial Tre from 36.8 °C to 37 °C, the protection efficacy was improved from original 24.7% to 57.1% for the PHS model. The protection efficacies were further improved to 71.2% through adjusting the maximum HR based on ages, and to 68.2% through adopting the real-time HR to predict metabolic rates. The proposed three methods improve the heat strain prediction in the PHS model for Chinese workers and are more applicable in practical hot working place. This benefits to policy decisions and 2 occupational safety protection for Chinese workers with heat exposure risks.