Abstract: The wind temperature prediction of extractive working space is an important basis for the management of underground thermal pollution. A model of wind temperature prediction in tunneling working site based on PSO−SVR (Particle Swarm Optimization-Based Support Vector Regression) was established to accurately predict the wind temperature in tunneling working site, and the advantages of the PSO−SVR algorithm were analyzed by comparing it with the Multiple Linear Regression (MLR) model and the traditional SVR model with the standardization of the parameters by the “Trial and Error Method”, and finally it was applied to wind temperature prediction and local cooling of Ji-24120 return airway bottom roadway in No.10 Coal Mine, Pingdingshan Tianan Coal Mining Co., Ltd. The results show that the PSO−SVR model has the best prediction performance, and the average error of the model is only 1.81%, which is reduced 62.6% compared with the traditional SVR model. It is obvious that the optimisation of the model parameters by PSO plays an important role in improving the fit, generalisation and prediction accuracy of SVR. In addition, based on the results of wind temperature prediction by PSO−SVR algorithm and the wind temperature requirements of “Coal Mine Safety Regulations”, calculated that the cooling capacity of Ji-24120 return airway bottom roadway is 1 083.28 kW, and using the above results for local cooling, the wind temperature of the working face is average reduced about 8.6 ℃, the cooling effect is significant, which shows the reliability and feasibility of the wind temperature prediction model of PSO−SVR.