Aiming at the limitations of traditional machine learning models in predicting the temperature of heavy tractor hydro mechanical continuously variable transmission(HMCVT)wet clutch, an improved grey wolf particle swarm optimization-support vector machine(IGWPSO-SVM) HMCVT wet clutch friction pair temperature prediction model was proposed. Firstly, the thermal analysis of the sliding friction of the wet clutch friction pair was conducted to determine the factors that affected the temperature of the wet clutch friction pair. Then a temperature prediction model was built based on support vector machine(SVM), and the structural parameters of SVM were optimized by using improved grey wolf particle swarm optimization(IGWPSO)algorithm. Finally, based on the HMCVT wet clutch test rig data, a sample database of the clutch friction pair temperature prediction model was established, and the IGWPSO-SVM model was tested and validated using the dual steel plate of the wet clutch friction pair. The experimental results showed that the mean absolute error(MAE), mean square error(MSE), root mean square error(RMSE), and mean absolute percentage error(MAPE)predicted by the IGWPSO-SVM model for the inner diameter, pitch diameter, and outer diameter of the dual steel sheet of the friction pair were 3.3557℃, 24.3212℃2, 4.5976℃ and 3.95%, respectively, the maximum temperature prediction errors were 7.8700℃, 5.4300℃ and 0.9900℃, respectively. The average values of three tests for MAE, MSE, RMSE and MAPE were 3.3522℃, 24.7380℃2, 4.9737℃ and 4.12%, respectively. The maximum temperature mean absolute error(MTMAE) for inner diameter, pitch diameter, and outer diameter was 4.3733℃, which was the lowest compared with that of the other four existing models. The research results can provide a theoretical basis for highprecision prediction of temperature of wet clutch of heavy-duty tractors and reliability of entire vehicle.