A Hybrid PCA-PSO-SVR algorithm was used to model the grade efficiency of PV cyclone separators to describe the complex non-linear relationship between structural parameters and operation parameters. Principal component analysis(PCA) was used to reduce the dimension of the experimental data set, and particle swarm optimization(PSO) was used to optimize the parameters for the support vector regression(SVR). The support vector regression model of the cyclone separator grade efficiency optimized with PSO was compared with the multivariate regression model and other machine learning models in the prediction accuracy, generalization, robustness, and the operating speed. The results show that the PCA-PSO-SVR algorithm is an accurate and effective method to model the graded efficiency of PV cyclone separator.