Radiant cooling systems have the potential for energy saving, better thermal comfort, and space-saving. The lack of cooling load calculation methods limits the engineering application of radiant cooling systems. A universal radiant and convection time series (RCTS) set is indispensable in radiant cooling system design. In this study, an RCTS dataset with 100,000 sample rooms is generated based on the combinations of 9 room characteristic parameters. Extreme gradient boosting (XGBoost) is used to grade the room characteristic parameters and classify RCTSs. Gaussian mixture model (GMM) is used to extract the representative RCTSs for each category. The RMSE is used to measure the accuracy of the representative RCTS. The results indicate that XGBoost and GMM perform very well. Non-solar radiant time series (RTS), convection time series (CTS), solar RTS for radiant ceiling panel systems and solar RTS for radiant floor cooling systems are classified into 12, 8 and 1 categories, respectively. The representative RCTSs of all categories constitute a universal RCTS set. Fifteen room characteristic parameters are taken into account to construct sample rooms for Changsha. The accuracy of the representative RCTS was compared with the self-RCTS and heat balance method. The relative deviations of the design cooling load calculated using the representative RCTS and self-RCTSs were within ±3 % for 90 % cases of the sample rooms. The relative deviations between the design cooling load calculated using the representative RCTS, self-RCTSs, and theoretical design cooling load are within ±7 % for 90 % cases. The universal RCTSs set is applied for radiant cooling systems design.