Abstract: Temperature slip exists in the phase transition process of the mixtures, which can better match the cold and heat sources, so as to improve the performance of organic Rankine cycle (ORC). Due to the large number and combination of mixtures, it is difficult to use conventional methods to select mixtures and optimize them synchronously with system parameters. A synchronous optimization model of three-stage screening and ORC system of mixtures based on the calculated physical properties of Perturbed-Chain Statistical Associating Fluid Theory (PC-SAFT) was established. The initial value model of gas-liquid equilibrium was predicted based on eXreme Gradient Boosting of machine learning. Genetic algorithm was used to realize the synchronous optimization of working fluid screening and system parameters. From the pure working fluid pool composed of 26 kinds of common working fluids, the suitable mixtures were selected according to the temperature slip and pressure, which effectively reduced the search range. The mixtures R12/Perfluoro-pentane and R40/R160 were obtained for net work and exergy efficiency, respectively. The optimization results of working fluid combination and ORC operation parameters under different heat source inlet temperatures and different targets were compared.