Abstract: In view of the difference between the actual performance curve of a centrifugal compressor and that provided by the manufacturer, the performance curve of compressor was predicted with the Least Squares Support Vector Machine(LSSVM) based on the SCADA operation data of a compressor station, and the LSSVM model was optimized with the Improved Particle Swarm Optimization Algorithm(IPSO). Besides, a performance prediction method of centrifugal compressor based on IPSO-LSSVM was developed. The results show that the iterative rate of IPSO algorithm is faster after variable inertia weight is used and the disturbance factor is added. Compared with the other prediction models, the prediction accuracy of IPSO-LSSVM model is the highest. The MRE and RMSE of outlet pressure are 0.57% and 0.055 6, respectively, while that of the outlet temperature are 0.30% and 0.137 4, respectively, indicating that the model has good prediction accuracy and fitting effect, thus capable of providing a theoretical basis for compressor performance prediction and development of antisurge measures.