Abstract: Compressors are the power equipment for transporting natural gas through pipelines, and their operation scheme directly affects the overall regulation and operation efficiency of pipelines. The performance curve of compressors is the main basis for making operation scheme, but the curve is time-varying due to the influence of the changes of pipeline operation conditions and compressor performance states. By comparing the accuracy of performance curves generated by the three methods of least square, BP neural network and Adaptive Neuro-Fuzzy Inference System (ANFIS), a self-adaptive performance curve generation method based on ANFIS was proposed. The ANFIS mapping network of performance curve was modified by inputting the real-time operation data of the compressor irregularly, and the real-time performance curves in different periods were obtained. The real-time performance curve of a compressor in 2018 was generated by the proposed method, and the validity and accuracy of the method were verified. The results show that the relative error between the compressor operation curve obtained by the self-adaptive performance curve generation method based on ANFIS and the actual operation curve is less than 3%, which meets the requirement for accuracy in field engineering. The research results could provide theoretical basis for compressor performance analysis, energy consumption evaluation, and operation scheme formulation.