To ensure the safe and efficient operation of the supercritical CO2 compressor, it is crucial to obtain its accurate performance maps. This study introduces an artificial neural network that has been successfully used in conventional compressor performance prediction. The supercritical CO2 compressor in The Chinese Academy of Sciences’ “High-Efficiency and Low-Carbon Gas Turbine Research Facility” is the research object. The 252 compressor data points were divided into training sets and test sets, and a complete regression work was performed through the multilayer perceptron network. Finally, the accuracy of the regression work is verified by comparing the predicted and real values of compressor power, pressure ratio, and efficiency. The results show that the prediction effect of this network is satisfactory and has considerable potential for further accuracy. After 200 epochs of training, the regression results basically meet the expected requirements, especially the performance maps prediction error of pressure ratio and power is small. This study can provide some guidance for the commissioning of the supercritical CO2 testbed, and the exploration of supercritical CO2 compressor system-level modeling method.