Abstract: In response to the problems of high-dimensional redundancy and low computational efficiency in feature extraction of axial flow fan bearings in thermal power plants, a research on rapid identification of axial flow fan bearing faults in thermal power plants based on ReleifF feature extraction is carried out. By utilizing the ReliefF algorithm to efficiently screen the initial feature set, a low dimensional sensitive fault feature subset is obtained to reduce data dimensionality. Combining correlation dimension calculation to enhance the nonlinear dynamic representation ability of fault features, in order to improve the model generalization and classification efficiency in small sample scenarios. Through practical applications, it has been proven that this method can effectively achieve rapid and accurate identification of bearing faults in axial flow fans, demonstrating significant advantages in both identification efficiency and accuracy. It provides an effective solution for the intelligent operation and maintenance of key equipment in thermal power plants.