The safe and stable operation of electrical equipment in hydropower plants is critical. In order to achieve the automatic early warning of equipment faults, through infrared features extraction and gray-associated analysis, it is proposed to establish an early warning model of equipment failure with principal component analysis (PCA) and density-based clustering algorithm (DBSCAN). Firstly, the missing data are made up through the data pre-processing, the abnormal data are eliminated, and the principal component analysis is performed to reduce the dimensionality and extract the new principal component features. Secondly, the new principal components are used to construct the feature sample set by DBSCAN algorithm, establish a gray association model, calculate the gray associated coefficient, and then fail to warn the degree of change point of the gray association coefficient carry out the fault early warning through the change degree of the grey correlation coefficient of the mutation points. The experimental results show that the proposed method can effectively extract infrared characteristics and achieve equipment fault warning under the abnormal state of the equipment, and the fault warning accuracy rate reaches 97.88%.