With the complexity and aging of rotating equipment, the frequency of equipment failure gradually increases. To ensure the normal operation of rotating equipment in thermal power plants, the condition monitoring and fault early warning model of rotating equipment is researched. It is innovatively proposed to use wavelet packet transform and multivariate state estimation technology to extract the fault characteristics of rotating equipment and build a fault early warning model based on similarity model, K-means clustering algorithm and principal component analysis method. The experimental results show that the proposed model can significantly improve the accuracy of the feature parameters;the early warning threshold is 0.77. The model issues an alarm when the similarity of the data is lower than the early warning threshold in about 58 min. This is consistent with the actual fault generation time. Therefore, the proposed model has good fault early warning effect and has certain feasibility and practical application value. This research helps to promote the development of fault monitoring and early warning technology, and can provide technical support for the management of rotating equipment in thermal power plants.