Abstract: Aiming at the problem of segmentation accuracy of boiler temperature field caused by the difference of temperature field feature parameters, with the goal of maintaining temperature field characteristics, the graph structure is introduced to express the field data, and the clustering analysis is carried out through the improvement of Node2vec algorithm, then the optimal segmentation of the boiler temperature field is realized. Based on multi-dimensional feature information, the method can segment the boiler temperature field and preserve the flow field characteristics more accurately. The proposed method is verified by experiments on standard datasets, and the results show that the segmentation effect is significantly improved on datasets with multi-dimensional features. Finally, the proposed method is applied to segment the temperature field of a power plant boiler, and the results show that the method can capture the local and global characteristics of the temperature field data well, and the results have good accuracy.