In order to realize the fast calculation of transient temperature field of oil-immersed transformer windings, this paper proposes a real-time reconstruction method that integrates the Gappy Proper Orthogonal Decomposition (POD) method and the temperature of discrete monitoring points. Firstly, the dominant modes are extracted based on the historical simulation data of the transformer windings using the POD method. Then, a greedy algorithm is used to select the best monitoring point locations. Finally, the modal coefficients are solved based on the Gappy POD method using the discrete monitoring point data, and then the rapid reconstruction of the temperature field is achieved. In this paper, based on the basic structure of product-level 110 kV transformer windings, a simulation model is constructed to verify the effectiveness of the method. The results show that, for the analyzed example, compared with the simulation software, the method proposed in this paper has a maximum average absolute error of about 0.60 K for the whole field and a maximum hot-spot error of only 0.12 K, which is highly accurate for the calculation of the whole field temperature and hot-spot temperature. After obtaining the snapshot data, the average time consumed by the Gappy POD method to calculate each time step is about 0.07 s, which has high computational efficiency.