With the rapid development of new power systems mainly based on renewable energy resources, coupled with the increase of seasonal loads, the problems of heavy/overload transformer during peak summer, winter and major holidays are becoming more and more serious every year, which seriously affects the service life of transformers and the reliability of power supply. One of the important means to solve the transformer heavy/overload problem is to adjust and transfer the load in advance by load prediction, so it is crucial to achieve accurate prediction of transformer load. In this paper, A shortterm load prediction model based on long and short-term memory recurrent neural network (LSTM) with deep learning is built to predict the load level of transformers in the coming week using the historical load of transformers, holiday information and meteorological data as input. The accuracy of transformer load prediction is verified on Matlab using two 35kV transformers with a capacity of 5000kVA as an example.