A prediction algorithm of tropical cyclone (TC) intensity change based on deep learning is proposed by exploring the distribution characteristics of atmospheric and oceanic elements. we adopted three dimensional convolutional neural network (3D-CNN), which is part of a most advanced approach, to learn the implicit correlation between the spatial distribution characteristics of three dimensional environmental variables and TC intensity change. Image processing technology is also used to enhance the data of a small number of TC samples to generate the train set. On the basis of TC instantaneous three dimensional state and the influence of sea surface temperature, we extract the spatial hybrid features from TC image patterns to predict 24 h intensity change. Experimental results show that the Mean Absolute Error (MAE) of TC intensity change prediction and the accuracy of strengthening and weakening classification are both have a significant improvement