Deformable registration of lung Computed Tomography (CT) is an important research topic in the field of medical image registration, which can help doctors better observe the changing pattern of lung respiratory motion of patients and is of great significance for tracking lung respiratory motion, disease diagnosis, and radiotherapy. In this paper, a fully convolutional deformable registration method with residual modules is proposed for the registration of 4D-CT images of lungs. The residual blocks are inserted into the ordinary Fully Convolutional Network (FCN) to increase the depth of the intermediate layer hence improving the feature representation ability of the network to register image pairs. At the same time, in order to improve the registration multi-scale convolution into the network. During training, unsupervised learning is used to deal with the problem of less labeled data. Experiments show that the proposed method can effectively improve registration accuracy, and the registration speed can meet the needs of practical use.