Data-driven method is limited to be applied in fault location of distribution network because of the relatively small number of fault samples. In order to solve the problem of poor learning effect with small samples in deep learning, a new fault area location method based on deep convolution neural network (CNN) with transfer learning is proposed in this paper. Firstly, the characteristics of transfer learning and CNN are analyzed, the feasibility and advantages of their applied in fault area location of distribution network are also discussed. Then, a deep CNN model based on transfer learning is constructed by using ResNet50 network. The validation of the IEEE 33 bus system shows that the proposed method can locate the fault area location accurately even with limited amount of samples by using the voltage and current information of only two measurement points, and is not easily affected by the factors of transition resistance, fault type and noise.