A method based on Double Tree Complex Wavelet Transform (DTCWT)-Deep Belief Network (DBN) for overvoltage identification in distribution network is proposed. The three-phase overvoltage signal of the 10 kV bus is subjected to the dual-tree complex wavelet transform, and then the singular value is reduced by the singular value decomposition. The resulting singular value is input into the trained deep belief network classifier as the eigenvalue, and the seven typical internal overvoltage type identification types are realized. The proposed algorithm is trained and tested using ATP/EMTP simulation data and fault waveforms on the physics experiment platform, and compared with other classification algorithms. The results show that compared with other methods listed in this paper, the proposed algorithm has stronger feature extraction capability and higher recognition accuracy. This work is supported by National Natural Science Foundation of China (No. 51677030) and Natural Science Foundation of Fujian Province (No. 2016J01218).