Since the 4G standard TD-LTE commercialization, TD-LTE network has entered a stage of steady development, with huge network scale and complex system deployment. Therefore, the network interference problem faced by the system becomes more and more serious, which seriously affects the objective quality of the network and the quality of service to users. The common inter-network interference includes stray interference, blocking interference, jammer interference and intermodulation interference. In addition, the communication base station and terminal in the system may cause serious intra-network interference problems including ultra-distant interference, clock out-of-synchronization interference, signal amplifier/repeater interference and physical uplink control channel interference and physical uplink shared channel. Interference type identification is an important task in network maintenance and optimization of mobile operators. In this paper, an interference identification algorithm based on a deep neural network is proposed and verified on the 4G current network data of some cities in Guangxi province provided by China Mobile Communication Corporation. Experimental results show that compared with the interference analysis tool platform inside the operation, the algorithm has a higher recognition rate for 10 manual calibration categories, which can improve the efficiency of interference identification.