Abstract: Deep reinforcement learning combines the perception ability of deep learning with the decision-making ability of reinforcement learning. It has achieved excellent control effects in multiple complex scenarios and has been applied to the optimization and control of distribution networks. This paper first analyzes the advantages and disadvantages of traditional optimization methods for the distribution network, and then summarizes the deep reinforcement learning algorithm. According to different application problem backgrounds, a literature review is carried out from the aspects of dispatching and control of energy storage devices, dynamic reconfiguration, restoration ability research, reactive power optimization, and voltage control. And the advantages and disadvantages of the deep reinforcement learning algorithm in different applications of the distribution network are analyzed in detail. After that, two application cases of deep reinforcement learning in engineering practice are used to further elaborate the application potential and disadvantages of deep reinforcement learning in guiding the production practice of the electric power industry. Finally, the challenges and prospects of the application of deep reinforcement learning in the distribution network are analyzed.