The invention discloses a deep reinforcement learning-based post-5G forward network transmission time synchronization method and device. The method comprises the following steps: obtaining synchronousservice request information and network resource state information; sending the synchronous service request information and the network resource state information to a pre-trained deep reinforcementlearning route optimization model, and obtaining a minimum asymmetric delay route according to a learning result of the deep reinforcement learning route optimization model; obtaining the time of a master clock and a slave clock, and calculating the time error of the master clock and the slave clock; and when the time error is kept in a preset range, allocating network resources to the synchronousservice request information based on the minimum asymmetric delay route, and generating a synchronous service result. According to the method and the device provided by the invention, the influence of asymmetric delay on the synchronization signal in the transmission process can be minimized, so that the synchronization precision is improved, and the later 5G application requirement is met.