In order to solve the problem that traditional routing algorithms and existing intelligent routing algorithms are difficult to take into account the advantages of energy consumption, delay, packet loss and throughput, this paper proposes an energy efficient routing algorithm Route-A3C based on improved asynchronous advantage actor-critic (A3C) in the software-defined network (SDN) environment. Route-A3C optimizes both network energy consumption and various performance indicators during the training process, and introduces the Next network into the A3C algorithm framework, which can predict the state of the network and reduce training complexity. The Route-A3C trains the weights of each link that can be output after convergence, so as to calculate the network transmission path and realize energy efficient routing planning. During the experiment, we compare the Route-A3C algorithm with ECMP, DQN-EER and DRL-R. The Route-A3C routing algorithm has a good convergence effect, and it reduces the energy saving percentage by 12.58%, reduces the delay by 6.71% and the packet loss rate by 11.63%, and increases throughput by 4.26%.