Power allocation is one of the most important factors that determine the system performance and communication quality in the densely distributed cellular networks.Because the existing power allocation algorithms often can not achieve the desired results,and the generalization ability is poor.On this basis,a power allocation algorithm based on dueling double deep Q network(D3QN)is proposed to optimize the transmission rate of the system. D3QN uses dual neural networks and competitive networks to optimize the structure of neural networks.Through the selection and evaluation of decoupling actions,the overestimation problem in DQN is solved.The simulation results show that the average rate obtained by this algorithm is 7.14%higher than that of DQN algorithm,has better performance in convergence speed and stability,and has strong generalization ability,which can be applied to different actual scenarios.