Green cognitive radio power distribution method based on deep reinforcement learning

The invention relates to a green cognitive radio power distribution method based on deep reinforcement learning, and the method comprises the steps: firstly building a power distribution model, and carrying out the training of the power distribution model according to the following steps: S1, initializing the number of times of round training, the capacity of a memory pool, and random parameters of a deep neural network; s2, initializing the state at the beginning of each turn; s3, selecting an action according to a greedy strategy in the tth step of each round; s4, the action is input into the cognitive wireless environment, the environment returns a report, the available energy of the battery is updated, the state is updated, and the state is transferred and stored into a memory pool; s5, randomly sampling set batch state transition from the memory pool, and executing a gradient descent step; and then power distribution is carried out through the trained power distribution model. According to the method, op