In multi-node cooperative sensing of cognitive networks, as the number of nodes increases, and the energy consumption must increase, but the sensing performance does not necessarily improve. The nodes with less information are not helpful for the sensing performance but will increase the unnecessary energy consumption. To improve the sensing performance and reduce the energy consumption of nodes, a dynamic node selection algorithm based on reinforcement learning is proposed in this paper. The algorithm can evaluate the reliability of sensing nodes in real-time, select the nodes with the highest reliability to participate in cooperative sensing, and update the reliability of nodes in real-time through the method of combining feedback energy consumption and sensing performance. In a real-time environment, nodes with high reliability are selected to participate in cooperative sensing, and the optimal balance between sensing performance and energy consumption is achieved. The experimental results show that the proposed algorithm can reduce energy consumption and improve the perception performance at the same time. Under the same conditions, the detection probability is 5% higher than that of the traditional method, while the energy consumption is only 16.7% of that of the traditional method.