Radar jamming task can be regarded as a resource scheduling problem. Many manual decision-making approaches have low efficiency and effectiveness. In this paper, we propose an automatic radar jamming strategy generation model based on deep reinforcement learning named EWD3Q, which takes the electronic jamming aircraft interfering with ground air defense radar as the typical combat scenario. Considering the characteristics of the radar jamming task, we carefully design the environmental state space, action space and the reward function in our model and finally propose the EWD3Q algorithm training the jamming strategy generation agent. Through simulation, we realize two radar jamming scenarios containing several electronic jamming aircrafts as well as ground radars. The experimental results show that the proposed algorithm can generate efficient radar jamming strategy for those aircrafts, which achieves better performance than the manual decision-making method. This research can provide some references and supports for studies on jamming strategies generation in complex combat environments in the future.