Satellite-ground Fusion Network Switch Algorithm Based on Kalman Filter-Deep Reinforcement Learning in Power Grid Scenes

Satellite networks can fill the blind areas of ground signal coverage of traditional communication networks, and it can provide safe and reliable communication services in remote areas and complex scenarios for the power industry. However, the satellite-ground converged network will experience network switch problems due to user mobility and network resource optimization. To better utilize the advantages of satellite and terrestrial networks, it needs to propose an efficient network switch algorithm. In view of this, this paper designed a star-ground fusion network architecture that meets the special communication needs of the power industry. And then, the paper established a received signal strength prediction method based on kalman filtering. This method can predict the user's position and speed at the next moment and realize the switch decision and execution of switch action in advance. Then, the paper proposes a deep reinforcement learning-based switch algorithm for star-ground fusion networks. The paper considers the design network as a learner that obtains an optimal decision set by continuously interacting with environmental variables. And this paper informs the user about the optimal decisions to achieve optimal network switch. Finally, The simulation experiments on the proposed network switch algorithm. The results show that the paper proposed network switch algorithm achieves the minimum switch drop rate and the number of drop times compared with other network switch algorithms. The paper proposed network switch algorithm can effectively increase the system revenue and adapt to high-speed motion scenarios and ensure the continuity of user services.