Using Deep Reinforcement Learning for 5G RAN Slicing Resource Allocation in New Power Load Management System

In recent years, extreme weather has caused unstable power generation on the power side, posing a great need to build a new power load management system to unleash the ability of flexible load regulation for ensuring the balance between power grid sources and loads. The new power load management system needs to monitor and regulate dispersed load resources, thus posing higher requirements for the real-time performance and resource guarantee capabilities of communication networks. Currently, there is no slicing algorithm and communication resource allocation algorithm specifically designed for the new power load management system in 5G communication networks. To address these issues, this paper categorizes the service of the new power load management system and uses 5G slicing to carry it out, establishing a new power load management system 5G radio access net-work(RAN) slicing resource allocation model. The slicing resource allocation problem is transformed into a Markov decision process, and the state space, action space, and reward function are designed. Additionally, a dueling deep Q network(DQN) based 5G RAN slicing resource allocation algorithm is proposed, which learns from the historical resource needs of slices to provide the optimal resource allocation strategy. Simulation results show that the proposed algorithm can meet the communication resource needs of the new power load management system and has very high resource utilization efficiency.