Abstract
The 5G network provides an appealing communication platform for the cyber-physical power system (CPPS) in support of wide access of terminals and tremendous sensing data transmission. In the meantime, the remote terminal unit (RTU) plays a salient role in sensing operation data and executing control commands in CPPS, which, however, usually suffers from energy scarcity. The 5G network provides an appealing communication platform for the cyber-physical power system (CPPS), enabling the transmission of tremendous sensing data from remote terminal units (RTUs). To address the issue of energy scarcity faced by RTUs, this paper presents an energy-harvesting RTU model for data sensing and transmission in 5G-enabled CPPS. A hybrid energy source scenario is investigated, where renewable energy is harvested from the surroundings while grid energy is used as supplementary. Then the RTU power scheduling issue is formulated to minimize the energy cost under the constraints of long-term queuing delay and average energy storage. Given the problem is NP-hard due to information uncertainty, Lyapunov optimization is adopted to transform it into a series of short-term power scheduling subproblems. Furthermore, a deep reinforcement learning (DRL) based power scheduling algorithm (TD3-PER-OPSA) is developed to facilitate online power scheduling decisions. Extensive experimental evaluations using real-world solar power generation data demonstrate that the proposed approach achieves the optimal energy-communication tradeoff and outperforms benchmark methods in terms of training convergence, energy cost, and transmission performance.
Funding Information
  • National Natural Science Foundation of China (62201212, 62271201)
  • Natural Science Foundation of Hebei Province (F2022502017)
  • Fundamental Research Funds for the Central Universities (2023JC003)
  • National Key Research and Development Program of China (2022YFB3105101)

This publication has 37 references indexed in Scilit: