­

Direct load control for energy harvesting systems

  • Scenario: Energy Harvesting Cellular Networks
    • Macro BS (MBS) powered by energy from the electricity grid
    • Small BSs (SBSs) powered only by energy harvested source (solar panel + battery)
    • SBSs perform traffic and energy offloading
    • Control problem: Throughput-energy trade off
    • Design of agents enabling different BS operation modes
  • Digital twin of the system based on real energy data and traffic demands
  • Control policies
    • Offline optimization
    • RL-based (centralized and distributed)
      • Tabular solutions (Q-learning)
      • DQNs
    • Imitation learning
    • Layered learning
­