Traffic-Energy Aware Base Station Sleep Control Strategy Based on Deep Reinforcement Learning

With the large-scale deployment of fifth-generation mobile communication technology, the energy consumption and associated carbon emissions of base stations (BSs) have exhibited exponential growth. This paper investigates a base station sleep control strategy that dynamically adapts to variations in traffic load and energy supply by switching the base station's ON/OFF modes. Moreover, this paper proposes a traffic-energy-aware collaborative optimization algorithm based on the Dueling Double Deep Q-Network (TEC-D3QN). The proposed framework effectively reduces overall system non-renewable energy consumption while maintaining user Quality of Service (QoS). Simulation results demonstrate that the TEC-D3QN method outperforms existing strategies.