Throughput-Constrained Antenna Sleep Management for Saving Power

<copy ©2024 IFIPIn cellular networks, the rapid increase in traffic demand needs to be met by additional equipment, but the resulting increase in power consumption significantly impacts CO2 emissions. The impact on power consumption is particularly substantial due to the large number of base stations (BSs). As a result, power consumption needs to be reduced by temporarily suspending BSs with low traffic usage through BS sleep management. On the other hand, the effect of BS sleep on communication quality also needs to be considered. This paper proposes a new antenna sleep management framework to reduce power consumption while maintaining communication quality that deals with sleep on an antenna-by-antenna basis. The framework uses a two-stage deep reinforcement learning (DRL) approach that incorporates the concept of safe DRL to ensure that users’ throughput is above the target value. Results indicate that for the sleep management problem of multiple antennas within a single BS, constraint violations on throughput were reduced from approximately 0 to 1.7 percent while achieving a constant power reduction close to the optimum value.