Applying Evolutionary Algorithms for Cell Switch-Off to Reduce Network Energy Consumption

Jesús Galeano-Brajones, Carlos Pupiales, Daniela Laselva, Javier Carmona-Murillo, Francisco Luna

Abstract: The rising concerns about climate change motivate the industry and academia to actively work on techniques and strategies that help reduce energy consumption in mobile networks. One avenue under exploration involves methods for intelligently and temporarily deactivating serving cells during low-traffic periods. However, cell switch-off (CSO) decisions need to be taken carefully because they may severely impact the service quality. In this paper, we delve into optimizing the CSO decisions using an optimization framework that uses a constrained distributed multi-objective evolutionary algorithm along with a 3D-maps-based radio planner simulator. Firstly, the distributed version of the evolutionary algorithm aims to find near-optimal solutions within practical computational time. The solutions should satisfy a trade-off between reducing network energy consumption and maximizing cell throughput while ensuring a minimum throughput for all the active users in the network. Secondly, the radio planner simulator evaluates the objectives and constraints of the candidate solutions obtained by the evolutionary algorithm. Our findings indicate that using an aggressive but optimized CSO strategy reduces network power consumption by up to 80% during low traffic periods compared to an all-cells-on scenario.