Comparative Analysis of a KNN-based Base Station Switching Mechanism for Energy Consumption Reduction in Different 5G RAN Deployment Scenarios

Résumé fourni par la source

The superior performance of $\mathbf{5 G}$ networks as compared to previous mobile technologies has contributed to their rapid worldwide deployment. However, energy consumption constitutes a major challenge to the long-term sustainability of 5G technology. Machine Learning (ML) techniques have emerged as reliable and powerful tools to reduce energy consumption in 5 G networks by harnessing resources more efficiently based on prevailing network conditions. ML-driven features such as adaptive sleep modes for base stations and smart resource allocation have significantly enhanced the energy efficiency of 5G systems whilst preserving the required levels of user experience. This paper analyzes the performance of a k-nearest neighbor (KNN) algorithm to optimize the operation modes of mMIMO base stations for three different 5G Radio Access Network (RAN) scenarios. Results show that reductions of 15,8, and 13% in energy consumption is achieved with KNN-based switching for the 5G D-RAN, CRAN, and Cloud-CRAN scenarios, respectively.