Sustainability in Mobile Networks: Analyzing the Feasibility and Benefits of Infrastructure Sharing

Currently the advent of 5G technology, offering higher data rates, stricter latency requirements, and increased connectivity, is posing relevant challenges to the deployment of future communication networks. As traffic volumes increase, concerns regarding the energy consumption and efficiency of base station infrastructure in mobile access networks grow. These infrastructures are typically designed to meet peak demands, but they may remain underutilized for most of the time, resulting in significant energy waste and financial costs. This thesis addresses these challenges by proposing a novel strategy called network sharing (NS), in which two or more mobile network operators (MNOs) accept to share their own radio access network (RAN) infrastructure, coming to an agreement to collaboratively utilize portions of it to serve their customers. This strategy aims to decrease energy usage by minimizing the active periods of infrastructure during off-peak periods of traffic demand. This approach maximizes energy efficiency while meeting communication requirements to improve the resilience of future mobile networks. Based on real mobile traffic data in different areas, from densely populated urban environments to less congested rural areas, this study devises data-driven methodologies to dynamically offload traffic among Base Stations owned by different MNOs, which allows to deactivate redundant resources. The results demonstrate that network sharing is effective in achieving huge energy saving and a significant reduction in electricity expenditures. In most cases, network sharing between pairs of base stations owned by two different MNOs could result in energy savings ranging from 30% to 40%. Subsequent considerations lead to develop an optimal strategy aiming at defining the upper limits on the maximum number of base station on/off switching operations to reduce excessively frequent switches, thus improving the strategy's practicality and better preserving the network devices from degradation phenomena. Furthermore, this study utilizes time series machine learning (ML) techniques to evaluate the impact of traffic prediction inaccuracies on system performance during the real-time application of network sharing algorithms. The results indicate a relatively high prediction accuracy of around 85% and show that even under traffic forecasts affected by some error, energy savings of up to 35% can be achieved. All these outcomes prove that network sharing is both a feasible and beneficial strategy, highlighting the economic benefits brought by green network development and enabling a sustainable evolution towards 5G and beyond networks.