Energy and Spectral Efficiency Optimization in Multi-Radio-Access-Technology (RAT) Networks

Diogo de Brito Pereira

Key information

Authors:

Diogo de Brito Pereira (Diogo Pereira de Brito)

Supervisors:

Pedro Manuel de Almeida Carvalho Vieira; António José Castelo Branco Rodrigues (António José Castelo Branco Rodrigues)

Published in

December 11, 2025

Abstract

Improving Energy Efficiency in mobile networks has become an increasingly critical challenge as data traffic grows and network architectures evolve across multiple radio access technologies. MNOs have primarily focused on quantifying and optimizing the energy consumption of their infrastructures, particularly in the Downlink, creating an opportunity to analyze the Uplink, where the transmission behavior of the UE plays a decisive role. This dissertation proposes a methodology that combines theoretical modeling, empirical data analysis, and Machine Learning techniques to evaluate uplink EE and Spectral Efficiency using real network data. The methodological framework consists of six sequential stages, ranging from the estimation of UE–eNodeB distance and transmission power to the analysis of trade-offs between EE and SE through the Energy Sustainability Class UL metric. The final stage incorporates a Root Cause Analysis based on LightGBM and SHAP explanations, enabling the identification of the most relevant parameters driving efficiency patterns. Validation using real network data shows strong agreement with the uplink power control model defined by 3GPP, confirming the robustness of the estimation process. The results indicate that UE transmit power is mainly influenced by propagation conditions, UE–eNodeB distance, and the MCS. Furthermore, most BSs operate below the average in both EE and SE, revealing systemic inefficiencies. Overall, the proposed methodology integrates deterministic and data-driven approaches, providing interpretable insights that can support network optimization and more energy-aware operational decisions.