The continuous increase in electricity demand, combined with factors such as political instability, cyberattacks, and the rising frequency of natural disasters linked to climate change, poses significant challenges to the reliability and stability of the power grid. Failures in the power grid can have cascading effects on the communication infrastructure, which heavily depends on a stable electricity supply. Enhancing the resilience of computing and communication facilities is therefore essential to support critical aspects of daily life. To address this, we view a group of Base Stations (BSs) in a Radio Access Network (RAN) as consumers and producers within a Micro Grid (MG) equipped with Photovoltaic (PV) panels, energy storage, and interconnected by dedicated power cables for energy exchange. We propose novel RAN resource and energy management strategies designed to maximize RAN Quality of Service (QoS) during Power Grid Outages (PGOs), given the available energy within the MG. Our evaluation considers factors such as the number of BSs in the MG, PV panel capacity, PGO duration, BS traffic profiles, PV panel placement, the energy battery employment, and the geographical extent of the MG. Results demonstrate that the proposed methodology improves QoS, increasing the hourly Managed Traffic by up to more than 300%, an improvement obtained when the MG consists of 4 2-kWp-PV-equipped BSs, compared with isolated 2 kWp PV-equipped BSs, during daily hours. When the MG is implemented, small PV panels ( ≤ 6kWp) perform comparably to large ones ( ≥ 12kWp) in isolated BS setups, making the solution space-efficient. Additionally, performance is mostly unaffected by BS traffic profiles or PGO duration. Effective energy management, accounting for cable losses, and the central placement of PV panels within the MG are critical for optimizing performance.

To design efficient Radio Access Networks (RANs) capable of handling the increasing power demand of modern networks, it is crucial to accurately assess the power consumption of a Base Station (BS). Since existing models are outdated, we propose a new data-driven model that accurately reflects the power demand of modern BSs. We derive the model using real-world measurements from operative three-sector BSs, differing for technologies and transmission frequencies. Our findings suggest that models tailored to specific technologies and transmission frequencies outperform generalized ones. Moreover, linear regression models consistently perform up to 96% better than those based on Multilayer Perceptrons (MLPs), Recurrent Neural Networks (RNNs) and benchmarks from the literature, highlighting a predominantly linear relationship between input features and power needs. Finally, the most accurate power estimations are produced by a linear model using traffic volume, load, maximum transmission power, and cable power losses, as regressors, with errors ranging from 4 W to 38 W.

The growing demand for mobile connectivity is pushing terrestrial radio access networks (RANs) toward congestion, especially during peak hours in dense urban areas. High Altitude Platform Stations (HAPSs) provide a flexible means to support overloaded base stations (BSs) through aerial traffic offloading. This paper presents a simulation-based framework that evaluates HAPS-assisted offloading under realistic traffic traces, considering baseline and upscaled demand scenarios. We assess packet-level delay, user denial, energy consumption, and infrastructure expansion. Results show that HAPS markedly reduce queuing delay and nearly eliminate user denial under moderate overload, while also lowering operational and capital expenditures by relieving congested BSs. These findings confirm the viability of HAPS as a complementary aerial layer for congestion mitigation, postponing the need for terrestrial densification in the path toward B5G networks.

Artificial Intelligence (AI) applications are becoming more widespread, raising significant environmental concerns due to the high energy use required to train large Deep Neural Networks (DNNs). To address this issue, we conduct a detailed quantitative analysis of the energy consumption of various models across different training steps and datasets. Our study measures energy use at each training stage with a comprehensive review of selected models. By tracking energy consumption during each training step, we find that the backpropagation phase is the most energy-intensive. Additionally, we evaluate the power limits and performance features of Graphics Processing Units (GPUs), collecting empirical data on their behavior under different GPU power settings. Based on these insights, we explore the integration of locally installed renewable energy sources, such as solar power and battery systems, with the electrical grid to enhance the energy sustainability of GPU operations. We introduce and test an innovative approach for managing energy and computing resources, aiming to optimize energy use and reduce operational costs. Our results demonstrate that this method can reduce energy consumption by more than 40% and operational costs by almost 25%, paving the way for greener AI solutions.

Telecom base stations increasingly rely on solar power and battery storage to achieve sustainable, cost-effective energy solutions, but battery degradation poses a significant challenge to system reliability and longevity. This paper introduces an innovative optimization framework that accounts for lithium-ion battery aging, modeling both calendar and cycle degradation with a novel segment-based approach. Designed for seamless integration into cost-effective energy planning, the framework optimizes photovoltaic (PV) panel and battery sizing to minimize costs and extend system lifespan. Validated using real-world base station power consumption data, our approach outperforms traditional rainflow-based aging models, reducing battery cycle wear by up to 65.5% compared to aging-unaware methods and by an additional 10% over rainflow-based methods. By enabling real-time battery health tracking, it supports dynamic energy management, offering a practical solution for sustainable telecom networks.

Unmanned aerial base stations (UABSs) are a promising solution for improving coverage and capacity in vehicle-to-everything (V2X) communications, particularly in dense urban areas. However, their operation is constrained by onboard energy consumption, required for both flight and communication. Beamforming, while enhancing network performance, adds to this challenge due to its energy-intensive nature.This paper proposes a sequential and hierarchical decision-making framework for UABS operations, considering trajectory planning, dynamic beamforming, and radio resource assignment (RRA). While heuristic and optimal solutions are employed for trajectory planning and RRA respectively, the beamforming model is modeled as Markov decision process (MDP) to maximize served user demand, weighted by time-varying priorities, under strict energy constraints. Leveraging a dueling double deep q-network (3DQN) algorithm that penalizes energy budget violations, an agent policy for the beamformer is then trained. Simulation results demonstrate that the proposed approach outperforms static beamforming benchmarks and closely matches an ideal step-wise oracle, achieving a balance between energy efficiency and served user demand while adapting to dynamic V2X traffic conditions.

In this paper, we consider a Radio Access Network (RAN) powered by the Smart Grid (SG), which provides monetary incentives to users who respond to the grid's explicit requests to increase or decrease their energy consumption. The typical solution for reducing the RAN's power needs is to employ Base Station (BS) sleep modes, but this may lead to a drop in user coverage. For this reason, we propose dynamically adjusting the bandwidth allocated to users in response to SG requests. Our study considers a realistic urban RAN and evaluates the effectiveness of bandwidth limitation in meeting the SG's power reduction requests. Results indicate that this approach is particularly effective in high-density user environments, reducing power requirements by up to 20% while maintaining sufficient bandwidth for essential applications such as audio and video streaming. In contrast, BS deactivation can lead to substantial coverage losses, with user coverage falling below 90%.

In spite of promising outcomes in optimizing energy usage for Radio Access Network (RAN) Base Station (BS) hardware, deployment, and resource management, existing methods frequently lack flexibility for scenarios involving multiple frequencies and technologies of BSs. This investigation presents a comprehensive BS switching strategy based on a threshold, tailored for real-world multi-frequency and multi-technology BSs within the RAN. The proposed approach strategically deactivates BSs using a threshold parameter that determines the maximum allowable growth in transmission power for active BSs, ensuring both coverage for users affected by BS deactivation and energy saving . Simulations conducted on a realistic multi-technology 5G New Radio (NR) RAN in an urban environment validate the efficacy of the proposed strategy, achieving up to 73% of energy saving. The study assesses the influence of the frequency order of BS deactivation and examines user re-association strategies aimed at minimizing either path loss or transmission power.

The increasing adoption of Deep Learning (DL) algorithms for time series forecast has led to a significant environmental concern due to the high computational demands and associated carbon footprint. This study investigates the environmental impact of DL models, particularly Long Short-Term Memory (LSTM) networks, for time series forecasting tasks where frequent retraining of models is essential. We conduct an empirical analysis of carbon emissions produced by LSTM models trained on two distinct time series datasets. By systematically varying model hyperparameters (epochs, train-test split, number of layers and neurons per layer), and by reducing the number of models or the number of input features, we aim to understand the impact of these changes on carbon emissions and model accuracy. Our contributions include a comprehensive analysis of carbon emissions during model training and the identification of possible tradeoffs between emissions and accuracy. The findings indicate that strategic adjustments can significantly reduce environmental impact while maintaining satisfactory accuracy levels.

Due to the evolution in communications technologies and antennas, as well as advances in solar panel efficiency, High Altitude Platforms (HAPS) have been recently considered as a promising aerial network component, to support Radio Access Networks (RANs). Through their directional antenna they can activate beams and provide coverage to up to 1.5 km radius ground area. In this work, we consider a HAPS equipped with a Multi Access Edge Computing (MEC) server, which provides caching capabilities. The HAPS is used to off-load content requests. We analyse an urban environment scenario, as well as the effects of the simultaneous activation of beams in different areas. Results demonstrate that the HAPS is a suitable solution to bring additional capacity to the RAN and highlight that the provided performance strictly depends on the traffic demand profile of the covered portion of RAN.

In the next generation of Radio Access Networks (RANs), Multi-access Edge Computing (MEC) is considered a promising solution to reduce the latency and the traffic load of backhaul links. It consists of the placement of servers, which provide computing platforms and storage, directly at each Base Station (BS) of these networks. In this paper, the caching feature of this paradigm is considered in a portion of a RAN, powered by a renewable energy generator system, energy batteries and the power grid. The performance of the caching in the RAN is analysed for different traffic characteristics, as well as for different capacity of the caches and different spread of it. Finally, we verify that the usage of a strategy that aims at reducing the energy consumption does not impact the benefits provided by the mobile edge caching.

The field of networking, like many others, is experiencing a peak of interest in the use of Machine Learning (ML) algorithms. In this paper, we focus on the application of ML tools to resource management in a portion of a Radio Access Network (RAN) and, in particular, to Base Station (BS) activation and deactivation, aiming at reducing energy consumption while providing enough capacity to satisfy the variable traffic demand generated by end users. In order to properly decide on BS (de)activation, traffic predictions are needed, and Artificial Neural Networks (ANN) are used for this purpose. Since critical BS (de)activation decisions are not taken in proximity of minima and maxima of the traffic patterns, high accuracy in the traffic estimation is not required at those times, but only close to the times when a decision is taken. This calls for careful processing of the ANN traffic predictions to increase the probability of correct decision. Numerical performance results in terms of energy saving and traffic lost due to incorrect BS deactivations are obtained by simulating algorithms for traffic predictions processing, using real traffic as input. Results suggest that good performance trade-offs can be achieved even in presence of non-negligible traffic prediction errors, if these forecasts are properly processed.

The use of base station (BS) sleep modes is one of the most studied approaches for the reduction of the energy consumption of radio access networks (RANs). Many papers have shown that the potential energy saving of sleep modes is huge, provided the future behavior of the RAN traffic load is known. This paper investigates the effectiveness of sleep modes combined with machine learning (ML) approaches for traffic forecast. A portion of an RAN is considered, comprising one macro BS and a few small cell BSs. Each BS is powered by a photovoltaic (PV) panel, equipped with energy storage units, and a connection to the power grid. The PV panel and battery provide green energy, while the power grid provides brown energy. This paper examines the impacts of different prediction models on the consumed energy mix and on QoS. Numerical results show that the considered ML algorithms succeed in achieving effective trade-offs between energy consumption and QoS. Results also show that energy savings strongly depend on traffic patterns that are typical of the considered area. This implies that a widespread implementation of these energy saving strategies without the support of ML would require a careful tuning that cannot be performed autonomously and that needs continuous updates to follow traffic pattern variations. On the contrary, ML approaches provide a versatile framework for the implementation of the desired trade-off that naturally adapts the network operation to the traffic characteristics typical of each area and to its evolution.