The defense of PhD thesis of Ali El Amine will be held on November 12th at 14:00 at IMT Atlantique, Rennes Campus , Petit Amphi. The thesis is entitled “Radio resource allocation in 5G cellular networks powered by the smart grid and renewable energies”.
The heated 5G network deployment race had begun between competitors to outperform one another and be the most innovative followed by a rapid progress towards standardization.
Unlike previous generations, 5G envisions to support extremely wide diversity of services and applications with different requirements in terms of reliability, network availability, data rate and latency with high energy efficiency. This thesis focuses on studying the role of energy and its behavior while designing and operating wireless cellular networks. We consider different and complementary approaches and parameters, including energy efficiency techniques (i.e., radio resource management and sleep schemes), renewable energy sources, smart grid and tools from machine learning, to bring down the energy consumption of these complex networks while guaranteeing a certain quality of service adapted to 5G use cases.
In the past decade, a lot of research efforts have been devoted to design energy efficient cellular networks not only to reduce the energy consumption, but also to limit the carbon footprint induced by these systems. Focusing on renewable energy to bring down the Operational Expenditure (OpEx) costs for telecom operators while respecting the environmental regulations opened opportunities for new business models. However, it is not trivial to design and operate such networks given the complex nature of these networks. In addition to radio resource management, green cellular networks require the optimization of using renewable energy that entails high management complexity due to its erratic and intermittent nature. Moreover, considering the energy storage element (i.e., battery), a critical component in renewable energy-equipped systems, and the Smart Grid environment provide additional dimensions to the problem and open new research challenges.
In this work, we start by a review of the literature in order to identify the different research directions in energy efficiency green wireless networks. Following the extensive research under these areas, we highlight several aspects in which the problematic of green cellular networks needs more exploration.
Consequently, we first study the effect of equipping base stations with renewable energy sources. Due to the high capital costs of these systems and the infeasibility to equip renewable energy systems to all base station sites, we study the percentage of sites to be powered with hybrid energy supplies (renewable energy and Smart Grid). In particular, we focus on the impact of equipping sites with renewables on the operational cost and the performance of a cellular network to decide how much to invest in renewable energy, i.e., number of sites equipped with renewable energy sources, sizing of renewable energy and battery capacity. Our study for instance shows that it is enough to equip 30% of sites with renewables in order to realize an operational cost gain of 60%.
Then, we evaluate the contribution of each service the network is providing and their effect on the network energy consumption. We consider Key Performance Indicators (KPIs) putting forward each service contribution to energy consumption. Using these KPIs, we propose some energy management strategies, leading to performance amelioration and energy savings up to 11:5% points compared to other benchmark algorithms, under renewable energy and smart grid environment.
Focusing on the storage element (i.e., battery) that requires expensive investment cost both in terms of Capital Expenditure (CapEx) and OpEx, we include important constraints on the battery that is prone to irreversible aging mechanisms to expand its life span. Then, we propose several energy management algorithms that aim at saving energy while respecting the battery constraints. Our results show a gain of 20% in terms of electric bill reduction compared to an existing algorithm, and 35% battery life time enhancement.
Recently, artificial intelligence has received significant attention as a highly effective alternative to conventional methods. In particular, we take advantage of Reinforcement Learning (Q-learning) to orchestrate different levels of sleep modes to save energy given a user Quality of Service (QoS). By considering advanced sleep mode levels compliant with 5G requirements, we demonstrate the performance of these sleep schemes under the energy-delay-tradeoff problem.
