On the trade‐off between energy saving and number of switchings in green cellular networks

Abstract

Cellular networks are optimized by targeting multiple objectives. Usually, the different objectives are not coherent: minimizing the transmit power; the number of base station (BS) sleep-mode switchings, ie, ACTIVE/SLEEP state transitions; and the activity of the BSs and guaranteeing the quality-of-service (QoS) of users. Hence, suitable trade-offs have to be managed by network planners to provide an efficient solution to the challenge of booming mobile data. In this paper, we propose a multiobjective optimization framework aimed at minimizing the power consumption and the number of BS sleep-mode switchings in cellular networks, by jointly considering QoS requirements. These requirements are expressed in terms of a required bit rate for each mobile terminal. The framework deals with network management, such as the number of BSs that should be switched on, considering common diurnal patterns of the traffic demand. The optimization technique proposed in this paper is mixed-integer quadratic programming, which solves the joint power allocation and user association problem while also considering optimized bandwidth allocation schemes. The trade-off between the conflicting objectives, as well as the performance analysis in terms of the throughput and energy consumption of the network, is shown for different traffic load cases. The proposed optimization can obtain up to 60% energy savings during off-peak hours, guaranteeing QoS target requirements. By optimizing the network configuration, a 70% reduction in BS switch on/off operations can be reached in a day with 3% more energy expense.

Multiobjective self‐optimization of the cellular architecture for green 5G networks

“…The additional design actions (tilt, power, etc) and 3D heterogeneous sectorization (macro and VSC) performed by MOSON have led to better results in terms of energy saving with network load balancing. Dolfi et al has minimized 70% of unneeded CSO operations in a day with 3% more energy expense. This result explains our Obj 4 aiming to minimize the number of HO initiated by CSO transitions with a maximum percentage of 30%.…”

“…It is obvious that there is no deterministic algorithm able to explore the whole search space to find a solution in a polynomial time. This is normal since MOSON is a CSO problem, which is well known as NP‐hard problem . To tackle MOSON, we have used the GA to find good suboptimal solutions.…”

“…It is important to note that all objectives are normalized (0 ≤ Obj ≤ 1) for robust normalized fitness calculation. Our multiobjective self-optimization framework addresses many combinatorial NP-hard problems: capacity, energy efficiency, dynamic network planning (sectorization and design tuning), and load balancing, 9,[11][12][13][14][15][16]22,23 and these cannot be resolved in deterministic time. Their realistic projection and implementation constraints in a live network increase the complexity even further, especially when the traffic and network layout vary in temporal and spatial planes.…”

“…This is normal since MOSON is a CSO problem, which is well known as NP-hard problem. 9,[11][12][13][14][15]22,23 To tackle MOSON, we have used the GA to find good suboptimal solutions.…”

“…• We have used 3D sectorization, 3GPP 3D channel model, and 3D UE dropping that we consider mandatory to zoom in the performance of UEs located in the elevation dimension such as high-rise buildings. Most previous research studies on sectorization have used 2D channel model, passive antenna systems, and UE dropping at 1.5 m. [3][4][5][6][8][9][10][11][12][13][14][15] • Unlike probabilistic traffic estimation, we have used a real history of hourly traffic and key performance indicators (KPIs) of a live LTE-Advanced network (3GPP Release 12) to forecast the nearby future traffic trends. It is a precise way to show the traffic fluctuation during the day and to exploit peak-off time and busy hours to trigger CSO.…”