Energy saving in a 5G separation architecture under different power model assumptions

Abstract

In this paper, a framework is developed to study the impact of different power model assumptions on energy saving in a 5G separation architecture comprising high power Base Stations (BSs) responsible for coverage, and low power, small cell BSs handling data transmission. Starting with a linear power model function, the achievable energy saving are derived over short timescales by operating small cell BSs in low power states rather than higher power states (termed Low Power State Saving (LPSS) gains) for single and multiple BS scenarios. It is shown how energy saving varies with different power model assumptions over long timescales in accordance with short timescale LPSS. Simulation results show that energy saving in the separation architecture varies across the six power models examined as a function of model-specific significant LPSS state changes. Furthermore, it is shown that if the architecture is based on existing small cell BSs modelled by state-of-the-art (SotA) power models, energy saving will be mainly dependent on sleep state operation. Whereas, if it is based on future BSs modelled by visionary power models, both sleep and idle state operations provide energy saving gains. Moreover, with future BSs, energy saving of up to 42% is achievable when idle state overhead is considered, while a higher saving is possible otherwise.

A Distributed Offloading Market for 5G Heterogeneous Networks

“…The processing rate of the APs are similar to that of the LTE MBSs with µ b = µ i . The MBSs are parametrized with P IN i = 780W and P A i = 1344W , with power data from Auer et al [15], and the APs are parametrized with data from Fisusi et al [20] where P IN i = 1W and P A i = 8.2W . The average system power price of Nordpool for 2013-2016 3 along with a time slot duration of 5 minutes gives C P d = 2.4 · 10 −6 Euro W per time slot .…”

Quantifying the capacity credit of IDC‐based demand response in smart distribution systems

“…The data load tasks received by the IDC are determined by the information requirements of terminal data users. Here, according to the survey, it is considered that the data load task ($$D_{i,t}^0$$) received by IDC from system user i in time period t conforms to truncated Gaussian distribution [24], and the mathematical expression is as follows: $$$$\begin{equation}f(D_{i,t}^0) = \frac{1}{{\sqrt {2\pi } \sigma _{i,t}}}\exp \left[ { - \frac{{{{(D_{i,t}^0 - \bar{D}_{i,t})}}^2}} {{2{{( {\sigma _{i,t}})}}^2}}} \right]\end{equation}$$$$ $$$$\begin{equation}{\rm{ 0}} \le D_{i,t}^0 \le D_{i,t}^{\max }\end{equation}$$$$where $$\bar{D}_{i,t}$$ is the mean value of user i data load demand in time period t ; $$\sigma _{i,t}$$ is the standard deviation of user i data load demand in time period t ; $$D_{i,t}^{\max }$$ is the maximum predicted value of user i data load demand in time period t .…”

“…The data load tasks received by the IDC are determined by the information requirements of terminal data users. Here, according to the survey, it is considered that the data load task (D 0 i,t ) received by IDC from system user i in time period t conforms to truncated Gaussian distribution [24], and the mathematical expression is as follows:…”

Energy Usage Forecasting for LTE: A Network-Wide Traffic Measurements Study

“…The time series of each base station has to be stationary in both variance and mean, in order for a model to be fitted to the measurements and used for forecasting [14]. The heteroscedasticity (variance in variance) was removed with a BoxCox power transformation [14] as shown in (3). The transformation requires that the time series is strictly positive, and since some measurements were zero, a small constant C was added to x it , such that x * it = x it +C.…”

“…Mobile broadband consumes more energy than ever, and for 5G there is a focus on energy usage in addition to increased throughput [1]. Energy models for LTE [2] and 5G [3] base stations consist of fixed and load dependent energy costs. The latter is driven by the number of resource blocks used [4], which is a product of both traffic and spectral efficiency.…”