1、 I n t e r n a t i o n a l T e l e c o m m u n i c a t i o n U n i o n ITU-T Series L TELECOMMUNICATION STANDARDIZATION SECTOR OF ITU Supplement 8 (12/2014) SERIES L: CONSTRUCTION, INSTALLATION AND PROTECTION OF CABLES AND OTHER ELEMENTS OF OUTSIDE PLANT ITU-T L.1300 Supplement on potential for prim
2、ary energy savings in TLC/ICT centres through free cooling ITU-T L-series Recommendations Supplement 8 L series Supplement 8 (12/2014) i Supplement 8 to ITU-T L-series Recommendations ITU-T L.1300 Supplement on potential for primary energy savings in TLC/ICT centres through free cooling Summary Supp
3、lement 8 to the L series of Recommendations refers to the best practices defined in Recommendation ITU-T L.1300. More precisely, this Supplement first provides an introduction of potential for primary energy savings in TLC/ICT centres through the free cooling solution. Then, a probabilistic model fo
4、r the inlet conditions is defined, and finally an analysis on room temperature and energy consumption is reported. History Edition Recommendation Approval Study Group Unique ID* 1.0 ITU-T L Suppl. 8 2014-12-19 5 11.1002/1000/12436 Keywords Best practice, data centre, energy efficient, information an
5、d communication technology and climate change (ICT a probabilistic model for the inlet conditions; and a room temperature and energy analysis. 2 Definitions This Supplement uses the following term: 2.1 power density: The energy consumption of ICT equipment per rack cabinet of floor area of a server
6、room. 3 Abbreviations and acronyms This Supplement uses the following abbreviations and acronyms: COP Coefficient Of Performance RH Relative Humidity 4 Introduction The growth of modern society is characterized by dynamism in communications, and in recent years the field of telecommunications has ac
7、hieved great importance in the socio-economic development of many countries. Such a trend requires increased power installation in data centres that serve to allocate routers, switches, computers, and many other equipment that have high energy requirements. Such data centres generate heat that must
8、be dissipated. To give an example, the mean value of electrical power density installed in Italian telecommunication plants is about 450 W/m2 and all this electrical power is converted into endogenous power. In 2007, electrical energy consumption for Italian telecommunications was 4000 GWh, which is
9、 1.25% of global electrical energy consumption for that year. Moreover, data shows that the global demand for electrical energy has increased by 0.4% with respect to 2006 data, but that the increase in telecommunications was 1.5%. The most important Italian telecommunication company is the second na
10、tional user for energy consumption; with 0.7% of total national energy demand (the largest user is the national railway transport company). More than half of the energy consumption is used for fixed network and mobile equipment, and approximately 16% of this part is used for cooling. In total, elect
11、rical energy purchased or generated by the group in 2007 amounts to 2.15 TWh, i.e., about 2.3% more than energy used in 2006, and 7.36% more than energy used in 2005. Equipment in data centres have to operate within defined temperature and humidity conditions, because their internal circuits are sen
12、sitive to excessively high or low temperature and humidity. Furthermore, the equipment cannot be switched off because data must be transmitted from centre to centre 24 hours a day, 7 days per week. ETSI standard b-EN 300 019-1-3 establishes the upper and lower limits for temperature and humidity in
13、which equipment must operate. 2 L series Supplement 8 (12/2014) To maintain correct operating conditions, the cooling system must operate continuously, absorbing large amounts of electricity. To contribute to energy saving policies, such as the 20/20/20 objectives established by the European Commiss
14、ion, the use of alternative cooling systems for data centres air conditioning is crucial. Free cooling consists of the direct use of external air to cool the environment. The temperature of external air can be reduced by injecting water spray, if the external humidity is lower than 100%. This is cal
15、led adiabatic free cooling. Free cooling is a rational option for the telecommunication sector, as the temperature tolerated in data centres is usually above outdoor temperature and because of the non-stop operation of the equipment. Knowledge of typical external temperature and relative humidity is
16、 particularly important for a correct design of a free cooling system. Such information is available for various locations, and otherwise a model may be used. A possible model is presented in the next clause. This is particularly useful because of the small number of parameters required for its appl
17、ication to a location. In addition, it provides a general behaviour that should be adopted for compact representation of the suitability of free cooling systems in a location. This Supplement describes a general approach to the energy analysis of free cooling systems and its application to telecommu
18、nication centres located in Italy. In the mid-term it could also be applied to data centres. First, a probabilistic model for ambient temperature and relative humidity is presented. This is the basis for predicting the air inlet condition. Then some theoretical and experimental data obtained from me
19、asurements in the Telecom Italia telecommunication laboratory are presented. These data allow one to relate the inlet air condition to the operating conditions of the equipment. Ultimately, the energy savings that can be achieved through use of free cooling systems in Italy is calculated. Italian lo
20、cations are classified in terms of energy saving due to free cooling and adiabatic free cooling use, and the opportunity of using such a technique is highlighted through a pictorial representation on the Mollier diagram, where the limits of the ETSI standard are also drawn. With respect to the ETSI
21、diagram, here the properties of the probabilistic model of temperature and relative humidity are used for a more compact representation. 5 Probabilistic model for the inlet conditions External air temperature and relative humidity (RH) define the applicability of free cooling. If the temperature is
22、too high, a large air mass flow rate may be necessary to keep the internal temperature within acceptable limits. However, it may be impossible if the external temperature is above the maximum internal temperature. A model that represents the daily temperature (and RH) trend, allows temperature distr
23、ibution to be foreseen for an assigned geographical site on the basis of little data, generally available for most locations. This model is composed of two parts. The first part is an equation that makes it able to describe hourly temperature and RH in the average day of each month. Then a stochasti
24、c model is applied in order to account for expected deviations. 5.1 Temperature model The model is aimed at obtaining an average daily temperature trend for an assigned site and an assigned time step; temperature distribution has been described by simplified mathematical functions that require the l
25、owest number of independent variables. The model has been generated thanks to a large amount of meteorological data referred to over a period of seven years. It has been created on the basis of data available from Turin. First, hourly data have been averaged to obtain 24 mean temperature values for
26、every month (e.g., the value corresponding to the average temperature in January at 1 a.m. is obtained from the mean 317 temperatures recorded at 1 a.m.). The average temperature of a typical day of January is shown in Figure 1. Similar curves are obtained for the other months. L series Supplement 8
27、 (12/2014) 3 Figure 1 Temperature for an average day of January The model has been obtained by considering two different periods: daytime and night-time. The sequence of the hours, during the day, has been modified and fictitious time units have been used: the day starts at sunrise (the first hour i
28、s the hour corresponding to sunrise) and finishes at the hours that come before sunset. The first hour will be different for different geographical areas and for different months. Night-time starts at sunset. Figure 2 shows the shifted temperature curve and the two curves used for modelling daytime
29、and night-time, which are described below. Figure 2 Temperature in a typical day of January and model curves A harmonic function is used to model the daytime temperature: y1 = A + Bsin( x k) (5-1.1) where A is the average between the maximum and minimum temperatures in an examined month; B is half o
30、f the difference between maximum and minimum temperatures; x is the current hour (sunrise x sunset ); k is an empirical coefficient and its value is constant in every month; is the pulsation: = 2 /(H FF) (5-1.2) where H refers to daylight time (from sunrise to sunset); FF is an empirical coefficient
31、, that assumes two different values: in winter FF is equal to 1.6, otherwise it is equal to 1.45. 4 L series Supplement 8 (12/2014) The curve corresponding to night-time is selected to describe the behaviour of a body that is losing heat, so it has been represented with an exponential function: y2 =
32、 b + (y1(xSUNSET ) b) exp(h (x xSUNSET) (5-1.3) where b is the minimum average value of the temperature Tmin, xSUNSET is the actual sunset time, h is an empirical coefficient: its value is constant in every month and is equal to 0.231, x is the current hour (x sunset). Only 4 parameters are required
33、 to obtain an average temperature daily trend for the chosen location: 1) highest monthly temperature (average value); 2) lowest monthly temperature (average value); 3) latitude; 4) longitude. Latitude and longitude are used to determine location sunrise and sunset times. 5.2 Relative humidity model
34、 Daily average distribution of the relative humidity has been obtained in the same way as described in the temperature model. Figure 3 shows the relative humidity for an average day in January and the two curves are used to model the daytime portion and the night-time portion. Figure 3 Relative humi
35、dity for a typical day in January, plus model curves The daytime portion has been represented with a harmonic function: Z1 = C + D (sin( x q) (5-2.1) where C is the mean between the lowest and the highest relative humidity (average value), D is obtained from half of the difference between the lowest
36、 and highest relative humidity (average value), q is an empirical coefficient, and its value is equal to 0.8. Compared to the previous model, this coefficient differs due to the different way of calculating lowest and highest air relative humidity. Because of the difficulty in obtaining such informa
37、tion, data has been taken from the mean relative humidity value, lowest temperature (average value) and corresponding saturated steam pressure. On the basis of the average humidity value, corresponding water vapour ratio, xM, has been calculated using the well-known expression: xM = 0.622 (M pvs (TM
38、)/(patm M pvs(TM) (5-2.2) where M is the average relative humidity, pvs is the saturated steam pressure calculated at reference temperature, patm is the atmospheric pressure, and xM is the water vapour ratio. Minimum and L series Supplement 8 (12/2014) 5 maximum relative humidity have been calculate
39、d with the hypothesis of keeping a constant water vapour ratio: min = (patm xM)/(0.622 pvs (Tmax) + xM pvs (Tmax) (5-2.3) max = (2 mean) min (5-2.4) The night-time portion has been modelled with a logarithmic function: z2 = z1 (xSUNSET) + M ln(x xSUNSET) + G (5-2.5) where G is an empirical coefficie
40、nt equal to 4.5, while M is: (z1(xSUNRISE) z1(xSUNSET) G)/ln(24 H) (5-2.6) Deviation analysis The aim of this part of the model is to introduce deviations with respect to the average hourly temperature and the relative humidity trend. Figure 4 Comparison between calculated and surveyed yearly temper
41、ature distributions The relative frequency distribution of the differences has been approximated as a normal distribution. The standard deviation on difference between calculated mean values and real temperature and relative humidity values has been determined for Turin. The deviation for temperatur
42、e is 3.5C, while for humidity it is 15.5%. The model has been validated with available data for Rome and Palermo. Figure 4 shows a comparison between calculated temperature and modelled temperature for a year. 6 Room temperature The cooling of data centre rooms depends on two different factors: exte
43、rnal conditions and internal conditions. Meteorological models allow the forecasting of external factors in order to classify location on the basis of the operating conditions of equipment. When external air goes into areas, it exchanges heat with the air of the rooms that benefit from the exchange
44、in different ways, according to different room configurations. For this reason, it is vital to know how external parameters need to be modified by referring to equipment suction conditions. A second model has been created to reproduce machine work modalities in Telecom Italia data centre rooms, loca
45、ted in Turin. Room pressure is less than outside pressure because external air has to enter and to cool the environment. The testing laboratory is divided into two contiguous parts (room 1 and room 2) divided by a watertight door. Room 1 is about three times wider than room 2, and has seven rows of
46、racks containing equipment that differ for power. Room 2 is characterized by a large power density but, 6 L series Supplement 8 (12/2014) compared to room 1, air entrance is penalized because of larger friction losses in the inlet air ducts. In room 2, there are four rows of racks and four air ducts
47、 for external air inlet. The layout is not optimum, because in the same passage there is equipment that expels air utilized for circuit cooling and also equipment that takes in air to cool the circuits. In this case, the second equipment will take in high temperature air. Room 2 has been selected fo
48、r the analysis shown hereafter. The results of such an analysis are used as the reference for further considerations. Four operating scenarios of the room are considered: 1) Ideal conditions are obtained by installing equipment of the same type with inlet vents all facing a cold aisle, and outlet ve
49、nts facing a hot aisle. Fresh air enters from the ceiling above the cold aisle and exhaust air is extracted from the ceiling above the hot aisle. Figure 5 shows a schema of this configuration, which theoretically allows one to feed the racks with fresh air at the same temperature as the inlet temperature of the room. This scenario is the best one, but is not always applicable in real data centres, especially if there is equipment with the inlet vent on the same side as the outlet vent. Figur
