International Journal of Refrigeration 122 (2021) 171–180

Contents lists available at ScienceDirect

International Journal of Refrigeration

journal homepage: www.elsevier.com/locate/ijrefrig

Free cooling potential for Brazilian data centers based on approach

point methodology
E.A. Amado, P.S. Schneider∗, C.S. Bresolin
Mechanical Engineering Department, Federal University of Rio Grande do Sul, Porto Alegre, RS, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Data centers cooling loads may reach about 40% of their electricity demands. The number of that type of
Received 9 November 2019 facility is continuously growing around worldwide and actions addressing energy saving are welcome to
Revised 14 November 2020
rationalize their operation. Free cooling appears as an effective solution to achieve energy savings, and the
Accepted 16 November 2020
present work proposes a methodology for assessing data center free cooling operation based on the heat
Available online 21 November 2020
exchanger approach point. The methodology can be applied to both airside and waterside economizer
Keywords: modes. The proposed methodology is applied to fourteen Brazilian state capitals that represent all coun-
Data center try climatic conditions. Free cooling environmental air temperature and humidity criteria are based on
Free cooling literature review and on ASHRAE thermal guidelines. Results show that cities such as Curitiba, São Paulo,
Air side economizer Porto Alegre and Brasília can potentially operate in both airside and waterside economizer modes for
Waterside economizer more than 30 0 0 h per year, even considering conservative data center thermal operating limits. Facilities
Approach point
placed on North and Northeast regions can only run under free cooling conditions if operational param-
ASHRAE TC 9.9 thermal guidelines
eters are to be relaxed to consider more flexible operating limits. It is shown that indirect economizers
are potentially more suitable to operate under ASHRAE recommended limits. The proposed methodology
is shaped for evaluating free cooling hours during system conception.
© 2020 Elsevier Ltd and IIR. All rights reserved.

Potentiel de refroidissement naturel basé sur la méthodologie des points

d’approche pour les centres de données brésiliens

Mots-clés: Centre de données; Refroidissement naturel; Économiseur câté air; Économiseur câté eau; Point d’approche; Directives thermiques TC 9.9 de l’ASHRAE

Abbreviations and symbols

Abbreviation
Ap : Approach Point [◦ C]
DC : Data Center DBT : Dry Bulb Temperature [◦ C]
ESD : Electrical Static Discharge DPT : Dew Point Temperature [◦ C]
FC : Free Cooling PUE : Power Usage Effectiveness [adim]
HX : Heat Exchanger RH : Relative Humidity [◦ C]
IT : Information Technology W BT : Wet Bulb Temperature [◦ C]
WCT : Water Cooling Tower Greek letters
Symbols γ : Summation of approach points [◦ C]
(continued on next page)

∗
Corresponding author.
E-mail address: pss@mecanica.ufrgs.br (P.S. Schneider).

https://doi.org/10.1016/j.ijrefrig.2020.11.010
0140-7007/© 2020 Elsevier Ltd and IIR. All rights reserved.
E.A. Amado, P.S. Schneider and C.S. Bresolin
Subscripts
in : Indoor condition
max : Maximum value
min : Minimum value
out : Outdoors conditions
1. Introduction
Data centers (DCs) are dedicated facilities whose primary func-
tion is to house computer servers and offer data services. Reliabil-
ity and availability are paramount for DCs since many companies
rely heavily on a proper and uninterrupted data center operation
all year round (ASHRAE, 2015).
Digital information has experienced massive growth in the last
decade, and data centers are the core infrastructure to support this
trend. Many countries are having rapid expansions of these facili-
ties to meet their demands for internet and cloud computing ser-
vices (Islam et al., 2015). Hence, DC financial and energetic impacts
have drawn a lot of attention worldwide.
The DC market has increased at accelerated rates. According
to Statista (2019), spending on DC systems worldwide has grown
from US$ 140 billion in 2012 to US$ 202 billion in 2018, and it
is expected to reach US$ 210 billion in 2019. Within the DC mar-
ket, stands out the collocation sector. A collocation is a special data
center facility that rents infrastructure for customers to host their
servers and other computing hardware. This sector has grown even
faster than the average DC market. The collocation market should
reach a revenue of US$ 48 billion in 2021.
The collocation market spread has contributed to segment DC
facilities worldwide (Shende, 2015). Partly driven by this, the Latin
America DC market generated a turnover of US$ 2.87 billion in
2017, and Brazil was responsible for 47.6% of this amount, which is
expected to reach US$ 4.37 billion until 2021 (Dino, 2018). Some
local environmental characteristics are desired to compete at a
world scale, such as cheap electricity cost, cold climate and low-
cost real estate. However, these characteristics are worthless unless
applied to high-level energy performance systems.
With respect to the energetic consequences of data centers, it is
fair to say that their energy consumption has increased at alarming
rates all around the world. Between 20 0 0 and 20 05 this value has
doubled, from 2005 to 2010 it increased another 56%, even though
the world was facing an economic crisis (Zhang et al., 2014). DCs
were responsible for about 1.5% of worldwide electricity consump-
tion in 2011, and an annual increase of 15–20% is expected in the
near future (Amoabeng and Choi, 2016). Furthermore, there is also
a major concern regarding data centers CO2 emissions. According
to Avgerinou et al. (2017), it is estimated that 2% of the global
CO2 is a result of energy usage in DCs. This value is expected to
increase at a 6% annual rate, and account for 12% of worldwide
emissions by 2020. In this context, DC energy performance has also
become an important factor.
As a result of all these factors, the trend in the IT and DC indus-
tries is to seek a high-level of reliability and availability followed
by a reduction in energy consumption. The main responsible for
energy consumption in data centers are the servers and the re-
frigeration system, each of them accounting for about 40% of the
entire facility consumption (Rong et al., 2016).
From this perspective, Power Usage Effectiveness (PUE) index is
acknowledged as the most accepted metric to describe DC energy
efficiency (Yuventi and Mehdizadeh, 2013). PUE is defined as the
ratio of the total amount of energy used by an entire data cen-
ter facility to the energy demanded only by the computing equip-
ment. PUE = 1 represents the hypothetical situation that the en-
172
International Journal of Refrigeration 122 (2021) 171–180
ergy consumption from all non-computing devices of a data center
facility; such as lighting, power supply, communication and cool-
ing systems; would be zero. An average PUE is just below 2.0.
High-level energy performance in a DC is equivalent to reach low
PUE values, close to 1.2. Reducing the energy consumption of the
cooling system is the key factor to achieve a low PUE value and
improve a data center facility in both energetic and financial as-
pects. Santos et al. (2019) reviewed and discussed the most com-
mon metrics used to address DC energy performance. The authors
proposed the Energy Usage Effectiveness Designs – EUED as a new
index, which tried to correlate the local weather to the DC energy
consumption on a year base.
The use of free cooling strategies is, nowadays, one of the most
used cooling efficiency techniques for data centers (Oró et al.,
2015). According to The Green Grid (2012), it is estimated that the
use of free cooling results in saving an average of 20% of money,
energy and carbon emissions caused by cooling when compared to
DCs that do not use this technique. The following section describes
free cooling in detail.
Acknowledging the urge for energy savings in data centers, free
cooling potential studies have an important role in developing this
technique and, consequently, in the increase of energy efficiency of
these facilities.
Methodologies have been proposed to assess the amount of
time that a DC can run compressor-free. Most of them re-
lies on local weather data and specific system simulation. Local
weather data and psychrometric operational criteria can suffice
for designing simple systems based on direct airside economizers
(Siriwardana et al., 2013; The Green Grid; The Green Grid, 2012),
whether more complete thermal system simulation may be re-
quired when complex systems are to be conceived (Bi et al., 2017;
Gözcü et al., 2017; Lee and Chen, 2013a).
This work is motivated by the rising trend for energy efficiency
in data centers and aimed to develop criteria to estimate the DC
free cooling potential of Brazilian reference cities applied to airside
and waterside economizers. The proposed criteria were meant to
be general and flexible to be suitable for different DCs and regions
worldwide.
2. Free cooling
Free cooling is a mode of operation defined as the use of a
natural cold source (air or water) to cool an indoor environment
(Oró et al., 2015). When the cold source temperature is below
compared to the indoor operating temperature, it is possible to
cool the data center without a compressor based refrigeration cycle
(Amoabeng and Choi, 2016). This mode of operation is commonly
known as economizer cycle (Zhang et al., 2014).
Air conditioning compressors are the highest energy demanding
device of a cooling system. Free cooling systems can be arranged
in parallel or in series with the conventional air conditioning sys-
tem. Free cooling can be used either on full economizer mode, i.e.,
without the compressor; or in partial economizer mode as an aux-
iliary step to reduce the cooling load. Even so, free cooling still
relies on auxiliary power to drive system pumps and fans (Lee and
Chen, 2013b), significant energy savings can be achieved whenever
compressors are turned off.
The economizer modes can be divided into airside and water-
side categories. The former rejects the DC heat to the air, and the
latter uses a water loop to transfer the DC heat outward.
Airside economizers cool the DC indoor environment using the
outdoor air directly or indirectly. In the direct category, filtered air
is supplied to the DC as in the outside condition by fans, Fig. 1a.
The outdoor air is drawn to the indoor environment without any
air conditioning or humidity control. The advantage of this mode is
to be a less complex system that does not require a pump, cooling
E.A. Amado, P.S. Schneider and C.S. Bresolin
Fig. 1. Air cooled economizer modes.
Fig. 2. A schematic of a direct water cooled system. Adapted from Clidaras et al.
(2009).
towers or heat exchangers. The disadvantages are the risk of dust
contamination and no humidity or temperature control can reduce
DC reliability.
Indirect airside economizer uses an intermediate air-air heat
exchanger, between the indoor recirculating airflow and outdoor
airflow, Fig. 1b. The additional heat transfer step causes a reduc-
tion in the effectiveness of the whole system, usually allowing for
less free cooling hours when compared to direct airside economiz-
ers. On the other hand, it brings the advantages of preserving the
indoor air quality, saving the need for additional air filtering and
being less dependent on the outdoor humidity (ASHRAE, 2016b;
Zhang et al., 2014). Besides, it can even provide more free cool-
ing hours than a direct type for DCs in regions with constant high
humidity levels.
The performance of direct type economizers depends only on
the regional climate and indoor operating conditions, while the in-
direct type performance relies also on parameters related to the
heat exchanger. Thus, direct airside free cooling potential assess-
ments are more general and straightforward, making this kind of
assessment a better starting point for DCs free cooling potential
studies. There is an optimum supply air temperature for airside
economizer. As the supply temperature increases, the data center
free cooling hours increases as well, but after an optimum point,
the fan energy consumption exceeds the compressorless hours en-
ergy economy. The optimum supply temperatures are in the 18 o C–
23 o C range, depending on operation conditions (Ham et al., 2015).
Waterside economizer can be classified as direct water cooled
system, air-cooled system and cooling tower system (Zhang et al.,
2014). In direct water cooled system, natural cold water is used
to cool the data center indoor air without any other steps of heat
transfer, Fig. 2. This application is limited by the dependency either
on purchasing water or on natural cold water availability, meaning
that the DC must be located close to water sources. Nevertheless,
direct water cooled economizer is an efficient operating mode, able
to maintain the facility temperature close to the average environ-
ment temperature for 24 h a day.
This system seems promising in Brazil, due to the country large
coastline, and its key cities near the ocean. However, cooling wa-
ter drawn from natural sources may impact the local ecosystem
due to the increased temperature of the discharged water. Be-
sides, water purchased from utilities for this purpose became very
173
International Journal of Refrigeration 122 (2021) 171–180
Fig. 3. Schematic of an air cooled waterside economizer system. Adapted from
Zhang et al. (2014).
expensive because of increased water supply and disposal costs
(ASHRAE, 2016a).
Air cooled system is based on sensible heat exchangers to cool
down the circulating water when the outdoor air dry-bulb temper-
ature is favorable, Fig. 3. It is a simpler solution than the direct
water cooled one and avoids the environmental impact caused by
natural water consumption. On the other hand, its effectiveness is
lower (ASHRAE, 2016a), promoting fewer free cooling hours.
The cooling tower system uses an evaporative tower to reduce
the circulating water temperature, Fig. 3. Cooling tower type econ-
omizers are the most widely used in the waterside category be-
cause they overcome most of the problems encountered by other
types. The water consumption rate of a cooling tower is only
about 5% of direct water system, making it more viable to op-
erate with purchased water. Furthermore, the amount of heated
water discharged is very small, so the environmental harm is
greatly reduced if compared with the direct water cooled system
(ASHRAE, 2016a).
2.1. Free cooling potential
Free cooling is considered an effective solution for reducing the
energy consumption of cooling systems in the data center when its
operation is viable for a high number of hours in a year. Thus, the
free cooling potential is defined as the number of hours in a year
that the DC can be cooled without compressor work (Malkamäki
and Ovaska, 2012; Oró et al., 2015; Siriwardana et al., 2013). The
free cooling potential varies significantly depending on local cli-
mate (Amoabeng and Choi, 2016).
In this context, free cooling potential assessments are studies
that evaluate the potential of a given city, region or country to cool
environments without mechanical cooling.
DC free cooling potential assessments commonly focus on gen-
eral direct airside economizers. Siriwardana et al. (2013) investi-
gated the potential of airside economizers in DCs housed in 20
Australian cities, using weather data from 20 0 0 to 2011. The cities
had their hourly dry-bulb temperature being compared with 15 ◦ C,
which the set authors as the maximum allowed temperature for
the DC supplied air intake. The results showed that Melbourne pre-
sented the highest free cooling potential from the cities assessed
allowing for direct airside economizers to be used for more than
50 0 0 h per year.
A thermo-economic analysis for three types of economizers
done in ten Iranian cities (Deymi-Dashtebayaz and Namanlo, 2019)
concluded that the PUE can be improved up to 12% using com-
bined air or water economizer modes for most cities.
Similarly, The Green Grid (2009) published the “Airside Free
Cooling Maps”, evaluating the potential of direct airside economiz-
ers for DCs in Europe, North America and Japan. These maps were
updated in 2012 (The Green Grid, 2012) to take into account newer
and more relaxed operating ranges limits. The maps were made by
comparing bin data collected from the WeatherBank with maxi-
mum limits of dry-bulb and dew-point temperatures for data cen-
E.A. Amado, P.S. Schneider and C.S. Bresolin
Table 1
Recommended and Allowable ranges for DC operation.
Adapted from ASHRAE (2015).
Range DBT (◦ C) DPT (◦ C) RH (%)
Recommended 18 to 27 −9 to 15 60
Allowable A1 15 to 32 −12 to 17 8 to
Allowable A2 10 to 35 −12 to 21 8 to
Allowable A3 05 to 40 −12 to 24 8 to
Allowable A4 05 to 45 −12 to 24 8 to
ters. The Green Grid Airside Free Cooling Maps have become an
important reference for DC operators and researchers in this field,
as it allows for a first estimate of the free cooling performance in
different regions.
Beaty et al. (2019) presented a waterside free cooling potential
assessment for several North American cities, by considering dif-
ferent values for the supply air temperature, ranging from 14 ◦ C
to 24 ◦ C. The approach between the outdoor wet-bulb temperature
and the chilled water supply was assumed as 3.9 ◦ C and the tem-
perature difference between the chilled water supply and the sup-
ply air temperature was assumed as 5.6 ◦ C. The author evaluated
the hour percentage in a year in which a DC under those con-
ditions could operate under waterside free cooling, based on the
local weather data. Denver displayed the highest potential among
the assessed cities, with 53% of free cooling potential for the worst
scenario.
In Brazil, works from Driemeyer (2016) and
Belizário (2018) modeled and assessed free cooling systems
as case studies, and Santos et al. (2019) applied their proposed
EUED index to select Brazilian cities. Nevertheless, there still is a
lack of general free cooling potential assessment for the country.
3. Free cooling assessment
To account for free cooling hours, an internal operation condi-
tion is needed for temperature and humidity, and compare it with
the outside weather condition. If the system has intermediate heat
exchangers, it is necessary to consider the temperature gap be-
tween the hot and cold flows. Two ways to compute the free cool-
ing potential are proposed, one for direct air cooling, and other to
waterside cooling tower. Also, a generalization for other indirect
modes is proposed.
3.1. DC thermal operating conditions
The latest ASHRAE thermal guidelines (ASHRAE, 2015) displays
the required conditions for reliable DC operation. Ranges of dry-
bulb temperature (DBT ), relative humidity (RH) and dew-point
temperature (DP T ) were featured for different classes of IT equip-
ment in this guideline as shown in Table 1 and Fig. 4.
The Recommended range is defined as a reference, as it is
the safest one and the most accepted by DC operators (Lee and
Chen, 2013b). The other four are called Allowable ranges, with dif-
ferent limits according to the IT equipment class. An equipment
class is a definition for IT manufacturers to design their products.
When it was first created, the Recommended range was in-
tended to be the most reliable, acceptable and reasonable power-
efficient operation zone, but it was never meant to be the abso-
lute limits of operation. As a result of the growing need for en-
ergy efficiency, and consequently, the interest to use economizers
as much as possible during a year, the Allowable ranges were pro-
posed, reinforcing the idea that it is acceptable to operate outside
the Recommended range for short periods of time without affect-
ing the overall reliability of the IT equipment. It is worth noticing
that new equipment allowed for a more relaxed operation, toward
International Journal of Refrigeration 122 (2021) 171–180
A3 and A4 classes, while the older ones are more suited to operate
within the A1 and A2 classes.
3.2. Free cooling assessment for direct airside
80
80 The Recommended and Allowable envelopes are shown in
85 Fig. 4 are intended to describe operating limits of DCs while using
90 mechanical cooling. As free cooling does not depend on compres-
sor some of the envelope boundaries are not required. An in-depth
survey in the literature was carried out in this subsection to under-
stand which limits could be discarded while establishing the direct
airside free cooling criteria.
Dry-bulb temperature The upper DBT limit is the most impor-
tant parameter for IT equipment operation and should be taken
into account in this evaluation, as most of the failure mechanisms
of electronic components are intensified by increasing their opera-
tional temperature (ASHRAE, 2016b). The maximum operating tem-
perature limit that an IT equipment should reach the chip level is
about 85 ◦ C (Amoabeng and Choi, 2016; Marcinichen et al., 2012).
The heat removal in a data center must guarantee that this limit is
not surpassed. The lower boundary of DBT suggested ranges are in-
tended to prevent a very inefficient energetic operation under me-
chanical cooling. This boundary was discarded in the present work
as it does not impact the free cooling potential assessment.
Lower humidity boundaries The RH and DP T parameters are
metrics to measure the relative and absolute humidity of the DC
environment, respectively. Their lower limits were introduced by
ASHRAE TC 9.9 (ASHRAE, 2015) to avoid the risk of electronic
discharges (ESD) in the IT equipment. The guideline reduced the
lower RH limit from 25% to 8% after studies (Gao et al., 2015; Hy-
deman and Swenson, 2010; Wan et al., 2013) had verified that this
reduction would not increase at high rates the ESD probability.
The assessment of Brazilian meteorological data (LABEE, 2018)
showed little occurrences of RH and DP T values below 8% and
−9 ◦ C respectively. Thus, it was decided to discard the lower RH
and DP T limits while assessing the free cooling potential.
Higher humidity boundaries High humidity levels can cause wa-
ter vapor to condensate on IT equipment, putting at risk their reli-
ability (Oró et al., 2015). However, the RH parameter is difficult to
be controlled inside a DC facility as it depends on both the abso-
lute humidity content and the dry-bulb temperature. Nevertheless,
its control can be performed by partially mixing exhaust air with
inlet air. DP T is a more stable parameter, as it is linked to the air
absolute humidity, which does not change significantly throughout
the DC environment (Hydeman and Swenson, 2010). Therefore, the
upper DP T limit was taken into account while the upper RH limit
was discarded in the present assessment.
Based on these assumptions, only two parameters among those
presented in the operating ranges suggested by ASHRAE were con-
sidered to evaluate the free cooling potential for direct airside:
upper dry-bulb temperature limit (DBTin,max ) and upper dew-point
temperature limit (DP Tin,max ), defined by the upper right corner of
envelopes in Fig. 4. Thus, the free cooling is considered possible
for the hourly step accounted by the double condition imposed in
Eq. (1),
DBTout ≤ DBTin,max
∴ Direct Airside Free cooling is possible. (1)
DP Tout ≤ DP Tin,max
for the hourly outdoor dry-bulb DBTout and dew-point DP Tout
temperatures in respect to the desired limits for DBTin,max and
DP Tin,max . It was assumed that the air temperature remains un-
changed along with its intake through the cold face of IT equip-
ment.
A graphical representation was made to illustrate the criteria
established on a psychrometric chart. Fig. 5 presents the free cool-
174
E.A. Amado, P.S. Schneider and C.S. Bresolin International Journal of Refrigeration 122 (2021) 171–180

Fig. 4. Recommended and Allowable operating envelopes on a psychrometric chart.

Fig. 5. Free cooling zone considering the Recommended range limits on a psychrometric chart.

175
E.A. Amado, P.S. Schneider and C.S. Bresolin
Fig. 6. Free cooling zones considering Allowable ranges limits.
ing zone considering the free cooling criteria applied to the Rec-
ommended range limits for DBTin,max and DP Tin,max .
The green color represents the zone where free cooling is
achievable (Zone I), the red color means that the free cooling is
not possible due to the high DBT level (Zone II), the red color rep-
resents that free cooling is not possible due to the high DP T level
(Zone III), and the orange zone is where free cooling is not practi-
cal due to both: DBT and DP T parameters (Zone IV).
Free cooling zones for Allowable ranges limits A1, A2, A3 and
A4 were similarly plotted as shown in Fig. 6.
3.3. Free cooling methodology for waterside cooling tower
Differently from direct airside economizer mode, indirect
modes use heat exchangers. Heat exchanger can be modeled as
a resistance device. To offset the resistance, the cold source tem-
perature should be lower than the DC internal required tempera-
ture, reducing the free-cooling available hours. That additional heat
176
International Journal of Refrigeration 122 (2021) 171–180
transfer resistance is taken into account by the approach point,
which is the minimum temperature difference between hot and
cold fluid flows. The use of the approach point to refine the free-
cooling potential assessment is shown as follows applied to a wa-
terside economizer with a cooling tower to be later expanded to
other indirect modes.
To calculate the W BTout,max , the approach point from the heat
exchanger and cooling tower ApWCT are discounted from the max-
imum internal dry-bulb temperature DBTin,max ,
W BTout,max = DBTin,max − ApHX − ApWCT (2)
the approach values can be evaluated through an energy balance
in the heat exchanger, using the inlet and outlet temperature dif-
ferences and effectiveness. The ApWCT is defined as the difference
between the cooling tower outlet cold water and the outdoor wet-
bulb temperature W BTout . The ApHX is defined in the cold side of
a counter-current heat exchanger. Eq. (2) is an expression to find
the maximum outdoor wet-bulb temperature which guarantees
E.A. Amado, P.S. Schneider and C.S. Bresolin
Table 2
Free cooling potential criteria summary.
Category Type Criteria
Airside Direct DBTout ≤ DBTin,max
DP Tout ≤ DP Tin,max
Indirect DBTout ≤ DBTin,max − ApHX
Waterside Direct water cooled DBTout ≤ DBTin,max − ApHX
Air cooled DBTout ≤ DBTin,max − ApHX,out − ApHX,in
Cooling tower W BTout ≤ DBTin,max
b
− ApHX − ApWCT
DBTin,max as the indoor supply air temperature considering that
all DC restrictive parameters had been set. Therefore, free cooling
is considered possible when the outdoor wet-bulb temperature is
less than or equal to this maximum calculated value (Eq. (3)).
W BTout ≤ W BTout,max ∴ Free cooling is possible.
The heat exchangers were taken as adiabatic, so the cooling
tower rejects the total dissipated heat from the IT equipment. Both
indoor air and water loops were modeled as systems with no need
for fluid make-up. There are no other heat sources along the air
path from intake to the IT cold face. The hot and cold aisles are
perfectly separated. The DBTin,max is the main parameter to be
monitored whenever indirect waterside economizers are used. The
ASHRAE thermal guidelines limits presented in Table 1 were used
as a reference to evaluate this parameter.
3.4. Graphical procedure
A graphical procedure was made to define the free cooling zone
on a psychrometric chart. The customized psychrometric chart pre-
sented in Fig. 7(a) displays the “γ isolines”. With γ being defined
as the sum of approaches from the cooling tower and heat ex-
changer as shown in Eq. (4).
γ ≡ ApHX + ApWCT
W BTout,max = DBTin,max − γ
For values of DBTin,max and γ , one can trace a constant DBT line
from the DBTin,max until it reaches the set γ value and then mark
this point. The W BT of this point is the W BTout,max . Thus, the free
cooling zone can be defined as the entire zone below a constant
W BT line which has that point as shown in Fig. 7(b). This pro-
cedure is useful to create free cooling zones and compare them
with weather data graphically. Fig. 7(c) and (d) presents applica-
tions of this graphical procedure to create different free cooling
zones, relying on different system parameters, but considering the
same weather data occurrences.
3.5. Generalization for other indirect types
Similar criteria to the ones developed for the cooling tower
economizer can be proposed to the indirect air, direct water and
air cooled water types since they are all indirect. Table 2 brings
the criteria for all the five main types in the airside and waterside
categories.
4. Results
To explore the proposed methodology, the free cooling poten-
tial for fourteen capital cities were selected based on the avail-
able Test Reference Year (TRY) weather data (LABEE, 2018). The se-
lected cities are the major economic centers in their regions. Al-
most all cities are in the ASHRAE Climate Zone 1 in Brazil. All
country regions were covered (South, Southeast, Midwest, North-
east and North). Table 3 presents the selected cities by Region and
International Journal of Refrigeration 122 (2021) 171–180
ASHRAE Climate Zone. Dry-bulb and dew-point temperatures were
employed to find the number of hours that each city could poten-
tially operate on free cooling along a year period (8760 h).
4.1. Airside direct results
The direct type economizers performance depends only on
the regional climate and indoor operating conditions. Thus, direct
airside free cooling potential assessments are more general and
straightforward, making a better starting point for DCs free cool-
ing potential studies.
Table 4 summarizes the airside free cooling potential assess-
ment results for the 14 Brazilian capitals listed in Table 3. Limits
from all ranges (Recommended and Allowable A1, A2, A3, A4) were
considered. Reddish tones represent lower free cooling potential,
(3) while bluish tones represent higher potential.
Curitiba displayed the best performance among all cities, being
capable to operate inside Zone I (Fig. 5) for 5575 h considering the
Recommended range limits. Capital cities such as São Paulo, Porto
Alegre and Brasília presented free cooling potential for more than
30 0 0 h considering the Recommended range. Most of the North
and Northeast cities displayed a negligible number of free cooling
hours for both Recommended and A1 range limits.
Overall better results were achieved when considering more
flexible ranges (A2, A3 and A4). Curitiba, São Paulo and Brasília can
operate in free cooling considering A2 range for almost the entire
year (80 0 0+ hours), and all cities aside from Fortaleza are able to
operate inside Zone I for 7800+ hours in a year when A3 or A4
ranges were considered.
By looking exclusively at the free cooling hours, one cannot find
whether the excess of DBT or the excess of DP T were responsible
to prevent more free cooling hours. An expanded assessment based
only on the DP T criterion was performed on the 14 capital cities to
investigate whether the DBT criterion was really required (Table 5).
(4) Results presented in Table 5 corroborate with the hypothesis
that DP T is indeed the critical parameter to the free cooling po-
(5) tential assessment in Brazil. The DBT criterion was only relevant
while assessing the Recommended range limits and for a few cities
such as Brasília and São Paulo, which displayed a variation in the
free cooling hours of 9.94% and 3.12%, respectively. Maceió had a
62.50% variation in its free cooling hours, but the absolute num-
ber is negligible since it raised from 8 to 13 h. All cities displayed
slightly increases or even no changes at all while considering the
limits of the Allowable ranges, which indicates that the direct air-
side free cooling assessment could be performed based only on the
DP T criterion without causing major changes in the free cooling
potential hours of the 14 Brazilian cities.
4.2. Cooling tower waterside results
The free cooling potential results for 14 Brazilian capitals are
presented in Table 6 for an average scenario considering γ = 10 ◦ C.
As reference values, Beaty et al. (2019) and Belizário (2018) used,
respectively, an equivalent value to γ = 9, 5 ◦ C and γ = 8 ◦ C in
their work. All of the ASHRAE thermal guidelines operating range
limits (Table 1) were taken into consideration to set the DBTin,max
values in this assessment.
Curitiba displayed the best performance among all cities, as
6301 h per year were found of free cooling potential considering
the Recommended range limit. Capital cities such as São Paulo,
Porto Alegre and Brasília came right after, with more than 30 0 0 h
in a year on the same range limit.
Most of the North and Northeast regions of the country dis-
played a negligible potential to free cooling considering Rec-
ommended and A1 limit, with exception of Salvador and Ma-
177
E.A. Amado, P.S. Schneider and C.S. Bresolin International Journal of Refrigeration 122 (2021) 171–180

Fig. 7. Graphical representation of the waterside free cooling potential criteria.

Table 3
Capital cities assessed by country region. Weather data obtained from LABEE (2018).

Region ASHRAE climate zone City (state)

South 2 Porto Alegre (RS), Florianópolis (SC), Curitiba (PR)

Southeast 2 São Paulo (SP)
1 Rio de Janeiro (RJ), Vitória (ES)
Midwest 2 Brasília (DF)
Northeast 1 Salvador (BA), Maceió (AL), Recife (PE),
Natal (RN), Fortaleza (CE), São Luís (MA)
North 1 Belém (PA)

ceió which achieved more than 40 0 0 h in the A1 limit. Over-
all better results were found when considering more flexible
limits.
All the 14 assessed cities displayed at least 5900 free cool-
ing hours with the waterside economizer mode when consid-
ering the A2 range limits, and that number can reach at least
8300 h (almost the entire year) if Fortaleza, São Luís and Belém
were excluded from the assessment. Free cooling is possible for
the entire year for all the 14 cities when A3 or A4 limits were
considered.
This assessment was performed over an average of γ = 10 ◦ C.
Results are very sensitive to changes on γ , and W BT cumulative
frequency distributions can aid to assess the free cooling potential
of a given city considering variations of γ .

178
E.A. Amado, P.S. Schneider and C.S. Bresolin International Journal of Refrigeration 122 (2021) 171–180

Table 4
Free cooling potential in 14 Brazilian capitals.

Free Cooling (FC) hours FC hours FC hours FC hours FC hours

considering considering A1 considering A2 considering A3 considering
City State Latitude Recommended range range range range A4 range

Porto Alegre RS 30◦ 02’S 3783 5309 7927 8735 8735

Florianópolis SC 27◦ 36’S 1925 3760 7004 8470 8470
Curitiba PR 25◦ 26’S 5575 7299 8726 8760 8760
São Paulo SP 23◦ 33’S 4259 6418 8665 8760 8760
Rio de Janeiro RJ 22◦ 54’S 255 1134 5481 8555 8555
Vitória ES 20◦ 19’S 447 1417 5304 8494 8494
Brasília DF 15◦ 47’S 3028 5079 8745 8760 8760
Salvador BA 12◦ 58’S 17 261 5061 8729 8729
Maceió AL 9◦ 40’S 8 130 5795 8759 8759
Recife PE 8◦ 03’S 0 4 2521 8694 8694
Natal RN 5◦ 48’S 0 6 3048 8697 8697
Fortaleza CE 3◦ 43’S 0 1 221 6657 6657
São Luís MA 2◦ 32’S 0 1 168 8141 8141
Belém PA 1◦ 27’S 0 2 387 7888 7888

Table 5
Free cooling hours while assessing only the dew-point temperature criterion and its percentage growth in relation to the assessment
considering the two criteria.

Free Cooling (FC) hours FC hours FC hours FC hours FC hours

considering considering considering considering considering
City State Latitude Recommended range A1 range A2 range A3 range A4 range

Porto Alegre RS 30◦ 02’S 3836 (+1.40%) 5323 7939 8735 (=) 8735 (=)
(+0.26%) (+0.15%)
◦
Florianópolis SC 27 36’S 1929 (+0.21%) 3760 (=) 7004 (=) 8470 (=) 8470 (=)
Curitiba PR 25◦ 26’S 5615 (+0.72%) 7299 (=) 8726 (=) 8760 (=) 8760 (=)
São Paulo SP 23◦ 33’S 4392 (+3.12%) 6425 8665 (=) 8760 (=) 8760 (=)
(+0.11%)
Rio de Janeiro RJ 22◦ 54’S 259 (+1.57%) 1134 (=) 5487 8555 (=) 8555 (=)
(+0.11%)
Vitória ES 20◦ 19’S 447 (=) 1417 (=) 5304 (=) 8494 (=) 8494 (=)
Brasília DF 15◦ 47’S 3329 (+9.94%) 5086 8745 (=) 8760 (=) 8760 (=)
(+0.14%)
Salvador BA 12◦ 58’S 17 (=) 261 (=) 5061 (=) 8729 (=) 8729 (=)
Maceió AL 9◦ 40’S 13 (+62.50%) 130 (=) 5795 (=) 8759 (=) 8759 (=)
Recife PE 8◦ 13’S 0 (=) 4 (=) 2521 (=) 8694 (=) 8694 (=)
Natal RN 5◦ 48’S 0 (=) 6 (=) 3048 (=) 8697 (=) 8697 (=)
Fortaleza CE 3◦ 43’S 0 (=) 1 (=) 221 (=) 6657 (=) 6657 (=)
São Luís MA 2◦ 32’S 0 (=) 1 (=) 168 (=) 8141 (=) 8141 (=)
Belém PA 1◦ 27’S 0 (=) 2 (=) 387 (=) 7888 (=) 7888 (=)

Table 6
Free cooling potential hours in 14 Brazilian capitals considering the average scenario (γ = 10 ◦ C).

Free Cooling (FC) hours FC hours FC hours FC hours FC hours

considering considering considering considering considering
City State Latitude Recommended range A1 range A2 range A3 range A4 range

Porto Alegre RS 30◦ 02’S 4291 7700 8613 8760 8760

Florianópolis SC 27◦ 36’S 2600 7077 8389 8752 8760
Curitiba PR 25◦ 26’S 6301 8664 8760 8760 8760
São Paulo SP 23◦ 33’S 5117 8569 8754 8760 8760
Rio de Janeiro RJ 22◦ 54’S 578 5238 8242 8760 8760
Vitória ES 20◦ 19’S 832 5215 8261 8760 8760
Brasília DF 15◦ 47’S 3032 8660 8760 8760 8760
Salvador BA 12◦ 58’S 79 4113 8553 8760 8760
Maceió AL 9◦ 40’S 34 5051 8700 8760 8760
Recife PE 8◦ 03’S 0 1726 8613 8760 8760
Natal RN 5◦ 48’S 0 2302 8380 8760 8760
Fortaleza CE 3◦ 43’S 0 309 5939 8760 8760
São Luís MA 2◦ 32’S 0 132 6922 8760 8760
Belém PA 1◦ 27’S 0 609 6875 8760 8760

5. Conclusion compressor-free. The heat exchanger resistance was used as a pa-

The present work presented a methodology for free-cooling
potential assessment able to incorporate system limitations. The
methodology uses the heat exchanger approach point applied
to criteria from ASHRAE thermal guidelines with local weather
data to estimate the number of hours that a DC can be cooled
rameter within the approach point method, which allowed it to be
of general to use with airside and waterside categories of econo-
mizers.
The maximum dry-bulb (DBTin,max ) and dew-point (DP Tin,max )
temperatures of the supplied intake air were considered as cri-
teria to estimate the direct airside free cooling potential. Results

179
E.A. Amado, P.S. Schneider and C.S. Bresolin
showed that Curitiba displayed the highest free cooling potential
in this category, achieving 5575 economizer hours in a year, even
considering the most conservative operating limit. Cities such as
Porto Alegre, São Paulo and Brasília also showed the potential to
free cooling, displaying more than 30 0 0 economizer hours in those
limits. Cities from North and Northeast did not show the potential
to free cooling considering conservative operating limits. However,
when considering more flexible ranges such as Allowable A3 and
A4 limits more than 70 0 0 economizer hours were achieved. A fur-
ther analysis detailing these results showed that the DP T parame-
ter was the critical criterion in this assessment.
The waterside free cooling potential assessment was made con-
sidering a cooling tower economizer type. The criteria to free cool-
ing proposed in this category considered the maximum supply
air temperature (DBTin,max ) and the system approaches (ApCT and
ApHX ) as parameters to compare with the outdoor wet-bulb tem-
perature (W BTout ). A graphical representation was developed to il-
lustrate the criteria dependence with the sum of the system ap-
proaches (γ ) on a psychrometric chart.
The waterside free cooling criteria were evaluated and dis-
cussed considering an average scenario of γ = 10 ◦ C based on lit-
erature values. The results behaved similarly to the airside free
cooling potential assessment. Curitiba displayed the highest free
cooling potential in this category, achieving 6301 economizer
hours in a year considering the most conservative operating lim-
its. Porto Alegre, São Paulo and Brasília also showed potential
for free cooling displaying more than 30 0 0 economizer hours in
that limit. North and Northeast cities of the country only pre-
sented the potential to free cooling considering Allowable A2 limits
onwards.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgments
Amado acknowledges Coordenação de Aperfeiçoamento de Pes-
soal de Nível Superior (CAPES) for his MSc grant; Smith Schneider
acknowledges Conselho Nacional de Desenvolvimento Científico e
Tecnológico for his research grant (CNPq-PQ 301619/2019-0).
References
Amoabeng, K.O., Choi, J.M., 2016. Review on cooling system energy consumption
in internet data centers. Int. J. Air-Cond. Refrig. 24 (04), 1630 0 08. doi:10.1142/
S201013251630 0 081.
ASHRAE, 2015. ASHRAE Thermal Guidelines for Data Processing Environments,
fourth American Society of Heating, Refrigerating and Air-Conditioning Engi-
neers, Inc, Atlanta, GA.
ASHRAE, 2016a. ASHRAE Handbook: HAVC Systems and Equipment, SI Edition.
American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc,
Atlanta, GA.
ASHRAE, 2016b. Data Center Power Equipment Thermal Guidelines and Best Prac-
tices. Technical Report. American Society of Heating, Refrigerating and Air-Con-
ditioning Engineers, Inc.
Avgerinou, M., Bertoldi, P., Castellazzi, L., 2017. Trends in data centre energy con-
sumption under the european code of conduct for data centre energy efficiency.
Energies 10 (10), 1470. doi:10.3390/en10101470.
Beaty, D.L., Quirk, D.P., Morrison, F.T., 2019. Designing data center waterside econo-
mizers. ASHRAE J. 57–61.
Belizário, A.C., 2018. Avaliação Energética e Financeira para Utilização de Sistemas
de ar Condicionado Acionados por Energia Solar em Ambientes de Missão
Crítica Para Diferentes Regiões Climáticas. Universidade de São Paulo, São Paulo
Ph.D. thesis.
180
International Journal of Refrigeration 122 (2021) 171–180
Bi, Y., Wang, Y., Ma, X., Zhao, X., 2017. Investigation on the energy saving potential
of using a novel dew point cooling system in data centres. Energies 10, 1732.
doi:10.3390/en10111732.
Clidaras, J., Stiver, D.W., Hamburgen, W., 2009. Water-Based Data Center.
Patent US7525207B2; 2009. https://patentimages.storage.googleapis.com/b6/d0/
d6/d1f4357032fea5/US7525207.pdf.
Deymi-Dashtebayaz, M., Namanlo, S.V., 2019. Potentiometric and economic analysis
of using air and water-side economizers for data center cooling based on vari-
ous weather conditions. Int. J. Refrig. 99, 213–225. doi:10.1016/j.ijrefrig.2019.01.
011.
Dino. https://www.terra.com.br/noticias/dino/com- receita- de- us- 28- bilhoes- na-
america- latina- mercado- de- data- centers- demanda- investimento- brasileiro,
e86e110777ea555792ccb818a87d4421eedextcb.html.
Driemeyer, R., 2016. Projeto de Melhoria de um Data Center com Ênfase em In-
fraestrutura e Melhoria Energética. Universidade do Vale do Taquari, Lajeado,
RS - Brazil Master’s thesis.
Gao, X., Talebzadeh, A., Moradian, M., Han, Y., Swenson, D., Pommerenke, D., 2015.
Dependence of ESD charge voltage on humidity in data centers: Part III – esti-
mation of ESD-related risk in data centers using voltage level extrapolation and
Chebyshev’s inequality. ASHRAE Trans. 121, 49–57.
Gözcü, O., Özada, B., Carfi, M.U., Erden, H.S., 2017. Worldwide energy analysis of ma-
jor free cooling methods for data centers. In: Proceedings of the 16th IEEE Inter-
society Conference on Thermal and Thermomechanical Phenomena in Electronic
Systems (ITherm), pp. 968–976. doi:10.1109/ITHERM.2017.7992592.
Ham, S.-W., Park, J.-S., Jeong, J.-W., 2015. Optimum supply air temperature ranges
of various air-side economizers in a modular data center. Appl. Therm. Eng. 77,
163–179. doi:10.1016/j.applthermaleng.2014.12.021.
Hydeman, M., Swenson, D., 2010. Humidity controls for data centers are they nec-
essary. ASHRAE J. 52, 48–55.
Islam, M.A., Mahmud, H., Ren, S., Wang, X., 2015. Paying to save: reducing cost
of colocation data center via rewards. In: Proceedings of the 21st Interna-
tional Symposium on High Performance Computer Architecture (HPCA). IEEE,
Burlingame, CA, USA.
LABEE, 2018. Climatic data files, 2018. https://labeee.ufsc.br/downloads/
arquivos- climaticos/formato- try- swera- csv- bin.
Lee, K.-P., Chen, H.-L., 2013a. Analysis of energy saving potential of air-side free
cooling for data centers in worldwide climate zones. Energy Build. 64, 103–112.
doi:10.1016/j.enbuild.2013.04.013.
Lee, K.-P., Chen, H.-L., 2013b. Analysis of energy saving potential of air-side free
cooling for data centers in worldwide climate zones. Energy Build. 64, 103–112.
doi:10.1016/j.enbuild.2013.04.013.
Malkamäki, T., Ovaska, S.J., 2012. Solar energy and free cooling potential in European
data centers. Proc. Comput. Sci. 10, 10 04–10 09. doi:10.1016/j.procs.2012.06.138.
Marcinichen, J.B., Olivier, J.A., Thome, J.R., 2012. On-chip two-phase cooling of dat-
acenters: cooling system and energy recovery evaluation. Appl. Therm. Eng. 41,
36–51. doi:10.1016/j.applthermaleng.2011.12.008.
Oró, E., Depoorter, V., Garcia, A., Salom, J., 2015. Energy efficiency and renewable
energy integration in data centres. Strategies and modelling review. Renew. Sus-
tain. Energy Rev. 42, 429–445. doi:10.1016/j.rser.2014.10.035.
Rong, H., Zhang, H., Xiao, S., Li, C., Hu, C., 2016. Optimizing energy consumption for
data centers. Renew. Sustain. Energy Rev. 58, 674–691. doi:10.1016/j.rser.2015.
12.283.
Santos, A.F., Gaspar, P.D., Souza, H.J.d., 2019. Evaluation of the heat and en-
ergy performance of a datacenter using a new efficiency index: energy us-
age effectiveness design - eued. Braz. Arch. Biol. Techonol. 62. doi:10.1590/
1678- 4324- smart- 2019190021.
Statista, 2019. Worldwide it Spending Data Center Systems 2012–2020 Statis-
tic. Available at: <https://www.statista.com/statistics/496373/datacenter-
colocation-revenueworldwide/>, Statista, 2019.
Shende, S., 2015. Colocation Market by Type (retail Colocation and Wholesale Colo-
cation), End-User (Large Enterprises and Small and Medium Enterprises) and
Industry Vertical (BFSI, Government & Public Sector, Telecom & IT, healthcare
& Life Sciences, Energy) – Global Opportunity Analysis and Industry Forecast,
2014–2020.
Siriwardana, J., Jayasekara, S., Halgamuge, S.K., 2013. Potential of air-side economiz-
ers for data center cooling: a case study for key Australian cities. Appl. Energy
104, 207–219. doi:10.1016/j.apenergy.2012.10.046.
The Green Grid. Updated air-side free cooling maps: the impact of ASHRAE 2011 Al-
lowable Ranges White Paper #46. 2012. Available at https://datacenters.lbl.gov/
sites/default/files/WP46UpdatedAirsideFreeCoolingMapsTheImpactofASHRAE2011
AllowableRanges.pdf.
The Green Grid, 2009. Air-Side Free Cooling Maps.
Wan, F., Swenson, D., Hillstrom, M., Stayer, C., Pommerenke, D., 2013. The effect of
humidity on static electricity induced reliability issues of ICT equipment in data
centers – motivation and setup of the study. ASHRAE Trans. 119, 341–358.
Yuventi, J., Mehdizadeh, R., 2013. A critical analysis of power usage effectiveness
and its use in communicating data center energy consumption. Energy Build.
64, 90–94.
Zhang, H., Shao, S., Xu, H., Zou, H., Tian, C., 2014. Free cooling of data centers: a
review. Renew. Sustain. Energy Rev. 35, 171–182. doi:10.1016/j.rser.2014.04.017.