Applied Energy 94 (2012) 355–363

Contents lists available at SciVerse ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Linking occupants’ thermal perception and building thermal performance

in naturally ventilated school buildings
Han-Hsi Liang a, Tzu-Ping Lin b, Ruey-Lung Hwang a,⇑
a
Department of Architecture, National United University, 1 Lienda, Miaoli 360, Taiwan
b
Program of Landscape and Recreation, National Chung Hsing University, 250 Kuo Kuang Rd., Taichung 402, Taiwan

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

Article history: Building envelope design has an almost unrivalled impact on indoor thermal conditions in naturally ven-
Received 15 September 2011 tilated spaces. This study investigated the effects of building envelope energy regulations on thermal
Received in revised form 27 January 2012 comfort level in naturally ventilated classrooms in primary and secondary schools in Taiwan. In the
Accepted 3 February 2012
study, a long-term survey was conducted to investigate the thermal perception of children and teenagers
Available online 1 March 2012
together with a year-round monitoring of indoor climatic conditions in classrooms of different thermal
characteristics. Based on the results of field comfort surveys an adaptive comfort model for children
Keywords:
and teenagers was developed and compared to ASHRAE’s model for characterizing the requirements of
Adaptive comfort model
Children and teenagers
thermal comfort among the students using naturally ventilated classrooms. An analysis on frequency
Naturally ventilated classroom and level of thermal discomfort in the cool and warm periods was conducted using the adaptive comfort
Building thermal performance model established for the students to develop criteria applicable in thermal discomfort assessment. The
building energy regulation was found to have a significant impact on the level of thermal comfort in nat-
urally ventilated classrooms.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction However, the current AWSG estimation as shown in Eq. (1) does
not take into account the optic performance of the glass, as the win-
The school buildings in Taiwan are typically constructed to al- dows of a NV classroom in Taiwan are usually open in school days
low for the installation of sufficient glass windows as a part of and generally made of single-pane clear glass.
the building’s envelope. The glass windows serve to maintain the The impact of AWSG on the internal environments of NV school
quality of indoor air and to admit the entrance of light and solar buildings has not been quantitatively clarified to date, partly due to
radiation into the building. However, the massive solar heat gained the uncertainty remains in if the traditional PMV model may be ap-
in the building during hot-and-humid months in Taiwan often plied as a criterion in the assessment of thermal comfort in NV
leads to a reduction in thermal comfort for the occupants of the spaces. With the adaptive thermal comfort model being introduced
naturally ventilated (NV) classrooms, affecting the performance as an optional method in the ASHRAE Standard 55 [4], ISO 7730 [5],
of the students in learning [1,2]. Improving the thermal conditions and EN 15251 [6] for evaluating thermal comfort, especially in the
inside a classroom is an urgent issue. NV spaces, many studies in recent years focused on using the adap-
Since 1998 Taiwanese government has established the energy tive thermal comfort model as a tool in assessing the performance
regulation of Average Window Solar Gain (AWSG) to limit the of buildings in terms of facilitating and maintaining an adequate
yearly average solar heat gain through glass windows in school level of thermal comfort (e.g., Radhi et al. [7]; Singh et al. [8]; Lin-
buildings [3]. In this regulation, the AWSG is defined as: den et al. [9]; Henzea et al. [10]; Barlow and Fiala [11]; Pfafferott
P et al. [12]; Holmes and Hacker [13]; and Yao et al. [14]). Here we
IHi  K i  Ai present a study that examined the effects of AWSG on the thermal
AWSG ¼ P ð1Þ
Ai comfort of occupants in NV classrooms in Taiwan using the adap-
tive thermal comfort model.
where IHi is the yearly incident solar radiation on window i in The adaptive thermal comfort model as described in the
kW h/m2 yr; Ki is the corrected factor for an external sun shading ASHRAE Standard 55-2004 is the method most commonly applied
device attached to window i; and Ai is the area of window i in m2. in the assessment of thermal comfort in the NV buildings. This
The Ai excludes the area of windows on the sides of corridors. model, however, was established on a global database of 21,000
measurements taken primarily in office buildings, and was
⇑ Corresponding author. Tel.: +886 37 381643; fax: +886 37 354838. applicable mostly to investigating the thermal comfort of the
E-mail address: [email protected] (R.-L. Hwang). adults. In Taiwan, the majority of the NV classrooms were built

0306-2619/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apenergy.2012.02.004
356 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363

in the primary and secondary schools, with the occupants being effort taken to increase the comprehensiveness of the index might
students consisting of children and teenagers in their pre- or cir- actually introduce more errors to the index than it removed, sug-
cum-pubescence, respectively. These students in general have a gesting that a simple index, such as the operative temperature
lower metabolism [15], faster heat loss [16], and greater sensitivity (top), would usually be sufficient. In the ASHRAE Standard 55, the
to change in the core temperature [17] compared to that of the top was selected as the thermal index, and two sets of allowable
adults. Owing to the age-dependent variation in thermoregulatory limits, one for 80% thermal acceptability and the other for 90%
capacity, the thermal comfort of children and teenagers can also acceptability, were developed to indicate the thermal comfort zone
vary significantly from that of the adults. Thus, in this study an in the adaptive model. To facilitate a comparison between the
adaptive comfort model was established specifically for the chil- observations on thermal comfort in the current study to those pro-
dren and teenagers based on field observations so to provide a jected by the adaptive model in the ASHRAE Standard 55, the top
mechanism for evaluating the indoor thermal comfort of the stu- was used as the thermal index.
dents in NV buildings.
3.2. Calculation of operative temperature
2. Field studies on adaptive thermal comfort in hot-and-humid
The top is the average of the air temperature (ta) and the mean
regions
radiant temperature (tr) weighted by the convective heat transfer
coefficient and the liberalized radiant heat transfer coefficient for
The adaptive comfort model is developed on the findings of field
the occupant. In this study, for occupants engaged in near seden-
surveys concerning thermal comfort [18]. In the governing theory
tary activity (i.e. with a metabolic rate ranging between 1.0 and
of adaptive thermal comfort, people are assumed to consciously
1.3 met) and not directly exposed to sunlight, the top was
or unconsciously respond to their exposure to a given thermal
calculated following the formula presented in the Appendix C of
environment in order to restore their own thermal comfort [19].
ASHRAE Standard 55:
The adaptive theory considers many opportunities to adapt, such
as taking on/off clothing, opening/closing windows, and switching top ¼ w  t a þ ð1  wÞ  t r ð2Þ
on/off fans. It also allows a broad range of thermally acceptable in-
door environments in relation to the outdoor climate. where the weighting factor (w) was a function of air speed. The tr
After de Dear and Brager [20] developed the adaptive model in was estimated using the equation defined in the ISO 7726 [34]:
the 1990s, many field studies on the change of indoor thermal
environment and the occupants’ strategies in acclimation were h i1=4
conducted around the world, including a substantial proportion tr ¼ ðt g þ 273Þ4 þ 2:5  108  v 0:6 ðtg  ta Þ  273 ð3Þ
of these performed in hot-and-humid regions of climates similar
to that of Taiwan. For instances, de Dear and Fountain [21] inves- where tg was the measured globe temperature and v the air velocity.
tigated the indoor climates in the air-conditioned office buildings With regard to the humidity in the air, there were no limits in
in the city of Townsville, Australia; Kwok [22] through field sur- humidity required when the adaptive model developed in the ASH-
veys evaluated the thermal comfort of the subjects in NV class- RAE Standard 55 was used. In addition, the findings reported in
rooms at Hawaii. Wong and Khoo [23] carried out a field study Hwang et al. [30] on thermal comfort among Taiwanese indicated
in NV classrooms in Singapore, and Wong and Feriadi [24] studied that the indoor occupants in Taiwan were not significantly varied
the thermal comfort conditions in NV houses in Indonesia. In Chi- as to if the indoor humidity should be adjusted in different seasons,
na, field studies were conducted in different cities located in the possibly a result of acclimation after continuous and consistent
hot-and-humid climatic region, such as those reported by Zhang exposure to a humid climate, i.e., 60–80% in the mean monthly rel-
et al. [25], Han et al. [26], Yang and Zhang [27], Ye et al. [28] and ative humidity throughout the year. As the air humidity did not ap-
Yao et al. [29]. In Taiwan, Hwang et al. [30,31] reported field sur- pear to exhibit a significant effect on the thermal sensation or
veys performed to realize the thermal comfort of occupants and preference of the students using NV classrooms, it was not consid-
the behaviors of thermal adaptation in different types of indoor ered a factor that influenced the thermal adaptation of students in
thermal environment, including the university classrooms, office this study.
buildings and residential buildings. van Hoof [32] reviewed various
studies conducted in the Asian countries, and found that the ther- 3.3. Criteria applied in assessment of thermal discomfort
mally neutral and preferred temperatures determined from field
observations were often higher than those based on the PMV mod- The Annex H of the ISO 7730 suggests two measures of thermal
el. In addition, Hoof’s analysis revealed considerable differences discomfort, taking into account both the frequency and severity of
between the subjects in different countries. discomfort, for application in the long-term evaluation of general
thermal comfort in an identified indoor environment. These two
measures are: (a) a cooling degree hour measure, which takes dis-
3. Materials and methods comfort to be linearly proportional to the degree by which the dis-
comfort threshold is exceeded by the top; and (b) a weighted
3.1. Thermal index measure concerning the level of thermal dissatisfaction among
the subjects, which takes discomfort to be non-linearly propor-
There are a number of indices for expressing the effect of tional to the departure of the top from the discomfort threshold.
thermal environments on the human body. Examples are the On each occasion, when the actual top falls beyond the lower or
PMV, Effective Temperature (ET), Standard Effective Temperature upper limit (tlimit) of the specified comfort range, the difference is
(SET), and operative temperature. These thermal indices integrate taken, and a weighting factor is applied to quantify the thermal
the effects of environmental variables (air temperature, air humid- discomfort. In the first measure, the weighting factor (wfop) is
ity, mean radiant temperature and air speed) and/or personal calculated as:
variables (clothing level and metabolic rate of activity) on thermal
comfort. Humphreys et al. [33] reviewed a number of studies jt op  t limit j
wherein the calculated values of such indices were correlated to wfop ¼ 1 þ ð4Þ
jtn  tlimit j
the actual comfort votes of the respondents. They found that the
 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363 357

while in the second measure the weighting factor (wfppd) is calcu- preferences as well as the adaptive behaviors of the classroom
lated as: users. The participant’s metabolic rate (met) and clothing insula-
tion (clo) were estimated using a checklist of typical activities/
PPDactual
wfppd ¼ 1 þ ð5Þ clothing and their corresponding numerical met/clo values in
PPDlimit accordance with the ASHRAE Standard 55. The survey procedures
where PPDactual is the actual percentage of the subjects dissatisfied employed were those first developed in Hwang et al. [35] investi-
with the identified thermal environment and PPDlimit the PPD limit gating the adaptive thermal comfort of the students in NV class-
of the specified comfort range. In this study, the significance of ther- rooms. In Hwang et al., an annual thermal neutrality band of
mal discomfort during a warm or cool period of the year was eval- 22.7–29.1 °C and a yearly comfort range of 17.6–30.0 °C were
uated as: determined for 80% acceptability, with the students’ clothing adap-
(i) For a warm period: tation behaviors relating to indoor temperatures considered. Build-
X X ing on the findings from Hwang et al., the current study analyzed
Iop;hot ¼ wfop  t or Ippd;hot ¼ wfppd  t for top
the variation of neutral temperature against outdoor climate and
> t limitðupperÞ ð6Þ validated the regression model used in predicting the level of dis-
satisfaction toward an identified thermal condition. The results
(ii) For a cool period:
X X from these analyses were then integrated to suggest an adaptive
Iop;cold ¼ wfop  t or Ippd;cold ¼ wfppd  t fortop comfort model specifically applicable to children and teenagers
< tlimitðlowerÞ ð7Þ who used the NV classrooms on a continuous and long-term basis.

The estimates of thermal discomfort as derived from Eqs. (6) 3.5. Year-round climatic monitoring
and (7) were denominated as Iop and Ippd and expressed in hours.
The weighting factors wfop and wfppd were those defined in Eqs. Four typical classrooms in one secondary school were chosen
(4) and (5), respectively. The symbol t represented the length of for the year-round monitoring of climatic conditions. These class-
time in hours during which the specified level of thermal discom- rooms were selected as they represented the classrooms of a typi-
fort was experienced. cal architectural layout in Taiwan. Two of them (Classrooms A and
C) were built prior to the implementation of the AWSG program
3.4. Thermal comfort survey and the other two built afterwards (Classrooms B and D). These
classrooms shared a similar floor area and student capacity. They
A series of field surveys were conducted to evaluate the require- however differed in AWSG due to differences in the geographical
ments of thermal comfort among the students using NV class- orientation and in the type of external sun-shading system
rooms in different primary and secondary schools located in equipped in them. The AWSG required for the studied school was
urban and suburban areas of Taichung, Taiwan. A total of 1614 stu- 200 W/m2 yr. All of the monitored classrooms were rectangle-
dents, consisting of 823 girls and 791 boys, in 48 classrooms were shaped and located in the first floor of the buildings, each with
recruited to participate in the comfort survey. Table 1 presents the two walls sitting in parallel exposed to the outdoor environment
gender and age distribution of the study participants. The field and the other two serving as a divide with the adjacent classrooms.
study was conducted from September 1, 2005, to February 5, The monitored classrooms relied on a combination of natural
2006, corresponding to the session of fall semester in local primary cross-ventilation through the windows on opposing walls and
and secondary schools. The survey was performed continuously the ceiling fans to achieve thermal comfort. The measurement of
each week in 2–5 classes throughout the semester. Approximately microclimatic variables indicated that the mean air velocity in
one-third of the participating classes were visited monthly, the classrooms was in the range of 0.6–0.9 m/s when the ceiling
whereas the other classes were visited once or twice during the fans ran and <0.3 m/s when the fans stopped. There were no blinds
study. A total of 3754 sets of environmental monitoring data and or drapes on the windows. The physical characteristics of the mon-
subjective questionnaires were collected, with 52% from high itored classrooms were detailed in Table 2; the locations of these
schools and the remainder from primary schools. classrooms on campus were shown in Fig. 1.
In the comfort survey, a questionnaire was used to evaluate the As the current study focused on the mean thermal comfort level
students’ thermal perception in classrooms of different physical in the NV classrooms as affected by the performance of the build-
configurations and thermal characteristics. The measurement of ing envelope, an indoor climate acquisition system was installed in
microclimatic variables including air temperature, relative humid- the center of the classroom to measure the environmental vari-
ity, air velocity, and mean radiant temperature was simultaneously ables. The measured data were used to indicate the mean thermal
conducted in the classroom. The questionnaires were handed to condition at the specific point in time of measurement and to cal-
the students 30 min after they entered the classroom to ensure culate the instantaneous top. The thermal comfort level of students
the return of their metabolic rates to a sedentary state. The ques- sitting near the window might be different from the averaged level
tionnaire was designed to examine the thermal sensations and in the classroom due to direct sun exposure. Lyons [36], Sullivan
[37] and Gennusa [38] had suggested methods to evaluate the
alteration in PMV and PPD of individuals situated in a sunny,
Table 1
Distribution of gender and age among the study participants.
glazed area. While the risk of overheating by solar radiation should
be considered when analyzing the thermal comfort of individuals
Age Male Female sitting near the windows, it was not addressed in the current study
Number (%) Number (%) as this topic was beyond the scope of our investigation.
12 103 (13) 89 (11) In the ASHRAE Standard 55, by relating the acceptable range of
13 159 (20) 172 (21) indoor temperature to the outdoor climate, the adaptive comfort
14 170 (21) 159 (19) model has accounted for the influence of clothing adjustment on
15 144 (18) 141 (17)
thermal comfort in NV spaces. Since it was not required in adaptive
16 121 (15) 147 (18)
17 94 (12) 125 (15) comfort modeling, the clothing values of the students were not re-
corded in the year-round monitoring. The environmental monitor-
Total 791 (100) 833 (100)
ing was performed from February 16, 2007, to February 15, 2008.
358 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363

Table 2
Physical characteristics of the monitored classrooms.

Parameters Classroom
A B C D
Orientation North–south North–south East–west East–west
WWRa 0.30 0.42 0.55 0.39
External shading South: corridor space South: corridor space north: balcony; West: corridor space West: corridor space east: balcony;
system north: none 2.2 m in depth east: none 2.3 m in depth
AWSGb 153 123 393 267
a
WWR is window wall ratio.
b
AWSG is expressed in kW h/m2 yr.

Table 3
Results of linear regression models established by fitting observed thermal sensation
votes to measured operative temperatures.

Month a b R2 tn toma
September 0.288 8.393 0.89 29.2 28.2
October 0.132 3.791 0.92 28.8 25.7
November 0.052 1.356 0.30 26.5 23.3
December 0.171 3.952 0.93 23.1 17.7
January 0.168 3.695 0.94 22.4 16.8
a
The average monthly outdoor temperatures were from Central Weather Bureau,
Taiwan.

monitored classrooms, with a top typically ranging between 24

and 29 °C, and the occupants could achieve thermal neutrality
through adjustment of clothing. In terms of thermal sensitivity,
the students were of a greater capacity to adapt to the change in
the thermal environment when the slope of the regression de-
creased. As shown in Table 3, the slope value derived in the regres-
sion model fitted for November (0.052) was close to zero and
significantly lower than its counterparts for the other months, indi-
cating that the students were better acclimated to the indoor ther-
mal status in this month. The enhanced acclimation of students to
the indoor microclimate in November could be a result of the stu-
dents’ thermal sensational change being compensated by the effect
of their clothing adaptation, which subsequently contributed to the
Fig. 1. Locations of the monitored classrooms (color block) on campus. (For relative small slope and low R2 values in the November regression
interpretation of the references to color in this figure legend, the reader is referred model.
to the web version of this article.)
Eq. (8) presented a simple linear regression equation for the cal-
culation of neutral temperature by which people experienced a
During the investigated period, the spring semester began on Feb- neutral thermal sensation to the microclimate. At TSV = 0, a ther-
ruary 25, 2007, and ended on June 30, 2007; the fall semester be- mally neutral temperature (tn) could be derived from Eq. (8) for
gan on September 1, 2007, and ended on February 2, 2008. each of the surveyed months. As shown in Table 3, the tn varied
significantly over the investigated period, ranging from 22.4 °C in
the coldest month (January) to 29.2 °C in the warmest month
4. Adaptive thermal comfort model for children and teenagers (September). Fig. 2 shows the distribution of the tns against the
mean monthly outdoor temperatures (tom) for each of the surveyed
4.1. Neutral temperature months. The field observations-fitted regression model revealed a
strong dependency of the tn on the tom, and the model could be
One recognized method for predicting the subjective comfort is mathematically expressed as:
linear regression analysis [39]. In the analysis, the thermal sensa-
tion vote (TSV) of the subjects is typically expressed as: tn ¼ 12:1 þ 0:62  tom R2 ¼ 0:923 ð9Þ
In comparison, the regression model established for NV build-
TSV ¼ a  t op þ b ð8Þ
ings according to the ASHRAE Standard 55 was:
Table 3 tabulated the regression models generating by fitting
tn ¼ 17:8 þ 0:31  tom ð10Þ
the TSVs observed and the tops measured in each of the surveyed
months to Eq. (8), and provided information on the thermal sensi- As demonstrated in Fig. 2, the tn in warm months (September to
tivity of the students (slope), the neutral temperature, and the level November) for children and teenagers in Taiwan was higher than
of statistical significance of the model (R2). The R2 values of these that predicted by the adaptive model in the ASHRAE Standard 55,
regressions suggested a robust relationship between the TSV and while in the cold months (December and January) the tn was close
the top established for the surveyed months except for November. to that recommended by the model in the ASHRAE Standard 55.
In November, the microclimatic conditions were mild in the Cautions should be taken that, as the model described in Eq. (9)
 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363 359

32 32
neutral operative temperature, C

comfort operative temperature,°C

30
28
28
ASHRAE 55
26 optimum
24 90% limits
80% limits
24
this study
22 20
optimum
90% limits
20 80% limits
15 18 21 24 27 30 16
average monthly outdoor temperature, C 15 18 21 24 27 30

observations ASHRAE model fitted model mean monthly outdoor temperature

Fig. 4. Comparison between adaptive comfort model established for children and
Fig. 2. Comparison between neutral operative temperature of children and
teenagers and that from the ASHRAE model.
teenagers and that from the ASHRAE model.

was derived from a regression analysis, this model was not to be

Pcold ¼ eð0:5768Dt4:4666Þ =½1 þ eð0:5768Dt4:4666Þ  R2 ¼ 0:957 ð12Þ
applied to produce a tn estimate if the tom fell outside 16.8–
28.2 °C, the range of tom observed in this study and used for model The denominated PPD index for child and teenagers was then
validation. calculated as the sum of the proportions of hot and cold
discomfort:
4.2. Acceptable deviations from neutral temperature
PPD ¼ Phot þ Pcold ð13Þ
The adaptive comfort model in the ASHRAE Standard 55 speci-
fies a constant width of comfort band centralizing on the tn for 80%
4.3. Limits of thermal comfort zone
and 90% thermal acceptability. While these comfort bands provide
temperature limits beyond which hot or cold discomfort may be
In the evaluation of thermal discomfort among the students in
identified in the NV buildings, they do not serve to measure the
the NV classrooms, a few subjects were found to experience the
significance of the thermal discomfort. A strategy was developed
hot discomfort when the top dropped (Dt < 0) or the cold discom-
in the current study to quantitatively evaluate the likelihood that
fort when the top increased (Dt > 0) (Fig. 3). In addition, about 4%
the students in the NV classrooms would experience thermal dis-
of the surveyed students expressed a status of thermal discomfort
comfort due to hot or cool conditions.
when the top was identical to the tn. Of the 4% experiencing thermal
In general, thermal discomfort may arise when the environmen-
discomfort, three-fourths of them expressed hot discomfort and
tal temperature departs from the comfort temperature; the dis-
the rest cold discomfort, suggesting a preference among the sur-
comfort deteriorates as the departure increases. Nicol et al.
veyed students toward a slightly colder thermal status in the NV
[40,41] suggested an approach to develop a non-linear likelihood
classrooms. This observation was supported by the higher level
model of discomfort. In the current study, the discomfort likeli-
of Phot compared to the level of Pcold at Dt = 0. As far as the thermal
hood model that Nicol et al. developed was employed to predict
comfort zone was concerned, the Dt values corresponding to 10%
the severity of thermal discomfort for the students in the NV class-
and 20% of hot discomfort were 1.7 and 2.9 °C, respectively,
rooms, and the results were shown in Fig. 3. In this model, a logistic
whereas those corresponding to the same proportions of cold dis-
regression analysis was employed to develop a quantitative rela-
comfort were 3.9 and 5.3 °C. These results suggested that the
tionship between the level of hot discomfort (Phot) or cold discom-
students surveyed overall were more sensitive to the increase in
fort (Pcold) and the top as shown in Eqs. (11) and (12), respectively:
top when staying in a NV classroom. Fig. 4 shows the adaptive ther-
Phot ¼ eð0:6802Dt3:7690Þ =½1 þ eð0:6802Dt3:7690Þ  R2 ¼ 0:942 ð11Þ mal comfort model developed for children and teenagers in this
study compared to the model established in the ASHRAE Standard
55. In the thermal comfort zone defined for the students in the NV
80 classrooms, the optimal, 80% acceptable, and 90% acceptable com-
70 fort top received a greater influence from tom than their counter-
parts defined in the comfort zone of the ASHRAE Standard 55 did.
60 The ISO 7730 established PPD limits corresponding to various
Phot or Pcold, %

50 categories of thermal comfort level [5]. To realize the thermal com-

fort requirements of the students in terms of top change in the NV
40
classrooms, the Dt limits corresponding to the different comfort
30 categories in the ISO 7730 were derived using Eqs. (11)–(13) fol-
20
lowing the ISO PPD limits and presented in Table 4.

10
5. Thermal comfort as a function of AWSG
0
-9 -6 -3 0 3 6
Δt, C 5.1. Environment monitoring

observed data for Pcold observed data for Phot Pcold Phot
During the investigated period, the daily maximum outdoor
Fig. 3. Best likelihood of thermal discomfort and proportions of hot or cold temperature was over 30 °C for 160 days, and the daily minimum
discomfort. temperature was less than 10 °C for 10 days. The tom ranged from
360 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363

Table 4
Categories of thermal comfort level for a NV classroom.

Category PPD limitsa Description Dt limitsb

I PPD < 6% Good comfort, no problem 2.8 °C < Dt < 0.7 °C
II 6% 6 PPD < 10% Acceptable, no significant restrictions 3.9 °C < Dt 6 2.8 °C; 0.7 °C 6 Dt < 1.7 °C
III 10% 6 PPD < 15% Comfort slightly reduced, slightly outside of limits 4.7 °C < Dt 6 3.9 °C; 1.7 °C 6 Dt < 2.4 °C
IV 15% 6 PPD Not acceptable Dt 6 4.7 °C; 2.4 °C 6 Dt
a
PPD limits proposed by ISO 7730.
b
Operative temperature departure (Dt) limits proposed by this study.

35 Table 5
AWSG: Statistical distribution of hourly operative temperature monitored in NV classrooms.
indoor operative temperature, °C

34 =123 =153 =267 =393

top (°C) Classroom

33 A B C D
Mean 25.5 25.5 26.3 26.1
32 Std. dev. 4.1 3.8 4.1 4.0
Max 32.2 32.5 33.8 33.3
31 Median 26.1 26.1 26.7 26.6
Min 14.6 15.1 16.2 15.4
30 5th Percentile 18.6 19.0 19.2 19.5
95th Percentile 30.9 30.6 31.9 31.5
29 neutral temperature

28
6/25 6/26 6/27 6/28 6/29
date 100% 4% 5% 7%
4% 3% 11% 11% 10%
7% 4%
Fig. 5. Indoor operative temperatures measured during the hottest week. 8% 7% 7% 29%
23%
10% > 2.4
80% 14%
1.7 –2.4
14%
0.7 –1.7
29% 17%
-2.8 – 0.7
frequency

24 60% 31% 25% -3.9 –-2.8

58% 58%
-4.7 –-3.9
indoor operative temperature, °C

59%
23 50% < -4.7
33%
neutral temperature 40% 17%
22 Unit: C Top
54%
21 11% 45%
20% 13%
5% 8% 11% 29% 27%
6%
20 12% 6% 5%
8% 5% 4%
0%
19 A B C D A B C D

18 cool period warm period

AWSG:
17
=123 =153 =267 =393 Fig. 7. Distribution of change in operative temperature grouped by the ISO 7730
16 thermal comfort categories.
1/22 1/23 1/24 1/25 1/26
date

Fig. 6. Indoor operative temperatures measured during a cold and sunny week. semester) are commonly considered as the warm period, whereas
the months of December, January, February and March are re-
garded as the cool period. In this study, the primary concern with
16.1 °C (January) to 28.4 °C (August). For 158 days, the daily mean the thermal comfort in the warm and cool period was the hot dis-
temperature was over 25 °C. The yearly mean relative humidity comfort and cold discomfort of indoor occupants, respectively. For
was 82.8%. the monitored classrooms, the school hours accumulated in the
Figs. 5 and 6 show the variations of indoor top in the monitored warm and cool period were 774 and 621 h, respectively.
classrooms compared to the tn determined during the hottest week In this study, the hourly indoor top was used to calculate the
and a cold but sunny week of the investigated period, respectively. departure of top from tn(Dt). Fig. 7 shows the distribution of Dt as
The AWSG appeared to exert impacts on the thermal comfort dif- grouped by the Dt limits established in consistence with the ISO
ferently in hot vs. cold days. In hot days, the departure of top from 7730 thermal comfort categories (Table 4) in the monitored class-
tn increased with increasing AWSG (Fig. 5); in cold days, however, a rooms during the warm vs. cold period. During the warm period, as
greater AWSG led to a smaller top departure (Fig. 6). To facilitate a far as the cold discomfort was concerned the Dt in all of the mon-
comparison between the monitored classrooms on the effect of itored classrooms were never below 2.8 °C, the lower boundary
AWSG, Table 5 summarized the statistical distribution of the of the ‘‘good comfort’’ category in the ISO 7730 classification. In
hourly top in the monitored classrooms during the investigated contrast, during the cool period there were 5–11% of the school
period. hours when the Dt exceeded 2.4 °C and rendered the classrooms
‘‘not acceptable’’ because of hot discomfort. Nonetheless, the top
5.2. Levels of thermal comfort in naturally ventilated classrooms in the monitored classrooms during the cool period was generally
less than 28 °C even with a top increase of more than 2.4 °C, and the
In Taiwan, during an academic year the months of May and June students were able to alleviate the hot discomfort by taking off ex-
(the spring semester) and of September and October (the fall cess clothing. Thus, the challenge of hot discomfort to the students
 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363 361

in the NV classrooms during the cool period was less than it ap- In the warm period:
peared from the ISO 7730 thermal comfort projection.
In the warm period, with the lowest AWSG of all monitored Ippd;hot ¼ 7:17 þ 2:22  AWSG R2 ¼ 0:99 ð14Þ
classrooms (123 kW h/m2 yr), Classroom B had 54% of the total
hours the room being used during investigation when the indoor
Iop;hot ¼ 16:71 þ 1:74  AWSG R2 ¼ 0:99 ð15Þ
Dt fell in the ‘‘good comfort’’ ISO 7730 comfort category. When
the ‘‘acceptable’’ category was included, the Dt in Classroom B In the cool period:
was within the ‘‘good comfort’’ and ‘‘acceptable’’ categories for
83% of the school hours. Ten percent of the hours in Classroom B Ippd;cold ¼ 210:68  0:33  AWSG R2 ¼ 0:82 ð16Þ
fell in the ‘‘not acceptable’’ category due to hot discomfort
(Dt > 2.4 °C). Classroom C had the highest AWSG (393 kW h/
Iop;cold ¼ 131:59  0:16  AWSG R2 ¼ 0:94 ð17Þ
m2 yr). However, for Classroom C the Dt was within the ‘‘good
comfort’’ category and the ‘‘good comfort’’ and ‘‘acceptable’’ cate- The magnitude of slope in these regression models revealed the
gories in combination for only 29% and 46% of the school hours, thermal sensitivity of the students (i.e. the significance of thermal
respectively, while for 29% of the hours the Dt change was ‘‘not discomfort) toward a change in the AWSG: a positive slope indi-
acceptable.’’ In the cases of Classrooms A and D, for 76% and 60% cated an increase in thermal discomfort with increasing AWSG; a
of the school time the Dt was in the ‘‘good comfort’’ and ‘‘accept- negative slope suggested a reduction in discomfort when the
able’’ categories in combination, respectively, while the Dt was AWSG increased. Overall, the level of thermal discomfort was more
‘‘not acceptable’’ for 11% and 23% of the hours. influenced by the AWSG change in the warm period than in the
During the cool period, for all of the monitored classrooms the cool period.
Dt in over half (50–58%) of the hours was in the ‘‘good comfort’’ As Fig. 8 shows, the Iop,hot dropped from 685 to 210 h and the
category and in 72–80% in the ‘‘good comfort’’ and ‘‘acceptable’’ Ippd,hot from 849 to 251 h when the AWSG decreased from 393
categories in combination. As far as cold discomfort was con- to 123 kW h/m2 yr. As the level of hot discomfort could be signif-
cerned, only Classroom A had 12% of its use time when the Dt fell icantly alleviated by reducing the AWSG, careful treatment of the
in the cold ‘‘not acceptable’’ category (Dt < 4.7 °C). In compari- AWSG received in a NV classroom could offer an effective oppor-
son, Classrooms B, C, and D had 8%, 5%, and 4% of the school tunity to improve the level of thermal comfort among the stu-
hours when the Dt fell in the cold ‘‘not acceptable’’ category, dents. Typically, the comfort level in a NV classroom may be
respectively. improved by adjusting the AWSG via two mechanisms: control
of the classroom orientation and attachment of external sun-
5.3. Correlation between AWSG and thermal discomfort shading systems to the windows. The properties of the two mech-
anisms were represented by the term IH (the yearly incident solar
As introduced in Section 3.3, in this study the significance of radiation on the window) and K (the corrected factor for an
thermal discomfort among the students in NV classrooms during external sun-shading device attached to the window) as de-
a warm or cool period was characterized as the thermal scribed in Eq. (1), respectively. By making a comparison among
discomfort indices Iop,hot/Ippd,hot (previously shown in Eq. (6)) or Classroom A (AWSG = 153 kW h/m2 yr), C (393 kW h/m2 yr) and
the Iop,cold/Ippd,cold (shown in Eq. (7)), respectively. To compute D (267 kW h/m2 yr), it became apparent that the change in class-
these indices, the hourly Dt collected from field monitoring was room orientation possessed a greater potential in alleviating the
introduced into Eqs. (11)–(13) to generate the hourly PPD index, severity of hot discomfort (demonstrated as the difference in Iop
which was then used to compute the weighting factors as or Ippd between Classrooms A and C in Fig. 8) than the attachment
described in Eqs. (4) and (5) for final derivation of the aforemen- of a sun-shading system did (demonstrated as the difference in Iop
tioned discomfort indices. Fig. 8 illustrates the distribution of the or Ippd between Classrooms C and D). A comparison among
level in thermal discomfort, expressed as Iop,hot/Ippd,hot in the warm Classrooms A, B, and C would lead to the same conclusion. If
period and Iop,cold/Ippd,cold in the cool period, against the AWSG of the classroom orientation and the use of window shades were
the monitored classrooms. These distributions were also fitted combined in the design of a strategy for reduction of thermal dis-
to generate regression models so to quantitatively evaluate the comfort in Taiwanese classrooms, it would be desirable to sit the
relationships between thermal discomfort and AWSG in different classroom in the south-north orientation and to use sun-shading
climatic periods: systems at the same time. However, the emphasis should be on
the architectural orientation, as it weighed more in reducing
thermal discomfort than the sun-shading systems did.
1000 As the findings show, the classroom with the highest AWSG,
Iop,hot Ippd,hot Iop,cold Ippd,cold Classroom C, provided the best thermal comfort during the cool
period. The Iop,cold or Ippd,cold of Classroom C was only 71 or 95 h,
800
respectively, the lowest values among all of the monitored class-
rooms. In comparison, the Iop,cold and Ippd,cold for the classroom of
Iop or Ippd, hrs

600 the lowest AWSG (Classroom B) were 113 and 190 h, respectively.
Increasing the AWSG might appear to have a positive impact on
400 alleviating the cold discomfort in the cold days. However, the ben-
efit of reducing thermal discomfort in the cool period by increasing
200
the AWSG was only marginal when compared to the negative im-
pact the AWSG increase might exert on the thermal comfort in the
warm period—a large AWSG inevitably led to significant thermal
0
0 100 200 300 400 500
discomfort in the warm period or even in the warm, sunny days
in the cool period. Besides, the winter days of extremely cold
AWSG, kWh/m2-yr
weather in Taiwan were frequently accompanied by lasting rains.
Fig. 8. Correlations between the AWSG of monitored classrooms and the severity of Attempting to alleviate cold discomfort in those rainy, cold days
thermal discomfort. by increasing the AWSG would be impractical.
362 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363

6. Conclusions NV buildings in natural ventilation, as the effectiveness of natural

ventilation as a mechanism of indoor environmental control could
A long-term thermal comfort survey was conducted in primary be influenced equally by the solar heat gain taking the form of
and secondary schools to establish the adaptive comfort model for AWSG as well as by the cooling potential of the buildings, as sug-
Taiwanese children and teenagers using NV buildings. The adap- gested by Yao et al. [14]. Identifying the variables involved in the
tive comfort model established for the children and teenagers process of building cooling and integrating these variables into
and the microclimatic conditions monitored in the NV classrooms the current energy regulation AWSG is a topic of interest.
were compared to the levels of AWSG to examine the effects of
AWSG as regulated in Taiwanese building energy codes on improv-
Acknowledgment
ing the thermal comfort of occupants in the NV classrooms. The
findings from this indicated that:
The authors are tremendously thankful to the financial support
from Taiwan’s National Science Council in grant under the Project
 The neutral temperature tn for the students in the hottest
Number: NSC-98-2221-E-239-032.
month was 29.2 °C, 2.3 °C higher than the level suggested by
the ASHRAE Standard 55, 26.9 °C; the tn in the coldest month
was 22.4 °C, which was close to the tn suggested by the ASHRAE References
Standard 55, 23.0 °C.
 A strong relationship was found between the tn and the mean [1] Wargocki P, Wyon DP. The effects of moderately raised classroom
temperatures and classroom ventilation rate on the performance of
monthly outdoor temperature tom. A linear regression model schoolwork by children. HVAC&R Res 2007;13(2):193–220.
was established to describe the dependency of tn on tom: [2] Kameda K, Shuzo M, Kazuhide I, Takamasa K. Study on productivity in the
tn = 12.1 + 0.62tom. The slope of this regression was twice of that classroom (part 3): nationwide questionnaire survey on the effects of IEQ on
learning performance. In: Proceedings of Clima 2007 wellbeing indoors; 2007.
presented in the model, tn = 17.8 + 0.31tom, as defined in accor- p. 1608–11.
dance with the ASHRAE Standard 55. [3] Construction and Planning Agency. Technical codes for energy conservation
 Logistically regressed models were developed for the children design of educational buildings. Ministry of the Interior, Taiwan; 2005.
[4] ANSI/ASHRAE, Standard 55-2004. Thermal environmental conditions for
and teenagers for estimating the PPD occurring in response to
human occupancy. American Society of Heating, Atlanta, Georgia,
temperature difference between the actual operative tempera- Refrigerating and Air-Conditioning Engineers (ASHRAE) Inc.; 2004.
ture top and the neutral top. The curve demonstrating propor- [5] EN ISO 7730. Moderate thermal environments—analytical determination and
interpretation of thermal comfort using calculation of the PMV and PPD
tions of hot discomfort was asymmetrical around the neutral
indices and local thermal comfort; 2005.
top to the curve for proportions of cold discomfort. The mini- [6] CEN EN15251. Indoor environmental input parameters for design and
mum PPD of the students was 4%, a level close to the minimum assessment of energy performance of buildings – addressing indoor air
PPD of 5% suggested in the ISO 7730 model. quality, thermal environment, lighting and acoustics; 2006.
[7] Radhi H, Eltrapolsi A, Sharples S. Will energy regulations in the Gulf States
 Based on the PPD curves established in this study, the upper make buildings more comfortable – a scoping study of residential buildings.
limits for the departure of top from tn for 90% and 80% accept- Appl Energy 2009;86:2531–9.
ability were 1.7 and 2.9 °C, respectively. The lower limits of [8] Singh MC, Garg SN, Jha R. Different glazing systems and their impact on human
thermal comfort-Indian scenario. Build Environ 2008;43:1596–602.
departure for 90% and 80% acceptability were 3.9 and [9] Linden AC, Boerstra AC, Rauec AK, Kurversa SR, de Dear RJ. Adaptive
5.3 °C, respectively. temperature limits: a new guideline in The Netherlands: a new approach for
 In the warm period, for the classrooms of an AWSG level in the assessment of building performance with respect to thermal indoor
climate. Energy Build 2006;38:8–17.
compliance with the 200 kW h/m2 yr norm defined in the build- [10] Henzea GP, Pfafferott J, Herkel S, Felsmann C. Impact of adaptive comfort
ing energy codes, the hourly top change Dt was considered ther- criteria and heat waves on optimal building thermal mass control. Energy
mally ‘‘unacceptable’’ according to the ISO 7730 thermal Build 2007;39:221–35.
[11] Barlow S, Fiala D. Occupant comfort in UK offices – how adaptive comfort
comfort classification in 10% of the hours the classroom was
theories might influence future low energy office refurbishment strategies.
used. In comparison, the Dt in the ‘‘unacceptable’’ thermal com- Energy Build 2007;39:837–46.
fort category was observed in 23–29% of the school hours in the [12] Pfafferott JU, Herkel S, Kalz DE, Zeuschner A. Comparison of low-energy office
buildings in summer using different thermal comfort criteria. Energy Build
classrooms of an AWSG > 200 kW h/m2 yr. During the cool per-
2007;39:750–7.
iod, a high percentage (72–80%) of school hours was identified [13] Holmes MJ, Hacker JN. Climate change, thermal comfort and energy: meeting
wherein the Dt fell in the ‘‘acceptable’’ comfort category. the design challenges of the 21st century. Energy Build 2007;39:802–14.
 Linear correlations between the energy regulation AWSG and [14] Yao RM, Li BH, S Koen, S Alan. Assessing the natural ventilation cooling
potential of office buildings in different climate zones in China. Renew Energy
the indices of thermal discomfort significance, Iop,hot/Ippd,hot for 2009;34:2697–705.
the warm period and Iop,cold/Ippd,cold for the cool period, were [15] Havenith G. Metabolic rate and clothing insulation data of children and
established. The slopes of the regression models revealed that adolescents during various school activities. Ergonomics 2007;50(10):
1689–701.
the Iop,hot/Ippd,hot was more sensitive to the AWSG received in [16] McCullough EA, Eckels S, Harms C. Determining temperature ratings for
the classroom in the warm period compared to the sensitivity children’s cold weather clothing. Appl Ergon 2009;40(5):870–7.
of the Iop,cold/Ippd,cold in the cool period. The Iop,hot/Ippd,hot [17] Anderson GS, Mekjavic IB. Thermoregulatory responses of circum-pubertal
children. Eur J Appl Physiol Occup Physiol 1996;74(5):404–10.
increased with increasing AWSG, whereas the Iop,cold/Ippd,cold [18] Hiroki K, Masaru N, Yoshitaka S, Akio I, Motoyuki S. Comfort and amenity of
decreased when the AWSG rose. classroom environment, part 1: survey of the classroom with air conditioning
and some equipment in elementary school. Summaries of technical papers of
annual meeting architectural institute of Japan, D-1. Environ Eng 1999; 1:
In summary, the adaptive comfort model for children and teen- 1113–4.
agers developed in this study provided a mechanism for the assess- [19] Nicol JF, Humphrey MA. Adaptive thermal comfort and sustainable thermal
ment of thermal comfort in the NV classrooms in primary and standards for buildings. Energy Build 2002;34(6):563–72.
[20] de Dear RJ, Brager G. Developing and adaptive model of thermal comfort and
secondary schools. The quantitative analysis of thermal comfort
preference. ASHRAE Trans 1998;104(1):145–67.
under influences of climatic conditions further illustrated how [21] de Dear RJ, Fountain ME. Field experiments on occupant comfort and office
the level of thermal comfort in a NV classroom changed as a func- thermal environments in a hot-humid climate. ASHRAE Trans 100(2): 457-75.
tion of AWSG. These findings should prove useful for improving the [22] Kwok AG. Thermal comfort in tropical classrooms. ASHRAE Trans
1998;104(1B):1031–47.
efficacy of natural ventilation as a primary means of ventilation in [23] Wong NH, Khoo SS. Thermal comfort in classrooms in tropical. Energy Build
Taiwan. Further study could focus on the cooling potential for the 2003;35:337–51.
 H.-H. Liang et al. / Applied Energy 94 (2012) 355–363 363

[24] Feriadi H, Wong NH. Thermal comfort for naturally ventilated houses in [33] Humphreys MA, Nicol JF, Raja IA. Field studies of indoor thermal comfort and
Indonesia. Energy Build 2004;36:614–26. the progress of the adaptive approach. J Adv Build Energy Res 2007; 1: 55–88.
[25] Zhang G, Zheng C, Yang W, Zhang Q, Moschandreas DJ. Thermal comfort [34] ISO, International Standard 7726, International Standard Organization:
investigation of naturally ventilated classrooms in a subtropical region. Indoor Geneva, ISO 7726:1998. Ergonomics of the thermal environment –
Built Environ 2007;16(2):148–58. instruments for measuring physical quantities; 1998.
[26] Han J, Zhang G, Zhang Q, Zhang J, Liu J, Tian L. Field study of occupants thermal [35] Hwang RL, Lin TP, Chen CP, Kuo NJ. Investigating the adaptive model of
comfort and residential thermal environment in a hot-humid climate of China. thermal comfort for naturally ventilated school buildings in Taiwan. Int J
Build Environ 2007;42:4043–50. Biometeorol 2009;53(2):189–200.
[27] Yang W, Zhang G. Thermal comfort in naturally ventilated and air-conditioned [36] Lyons PR. Window performance for human thermal comfort. ASHRAE Trans
buildings in humid subtropical climate zone in China. Int J Biometeorol 2000;106:594–602.
2008;52(5):385–98. [37] Sullivan R. Thermal comfort issues in the LRI study, internal memorandum/
[28] Ye XJ, Zhou ZP, Lian ZW, Liu HM. August, Field study of a thermal environment report. Windows and Day lighting Group, Lawrence Berkeley National
and adaptive model in Shanghai. Indoor Air 2006;16(4):320–6. Laboratory, Berkeley, CA, USA; 1986.
[29] Yao RM, Liu J, Li BH. Occupants’ adaptive responses and perception of thermal [38] Gennusa ML, Nucara A, Pietrafesa M, Rizzo G. A model for managing and
environment in naturally conditioned university classrooms. Appl Energy evaluating solar radiation for indoor thermal comfort. Sol Energy
2010;87(3):1015–1022,. 2007;81:594–606.
[30] Hwang RL, Cheng MJ, Lin TP, Ho MC. Thermal perceptions, general adaptation [39] de Dear RJ. A global database of thermal comfort field experiments. ASHRAE
methods and occupant’s idea about the trade-off between thermal comfort Trans 1998;104(1):1141–52.
and energy saving in hot-humid regions. Build Environ 2009;44:1128–34. [40] Nicol F, Humphreys MA. Maximum temperatures in European office buildings
[31] Hwang RL, Lin TP, Kuo NJ. Field experiments on thermal comfort in campus to avoid heat discomfort. Sol Energy 2007;81:295–304.
classrooms in Taiwan. Energy Build 2006;38:53–62. [41] Nicol F, Hacker Jake, Spires B, Davies H. Suggestion for new approach to
[32] van Hoof J. Forty years of Fanger’s model of thermal comfort: comfort for all? overheating diagnostics. In: Proceedings of conference: air conditioning and
Indoor Air 2008;18:182–201. the low carbon cooling challenge; 2008.