Building and Environment 111 (2017) 24e32
Contents lists available at ScienceDirect
Building and Environment journal homepage: www.elsevier.com/locate/buildenv
Technical note
Modelling window status in school classrooms. Results from a case study in Italy Francesca Stazi a, *, Federica Naspi b, 1, Marco D'Orazio b, 1 a Politecnica delle Marche, Via Brecce Bianche, Ancona, 60131, Department of Materials, Environmental Sciences and Urban Planning (SIMAU), Universita Italy b Politecnica delle Marche, Via Brecce Bianche, Ancona, 60131, Italy Department of Construction, Civil Engineering and Architecture (DICEA), Universita
a r t i c l e i n f o
a b s t r a c t
Article history: Received 2 August 2016 Received in revised form 1 October 2016 Accepted 22 October 2016 Available online 24 October 2016
In free running buildings users restore their thermal comfort usually opening and closing windows. In school buildings the window use is also useful to achieve good IAQ and avoid health hazards among students and teachers. In last decades many surveys focused on this topic, in order to understand which environmental parameters are the main triggers for users' actions. This paper investigates the relationship between window use and environmental stimuli in an Italian classroom. The survey concerned the monitoring of indoor and outdoor environmental variables and occupants' actions on windows to assess if occupants were influenced by the environment and the daily routine. Linear and logistic regression analyses were carried out to evaluate the relationship between variables and actions. The results highlight that indoor and outdoor temperatures are the main action trigger, while the relationship with CO2 concentration is weak. Also the daily routine affected students' actions, in fact the opening frequency is higher during breaks. Findings from previous studies are confirmed and new insight on behavioral pattern for school classrooms are presented. © 2016 Elsevier Ltd. All rights reserved.
Keywords: Adaptive actions Windows use Schools Regression analyses Behavioral models Trigger parameters
1. Introduction During the last decades several researchers focused their work in observing and understanding the way occupants interact with building controls. One of the most studied action is window opening, because of its frequency [1] and its impact both on users' thermal comfort and on building energy consumptions. This topic acquired an increasing relevance during recent years in particular for the recorded gap between simulated and actual energy consumptions [2e5]. Many surveys were conducted in order to understand what kind of events and parameters were the triggers for users' adaptive actions. Two main usage typologies are usually under study: offices and residential buildings. In relation to the office category, the opening probability was widely studied. Li et al. [6] noted that, in the intermediate season, the opening probability follows the outdoor temperature trend: when the temperature increases also the
* Corresponding author. Fax: þ39 071 220 4729. E-mail addresses:
[email protected] (F. Stazi),
[email protected] (F. Naspi),
[email protected] (M. D'Orazio). 1 Fax: þ39 071 220 4582. http://dx.doi.org/10.1016/j.buildenv.2016.10.013 0360-1323/© 2016 Elsevier Ltd. All rights reserved.
probability becomes higher. Zhang and Barrett [7], analysing an entire office building, assessed that the season affects the proportion of opened windows, reaching the maximum in summer. Among environmental parameters, outdoor temperature is the main trigger but relative humidity and wind direction were significant factors too. Many studies evaluated both window opening and closing patterns. Rijal et al. [8], collecting data from 15 offices, developed an adaptive algorithm (i.e. the Humphreys adaptive algorithm) to predict window status as a function of indoor and outdoor temperature. Haldi and Robinson [9] investigated the influence of occupancy patterns and environmental parameter in 14 offices in Switzerland for seven years. They observed that indoor conditions were better describers for opening actions, instead outdoor parameters have the greatest influence on closings. They also noted different window-uses during the day and, following this pattern, they developed different models for arrival, intermediate period and departure. D'Oca and Hong [10] and ShakibEkbatan et al. [11] confirmed that indoor and outdoor temperature were the most influential factors on windows use, in addition to presence of people and time-related parameters (i.e. time of arrival and leaving). Yun and Steemers [12] endorsed the relationship between time of the day and window use: in offices
F. Stazi et al. / Building and Environment 111 (2017) 24e32
without night ventilation 61% of windows had a change of status from close to open at the first arrival. The authors found also a statistically significant relationship between window opening and closing behaviors and indoor temperature. Analyses in residential buildings outlined similar results. Schweiker et al. [13], collecting data from two monitoring campaigns in Switzerland and Japan, confirmed for both cases, the prevalence of outdoor temperature, immediately followed by the indoor one, on both opening and closing behaviors. Andersen et al. [14] supported the influence of outdoor temperature as a trigger parameter for window closings but they highlighted that the main stimulus for opening was CO2 concentration. Calì et al. [15], gathering measurements from 60 identical apartments in the southern Germany, observed a different window-pattern depending on the time of the day (e.g. more openings in the morning). CO2 concentration, indoor temperature and indoor RH were found to be the main triggers for openings, while the daily average outdoor temperature and indoor temperature influenced the closing probability. Schools were poorly studied with this aim, only Santamouris et al. [16] remarked a correlation between openings and indoor temperature. Table 1 reports the above mentioned surveys: the studies are listed according to the building use (i.e. first the offices, then dwellings and finally the schools) and in chronological order (from 2007 to 2016). To develop behavioral models driven by environmental stimuli, several researchers [17e19] adopted linear correlations to assess the probability of the adaptive action as a function of one or more predictor variables. However, when the output variable is a dichotomous one, the logistic regression analysis is applied more often [8,9,20,21]. The aim is to infer a relationship between the outcome (i.e. the dependent variable y) and a set of predictors (i.e. the independent variables x1, …,xn). Logistic regression is usually preferred to the linear one because the output values (i.e. the occurrence probability p) cannot fall outside the range [0,1] and a better prediction of the upper and lower bounds of the observations are noted [20]. Using this approach the probability distribution is named the logit distribution and is defined as:
pðx1 ; …; xn Þ ¼
expðb0 þ b1 x1 þ … þ bn xn Þ 1 þ expðb0 þ b1 x1 þ … þ bn xn Þ
(1)
where the coefficients b0, …,bn are constants estimated by the regression analysis through maximum likelihood estimation and x1, …,xn are the environmental parameters. This analysis allows to predict the probability for the output variable p to take the “true value”, usually fixed at one for the window state “open” and zero for “close”. This method was successfully adopted in many of the abovementioned surveys [6,9e11,13,14], while the development of such models for school buildings is still lacking. Considering that children spend about one third of their day inside schools and that they need a healthy environment to improve attention and productivity, a model that takes into account students' perceptions and needs (that are reflected in the actions they take) would be recommended. Moreover, occupants' actions have an immediate impact on their thermal comfort and IAQ but also a following effect on building energy consumptions, that is an increasing issue dealing with more and more stringent energy saving standards. In order to fill this gap, the paper aims at investigating students' adaptive actions on windows, identifying their frequency and triggers and assessing the presence of any correlation between actions and environmental variables. 2. Methodology The survey consists of an experimental monitoring campaign
25
that involved the data collection and the simultaneous observation of people's behaviors. The recorded data were analysed and the results were statistically evaluated and compared to previous findings from literature. The following list explains each phase of the survey: 1. Experimental phase: continuous recording of environmental parameters (both indoor and outdoor) and monitoring occupancy patterns, students daily routine and actions on windows; 2. Data processing: associating windows status with time of the day and environmental variables to assess which stimuli are the drivers for users' actions; 3. Analysis: using linear and logistic regression analysis to assess the influence of recorded parameters on windows status and employing goodness-of-fit estimators to evaluate the level of statistical significance of the correlations; 4. Interpretation: comparing results achieved in this paper to findings from previous studies carried out in different building uses to evaluate differences and similarities.
2.1. Case study: school and classroom The survey was carried out in a high school in Italy, built in 2010 in the city of Ancona, located in central Italy (Latitude: 43 580 4900 30 N; longitude: 13 520 5700 01 E; altitude: 67 m). The location is characterized by a hot-summer Mediterranean climate €ppen climatic classification) and by 1688 heating degree days. (Ko The complex is situated in a recent residential zone, without noise sources. The school top view is showed in Fig. 1. The complex is composed by three main blocks: the office tower acts as a pivot and links both the classrooms wing and the auditorium plus gym block. The north-west orientation provides the optimal setting for school activities taking place during the morning, avoiding glare effects and direct solar radiation. The school was built according to European and Italian energy saving regulations, in particular the Law No. 10 of January 9, 1991 [22] and subsequent decrees [23,24]. The school is provided with an heating system with radiators: it is usually turned on from November to April (in relation to outdoor conditions). During the non-heated period the school is naturally ventilated. A single classroom located at ground floor level was selected as a case study: Fig. 2 shows plant and section view of the classroom and Table 2 reports its fundamental attributes. The classroom was occupied by 16 students of 14/15-years-old (5 boys and 11 girls). Lessons were held from Monday to Friday, from 8:00 a.m. to 2:00 p.m. Students' presence in the classroom was variable along the week, because sometimes they moved from the classroom to the laboratories. The tested space presents two double glazing windows in aluminium frame (casement plus hopper window on the top). The hopper windows are 190 cm long and 50 cm high, whereas each shutter of the operable casement is 80 cm long and 130 cm high. 2.2. Experimental methods The survey lasted from the 19th of March to the 29th of April 2015, amounting to 25 days of regular lesson. The heating system was switched on until the 16th of April (included), after this date the classroom was in a free-running condition. Fig. 3 reports a temporal scheme of the monitored period: it includes indoor and outdoor temperature trends and classroom occupancy (hatched areas). The following measurements were carried out according to ISO 7726:2001 [25]:
26
F. Stazi et al. / Building and Environment 111 (2017) 24e32
Table 1 Summary of referenced studies of window use. Authors/year Location
Building use
Monitoring duration
Influencing openings
Influencing closings
Rijal et al. [8] Oxford and Aberdeen, UK Cambridge, UK Yun and Steemers [12] Lausanne, Haldi and Robinson Switzerland [9] Sheffield, UK Zhang and Barrett [17] D'Oca and Frankfurt am Hong [10] Main, Germany Li et al. [6] Chongquing, China Frankfurt am SchakibMain, Germany Ekbatan et al. [11] Schweiker Neuch^ atel, et al. [13] Switzerland Tokyo, Japan Andersen Copenhagen, et al. [14] Denmark Calì et al. Southern [15] Germany
15 offices
Indoor temperature; Outdoor temperature.
6 offices
March 1996eSept. 1997 JuneeSept. 2006
Indoor temperature; Time of the day.
Indoor temperature; Outdoor temperature. Indoor temperature; Time of the day.
14 offices
2001e2008
Indoor temperature.
Outdoor temperature.
an office building
Jan. 2005eApr. 2006
Outdoor temperature; RH; Wind direction.
e
16 offices
Annex 53 database
5 offices
Sept.eOct. 2012
Indoor temperature; Outdoor temperature; Time of the Indoor temperature; Time of the day. day; Outdoor temperature. Outdoor temperature e
16 offices
2004e2009
Indoor temperature; Outdoor temperature.
Indoor temperature; Outdoor temperature.
2 residential buildings e student dormitory 15 dwellings
6 months Summer Outdoor temperature; Indoor temperature. 2007 and winter 2007/ 2008 Jan.eAug. 2008 CO2 concentration.
Outdoor temperature; Indoor temperature.
60 apartments
1 year (2012)
schools
2003e2007
Santamouris Athens, Greece et al. [16]
Time of the day; CO2 concentration; Indoor temperature; Indoor RH; Daily average outdoor temperature; Outdoor RH. Indoor temperature.
Outdoor temperature. Daily average outdoor temperature; Time of the day; Indoor temperature. e
closing door and windows) and contingent relevant events were manually noted by supervisors. During the monitoring the occupants could freely interact with the windows while the door was always closed, except for 20 min during the morning break (from 10:50 to 11:10) and for a few minutes between lessons with different teachers. The monitoring allowed students' adaptive behavior to be related to environmental parameters at the exact time the action was taken. Students' actions were accurately analysed in order to identify if occupants were influenced by the daily routine and to understand what kind of environmental parameters were the main stimuli. 3. Results and discussion
Fig. 1. Plan view of the school and identification of monitored classroom.
The presentation of the results is organized in three subsections, in relation to different topics. Section 3.1 reports data of the recorded environmental parameters and the evaluation of indoor conditions. Section 3.2 analyses the window opening and closing patterns and Section 3.3 presents the results from linear and logistic regression analyses. 3.1. Analysis of environmental conditions
1. Outdoor environmental conditions: outdoor air temperature, solar radiation, wind speed and direction were collected by a climate station located at about 1 km from the school. The station is a ownership of Marche Civil Defence and it collects data from 2009; 2. Indoor climate conditions: indoor air temperature, mean radiant temperature, air speed and CO2 concentration were recorded by the data-logger Babuc/A. The sensors were placed at about 1.1 m above the floor according to ISO 7726:2001 [25] for seated persons, in an almost central position (Fig. 2). Table 3 summarizes the features of the installed equipment. Every day of lesson real occupancy (i.e. number of students and teachers), adaptive actions (i.e. opening/
Table 4 shows a summary of monitored parameters during the survey, their variation range and a statistical analysis including the mean, the median and the standard deviation. The analysis of wind speed and direction is shown in Fig. 4: the hatched area represents the wind mean speed in relation to the different wind directions, while the dashed line displays the maximum intensity reached during the survey. The prevalent directions are both from East and from North West, respectively opposite and in line with the classroom orientation. High wind speed in the NW direction was rarely monitored, while the mean intensity was too low (about 2.5 m/s) to provoke discomfortable air drafts. In relation to CO2 concentration, Table 4 shows that the mean
F. Stazi et al. / Building and Environment 111 (2017) 24e32
27
Fig. 2. Plant and section of the selected classroom: sensor network distribution inside the building.
period, while the non-heated period is reported in the lower part. The proportion of windows open is represented by the hatched areas (considering that one window fully open accounts for 50%). When the heating system was on the indoor temperature was usually higher than 20 C (about 21 C) but window openings provoked sudden and sensible decreases, due to the high gradient between indoor and outdoor temperatures. When the heating
Table 2 Main features of the classroom. 7.08 7.88 54.65 3.20 174.88 0.313
Size (m) Net floor area (m2) Internal height (m) Heated volume (m3) Ratio S/V
Fig. 3. Temporal scheme of the monitored period (from 19/3/’15 to 29/4/’15).
Table 3 Features of sensors installed in the classroom. Sensors
Environmental parameter
Frequency
Output signal
Range
Accuracy
PT 100 Globe thermometer Anemometer CO2 sensor
Indoor air temperature Mean radiant temperature Indoor air speed CO2 concentration
5 min. 5 min. 1 min. 15 min.
50 ÷ 80 C 40 ÷ 80 C 0,01÷ 20 m/s 0 ÷ 3000 ppm
0,15 C 0,15 C 0 ÷ 0,5 m/s 20 ÷ 70 ppm
value is lower than the Building Bulletin limit (1500 ppm) but overcomes the ASHRAE one (1000 ppm). The highest value of 3000 ppm was reached few times, but always in conditions of maximum occupancy (18 people), with all windows closed and the heating system switched on (indoor temperature about 22 C). Fig. 5 shows the indoor and outdoor temperature trends in two different periods: on the top are shown 5 days during the heated
C C m/s ppm
system was turned off the indoor temperature remained on the average value of 20 C (lower than in the previous period). In this case the windows opening didn't cause consistent fluctuations for the little difference between indoor and outdoor temperature. The most frequent and lasting openings occurred during the heated period, suggesting that occupants felt frequently discomfort sensations for warm environment. In order to assess if comfort
28
F. Stazi et al. / Building and Environment 111 (2017) 24e32
conditions were matched, the PMV and PPD indices were calculated. Table 5 reports the mean values of PMV and PPD (according to ISO 7730 [26] for buildings in Category II) and the percentages of discomfort for hot and cold environment for the heated and nonheated periods, respectively. The average PMV for the heated period can be considered satisfactory because it falls within the interval [-0.5 0.5], even if it is very near to the lower limit, while the PPD is slightly over than the limit of 10%. For the non-heated period both PMV and PPD indices are fully outside the comfort range. The discomfort for both the periods is due to cool environments: about 33% of discomfort in the first period and 100% in the second one. The frequent window opening behavior suggests that the model slightly underestimates the users' thermal sensation in the specific case study. 3.2. Environmental variables and opening behaviors A preliminary phase regarded the calculation of air renovation from different window positions, according to ISO 13779:2008 [27] and EN 15251 [28]. Analysing the CO2 concentration decay it has been found that the air renovation when a single casement is open is about 8 l/s per person, while one window fully open (two casements and one hopper window) provide 20 l/s per person. To assess whether occupants are influenced by the daily routine, the openings frequency was analysed. The lesson time (from 8:00 to 14:00) was divided into time intervals of 10 min. Fig. 6 highlights that the openings frequency is higher in the first half of the morning, and in particular during the mid-morning break (10:50e11:10), at students' arrival (8:00e8:10) and between the end of one lesson and the beginning of the following one (each lesson lasts 1 h). Breaks are the preferred moments to intervene [16] because students can act freely, without interrupting teachers' explanations and compromising their own attention. Moreover, when students' concentration is focused on lessons they are less sensitive to environmental stimuli. According to previous studies [9,10,12,15], the results confirm that the daily routine has a great influence on users' actions and frequency both in offices and dwellings and school as well. Opening and closing patterns were analysed in relation to the recorded environmental parameters. In addition to the total one, the action frequency was studied separately for the heated and non-heated seasons to evaluate the differences between the periods (Figs. 7 and 8). According to results from previous studies (reported in Section 1), the actions were analysed in relation to indoor temperature, outdoor temperature and CO2 concentration. In particular, window openings were directly correlated with indoor temperature and CO2 concentration. Fig. 7a and c report, respectively, openings absolute frequency in connection to indoor temperature and CO2 concentration, while Fig. 7b and d shows the total cumulative frequency distribution (FCD). No opening was recorded for indoor temperature lower than 19 C and for CO2 concentrations lower than 498 ppm. The highest frequency of openings was recorded for indoor temperature between 21 and
Table 4 Summary of recorded parameters. Variable
Max
Min
Mean
Median
St. Dev.
Outdoor temperature ( C) Solar radiation (W/m2) Wind speed (m/s) Indoor temperature ( C) Mean radiant temperature ( C) Indoor air speed (m/s) CO2 concentration (ppm)
25.62 993.00 9.60 23.10 22.76 0.45 3000
3.80 0.00 0.00 15.15 16.10 0.00 443
16.55 501.75 2.25 21.06 20.61 0.01 1170
16.65 514.67 1.80 21.16 20.66 0.00 1059
3.95 269.61 1.85 0.99 0.96 0.02 474.55
Fig. 4. Analysis of wind speed and direction.
22.5 C and for CO2 concentrations between 1000 and 2000 ppm. About 37% of opening occurred for indoor temperature lower than 21 C, 57.5% in the range from 21 to 22.5 C and only 5% for temperature higher than 22.5 C. In relation to CO2 concentrations, about 33% of openings took place for values lower than 1000 ppm, the 59% in the interval 1000e2000 ppm and only the 8% when the CO2 overcomes 2000 ppm. The data concerning window closing frequency, joined to indoor and outdoor temperature, are reported in Fig. 8a and c, while in Fig. 8b and d their respective cumulative frequency distributions (FCD) are plotted. Closing events related to the end of morning lessons were excluded from the data, because they are not environmentally triggered. The results show that total closing events are concentrated in the interval between 20 and 22 C for indoor temperature and their occurrence was higher when outdoor temperature was lower than 17.5 C. The highest percentage (54%) of the window closings was reached for indoor temperature lower than 21 C, 42% in the range 21e22.5 C and only 4% for temperature higher than 22.5 C. Regarding the relationship between closing actions and outdoor temperature, 42% of closing occurred when the temperature was lower than 14.5 C, 34% in the range 14.5e19 C and a percentage of 24% for values higher than 19 C. This trend highlights that the lower the outdoor temperature was, the higher was the probability of a closing event (Fig. 8c). Analysing separately the patterns for the heated and non-heated periods, it can be noted that the highest interaction with windows occurred during the heated period. Openings increased when the indoor temperature overcame 21.5 C, while the peak for closing events occurred when outdoor temperature is about 13 C. The low actions frequency recorded during the non-heated season is a consequence of the lower gradient between indoor and outdoor temperature than that occurred during the heated one. 3.3. Adaptive behavioral models Stochastic models have the function to estimate the human behaviors and to develop a probabilistic relationship with the independent variables. In order to perform a linear regression
F. Stazi et al. / Building and Environment 111 (2017) 24e32
29
Fig. 5. Proportion of window open, indoor and outdoor temperatures during 5 days when the heating was on (on the top) and off (at the bottom).
Table 5 PMV and PPD calculations for heated and non-heated periods. Periods
Mean PMV
Mean PPD
% Discomfort hot
% Discomfort cold
Heated Non-heated
0.42 1.04
10.24 29.36
0 0
33.33 100
concentration (p > 0.001) as for the linear one. Values for area under ROC curve (AUC), McFadden's R2 and Neglekerke's R2 are provided in Table 8 and are used as goodness-of-fit estimators. Predictions of window opening and closing probability as a function of indoor temperature are significant, in fact the AUC (0.719)
Fig. 6. Window opening occurrence during the teaching period.
analysis with a larger sample, the evaluation included not only the single “opening event” but also the probability to find at least one window open. In the same way the “window closing” is translated in the probability to find all windows closed. Fig. 9 shows the obtained linear regression models (dashed lines) on data (square points), while Table 6 reports the equations coefficients and the R2 for each correlation. The best correlations are with indoor temperature and both window-open and close probability (R2 ¼ 0.851 and R2 ¼ 0.976), while the lowest is the one relating window opening to CO2 concentration (R2 ¼ 0.268). Logistic regressions were performed using the same variables for those linear. Fig. 9 shows the regression curves (solid lines) on data (square points). Table 7 reports outcomes from logistic regression: regression parameters are all significant, except for CO2
and the generalized R2 (0.112 and 0.192) are of better statistical quality. Logistic regressions with indoor temperature produce the same coefficients with the opposite sign, because data for the “opening” group are the complementary of the “closing” one. The relation between closings and outdoor temperature is statistically less significant, because of its lower values both for the AUC (0.582) and the R2 (0.017 and 0.031). As expected, a monotonically increasing function relates the probability to find a window open and the indoor temperature, while a decreasing function links windows closing and both indoor and outdoor temperature. Results for the correlation between openings and CO2 are not satisfying, confirming previous findings [16]. It means that users are not driven by this stimulus because of their unawareness of indoor CO2 concentration increasing.
Fig. 7. Openings frequency in relation to indoor temperature (a) and CO2 concentration (c) and the total cumulative frequency distribution (b and d).
Fig. 8. Closings frequency in relation to indoor (a) and outdoor temperature (c) and the total cumulative frequency distribution (b and d).
F. Stazi et al. / Building and Environment 111 (2017) 24e32
31
Fig. 9. Observational data, linear and logistic regressions.
Table 6 Parameters for linear correlations. Correlations
a
b
R2
open open close close
3.0877 0.6083 4.5894 0.9336
0.1723 0.0001 0.195 0.0305
0.851 0.268 0.976 0.587
-
T.in. CO2 T.in. T.ext.
In summary, the outcomes from the analyses highlight that the best predictor was the indoor temperature, both for the probability to find at least one window open and for the probability to find all windows closed, supporting previous results [8e12,16]. The outdoor temperature is an influencing parameter too, even if it has a lower impact than the indoor one. Many researchers assessed that some physical factors (different from temperature) can influence occupants' behaviors on windows [20]. Outdoor noise and perceived illumination are cited among the triggers for keeping the windows closed [9,18,29]. Even if these are important factors, in the present survey they were not considered because in the school surrounding any noise and pollutant sources are present and the classrooms orientation avoid the penetration of
direct sunlight. The results obtained in the survey can't be generalized for all the temperature intervals. The proposed models can be considered accurate only within the temperature ranges occurred during the monitoring period, namely between 19 C and 23 C for the indoor temperature and about 11e25 C for the outdoor one. In fact the opening/closing functions may assume different trends for indoor and outdoor temperatures higher than those occurred during the survey. Many studies [9,18,30] have found a decreasing in window opening when the indoor temperature overcomes about 27 C to avoid the heat entrance from the outside. However it should be noted that in the Italian climate such conditions are usually reached during the summer season, when schools are already closed (in Italy the school year ends at the beginning of June). Even if the temperature intervals are limited, it was found that the correlation with window opening follows a trend similar to those obtained in other studies [9,13]. Thus, indoor and outdoor temperatures are variables to be considered in an adaptive behavioral model. These results confirm findings from previous studies [9e13,16], although they were developed for different climates and building uses. The daily routine affects the frequency of
Table 7 Regression parameters for logistic models. Correlations
b0
Z
b1
Z
p-value
Open e T.in. Open e CO2 Close e T.in. Close e T.out.
19.427 ± 1.344 0.562 ± 0.299 19.427 ± 1.344 1.251 ± 0.507
13.414 3.679 13.414 4.827
0.922 ± 0.0685 0.00065 ± 0.001 0.922 ± 0.0685 0.087 ± 0.031
13.504 3.190 13.504 5.536
0.000 0.001 0.000 0.000
32
F. Stazi et al. / Building and Environment 111 (2017) 24e32
Table 8 Goodness-of-fit estimators for logistic models. Correlations
AUC
McFadden's R2
Neglekerke's R2
Open e T.in. Open e CO2 Close e T.in. Close e T.out.
0.719 0.512 0.719 0.582
0.112 0.005 0.112 0.017
0.192 0.010 0.192 0.031
actions too, confirming that users are deeply influenced by their activities and habits [9,10,12,15]. 4. Conclusions The aim of the present study was the investigation of window use in an Italian school classroom. Environmental parameters and users' adaptive actions were continuously recorded for about one month during the intermediate period and linear and logistic correlations were carried out. The findings highlight that indoor temperature is the best predictor both for openings and closings. Outdoor temperature has a lower influence on occupants' actions but it is statistically significant too. CO2 concentration is found to have no statistical meaning. Daily routine and habits highly influence students' behaviors, in fact window openings are focused during breaks especially in the first part of the morning. Building simulation programs should introduce the real user behavior, the time-related events and the trigger parameters in order to obtain more realistic results. A behavioral model, developed on observational data, would improve simulation outcomes and would guide the design of smart energy efficient buildings. The findings from the surveys provide a first aid in this direction, however further researches are needed to validate the obtained results for a wider sample and to analyse students' actions in different seasons and climates. References [1] S. Barlow, D. Fiala, Occupant comfort in UK offices-How adaptive comfort theories might influence future low energy office refurbishment strategies, Energy Build. 39 (2007) 837e846, http://dx.doi.org/10.1016/ j.enbuild.2007.02.002. [2] A. Roetzel, A. Tsangrassoulis, U. Dietrich, S. Busching, A review of occupant control on natural ventilation, Renew. Sustain. Energy Rev. 14 (2010) 1001e1013, http://dx.doi.org/10.1016/j.rser.2009.11.005. [3] A. Al-Mumin, O. Khattab, G. Sridhar, Occupants' behavior and activity patterns influencing the energy consumption in the Kuwaiti residences, Energy Build. 35 (2003) 549e559, http://dx.doi.org/10.1016/S0378-7788(02)00167-6. [4] I. Gaetani, P.J. Hoes, J.L.M. Hensen, Occupant behavior in building energy simulation: towards a fit-for-purpose modeling strategy, Energy Build. 121 (2016) 188e204, http://dx.doi.org/10.1016/j.enbuild.2016.03.038. [5] P. De Wilde, The gap between predicted and measured energy performance of buildings: a framework for investigation, Autom. Constr. 41 (2014) 40e49, http://dx.doi.org/10.1016/j.autcon.2014.02.009. [6] N. Li, J. Li, R. Fan, H. Jia, Probability of occupant operation of windows during transition seasons in office buildings, Renew. Energy 73 (2015) 84e91, http:// dx.doi.org/10.1016/j.renene.2014.05.065. [7] Y. Zhang, P. Barrett, Factors influencing occupants' blind-control behaviour in a naturally ventilated office building, Build. Environ. 54 (2012) 137e147, http://dx.doi.org/10.1016/j.buildenv.2012.02.016. [8] H.B. Rijal, P.G. Tuohy, M. Humphreys, J.F. Nicol, A.A. Samuel, J. Clarke, Using results from field surveys to predict the effect of open windows on thermal comfort and energy use in buildings, Energy Build. 39 (2007) 823e836, http://
dx.doi.org/10.1016/j.enbuild.2007.02.003. [9] F. Haldi, D. Robinson, Interactions with window openings by office occupants, Build. Environ. 44 (2009) 2378e2395, http://dx.doi.org/10.1016/ j.buildenv.2009.03.025. [10] S. D'Oca, T. Hong, A data-mining approach to discover patterns of window opening and closing behavior in offices, Build. Environ. 82 (2014) 726e739, http://dx.doi.org/10.1016/j.buildenv.2014.10.021. [11] K. Schakib-Ekbatan, F.Z. Çakıcı, M. Schweiker, A. Wagner, Does the occupant behavior match the energy concept of the building? e Analysis of a German naturally ventilated office building, Build. Environ. 84 (2015) 142e150, http:// dx.doi.org/10.1016/j.buildenv.2014.10.018. [12] G.Y. Yun, K. Steemers, Time-dependent occupant behaviour models of window control in summer, Build. Environ. 43 (2008) 1471e1482, http:// dx.doi.org/10.1016/j.buildenv.2007.08.001. [13] M. Schweiker, F. Haldi, M. Shukuya, D. Robinson, Verification of stochastic models of window opening behaviour for residential buildings, J. Build. Perform. Simul. 5 (2012) 55e74, http://dx.doi.org/10.1080/ 19401493.2011.567422. [14] R. Andersen, V. Fabi, J. Toftum, S.P. Corgnati, B.W. Olesen, Window opening behaviour modelled from measurements in Danish dwellings, Build. Environ. 69 (2013) 101e113, http://dx.doi.org/10.1016/j.buildenv.2013.07.005. [15] D. Calì, R.K. Andersen, D. Müller, B. Olesen, Analysis of occupants' behavior related to the use of windows in German households, Build. Environ. 103 (2016) 54e69, http://dx.doi.org/10.1016/j.buildenv.2016.03.024. [16] M. Santamouris, A. Synnefa, M. Asssimakopoulos, I. Livada, K. Pavlou, M. Papaglastra, et al., Experimental investigation of the air flow and indoor carbon dioxide concentration in classrooms with intermittent natural ventilation, Energy Build. 40 (2008) 1833e1843, http://dx.doi.org/10.1016/ j.enbuild.2008.04.002. [17] T. Johnson, T. Long, Determining the frequency of open windows in residences: a pilot study in Durham, North Carolina during varying temperature conditions, J. Expo. Anal. Environ. Epidemiol. 15 (2005) 329e349, http:// dx.doi.org/10.1038/sj.jea.7500409. [18] P. Warren, L. Parkins, Window-opening behaviour in office buildings, Build. Serv. Eng. Res. Technol. 5 (1984) 89e101. [19] V. Inkarojrit, G. Paliaga, Indoor climatic influences on the operation of windows in a naturally ventilated building, in: Proc. 21th Int. Conf. Passiv. Low Energy Archit. Netherlands, 2004, pp. 427e431. [20] H.B. Gunay, W. O'Brien, I. Beausoleil-Morrison, A critical review of observation studies, modeling, and simulation of adaptive occupant behaviors in offices, Build. Environ. 70 (2013) 31e47, http://dx.doi.org/10.1016/ j.buildenv.2013.07.020. [21] H.B. Rijal, P.G. Tuohy, J.F. Nicol, M.A. Humphreys, A.A.A. Samuel, J.A. Clarke, Development of adaptive algorithms for the operation of windows, fans and doors to predict thermal comfort and energy use in Pakistani buildings, ASHRAE Trans. 114 (2008) 555e573. http://strathprints.strath.ac.uk/9382/. [22] Legge 9 Gennaio 1991 n.10, Norme per l’attuazione del piano energetico nazionale in materia di uso razionale dell’energia, di risparmio energetico e di sviluppo delle fonti rinnovabili di energia, 1991 (in Italian). [23] Decreto Legislativo 19 Agosto 2005, n.192 Attuazione Della Direttiva Europea 2002/91/CE Relativa Al Rendimento Energetico Nell’edilizia, 2005 (in Italian). [24] Decreto del Presidente della Repubblica 2 Aprile 2009, N.59 Regolamento di attuazione dell’articolo 4, comma 1, lettere a) e b), del decreto legislativo 19 agosto 2005, n. 192, concernente attuazione della direttiva 2002/91/CE sul rendimento energe, 2009. [25] ISO, EN ISO 7726:2001, Ergonomics of the Thermal Environment - In Europe en de Normalstruments for Measuring Physical Quantities, Comite isation, Bruxelles, 2001. [26] I. 7730, Ergonomia degli ambienti termici Determinazione analitica e interpretazione del benessere termico mediante il calcolo degli indici PMV e PPD e dei criteri di benessere termico locale UNI EN ISO 7730, 2006. [27] UNI EN 13779, EN 13779 Requisiti di prestazione per i sistemi di ventilazione e di climatizzazione, 2008. [28] CEN Standard EN 15251, Indoor Environmental Input Parameters for Design and Assessment of Energy Performance of Buildings- Addressing Indoor Air Quality, Thermal Environment, Lighting and Acoustics Contents, 2006. [29] I.A. Raja, J.F. Nicol, K.J. McCartney, M.A. Humphreys, Thermal comfort: use of controls in naturally ventilated buildings, Energy Build. 33 (2001) 235e244, http://dx.doi.org/10.1016/S0378-7788(00)00087-6. [30] S. Herkel, U. Knapp, J. Pfafferott, Towards a model of user behaviour regarding the manual control of windows in office buildings, Build. Environ. 43 (2008) 588e600, http://dx.doi.org/10.1016/j.buildenv.2006.06.031.