A SENSITIVITY ANALYSIS OF NATURAL VENTILATION DESIGN PARAMETERS FOR NON RESIDENTIAL BUILDINGS Annamara Beller* 12, Spencer Dutton 3, Ulrch Flpp Oberegger 1, Roberto Lolln 1 1 Insttute for Renewable Energy, EURAC research, Bolzano, Italy *annamara.beller@eurac.edu 2 Unverstà degl Stud d Bergamo, Italy 3 Lawrence Berkeley Natonal Laboratory, Berkeley, Calforna ABSTRACT Desgners can use natural ventlaton to reduce buldng energy use and mprove occupant comfort. Ths study ams to provde gudance for desgners of naturally ventlated buldngs on the mpact varous desgn parameters have on natural ventlaton performance. We modelled a naturally ventlated reference offce buldng, under a range of clmate condtons and ventlaton strateges. We performed a senstvty analyss of a range of unknown desgn parameters used n the early desgn stage. The results underlne the most mportant parameters and ther effect on natural ventlaton strateges n dfferent clmate types. Of the parameters assessed, arflow network parameters had the most sgnfcant mpact on mprove thermal comfort. Envelope characterstcs were found to have a sgnfcant mpact also on nght coolng strateges n clmates wth large durnal swng. The results showed a domnant nfluence of the parameters that drectly mpact solar and nternal gans on natural ventlaton performance. Quanttatve measures of the mpact of varous nput parameters are presented here whch, wth the use of dynamc buldng smulaton, can be used to support natural ventlaton desgn. INTRODUCTION Natural ventlaton offers potental benefts to buldng owners and operators, both n terms of reduced energy use, resultng n lower operatng costs, and mproved occupant satsfacton. However the performance of any natural ventlaton strategy s hghly dependent on factors that are decded on early n the desgn process. Many of the key desgn decsons, ncludng the aspect rato of a buldng, the presence of courtyards or atrums, and the sze and locatons of wndows and openngs, have a sgnfcant mpact on whether or not natural ventlaton s feasble. To explot ts full potental, natural ventlaton desgn has to be consdered throughout the complete desgn process, partcularly n early desgn stages. Dynamc buldng smulaton tools can be appled durng early desgn phases to nform natural ventlaton desgn and test desgn strateges. Coupled buldng energy and arflow smulaton software, such as EnergyPlus can be used to smultaneously model buldng energy use, natural ventlaton arflow, and occupant comfort. Often however many of the buldng desgn parameters that mpact natural ventlaton performance, are not defned durng the early desgn stages. Ths uncertanty presents a sgnfcant challenge for smulaton engneer when modelng potental ventlaton strateges. Parametrc analyss can be used to some extent to gve a range of potental outcomes gven uncertanty n the model nput parameters. However performng a parametrc analyss that ncludes all of the unknown desgn parameters requres sgnfcant tme and resources. Several pror studes have attempted to dentfy whch smulaton parameters have the most mpact on buldng performance (Domınguez- Munoz F., 2010) (Hopfe C., 2009) (Breesch H., 2010) (De Wt S., 2001). These studes focused prmarly on smulaton varables rather than the buldng desgn parameters that are lkely to evolve throughout the desgn process. No pror studes have been dentfed that focus on the uncertanty n buldng energy models that are coupled wth an arflow network model. Ths study ams to provde gudance for desgners of naturally ventlated non-resdental buldngs on the mpact varous desgn parameters have on natural ventlaton performance. We performed a senstvty analyss on key desgn parameters that cannot be clearly specfed durng early desgn stages, these nclude, nternal and solar gans, envelope characterstcs and wndow geometry and openng type. The results of ths work can be appled to support natural ventlaton desgn usng buldng smulaton. METHOD Reference buldng The senstvty analyss s performed on a four-storey offce buldng north-south orented, ntended to be representatve of a typcal European medum szed offce. The buldng layout s symmetrc around the central starway and servces. Fgure 1 shows the plan vew of one of the four stores. Each floor has four open plan offce zones connected through vents to the starwell passve stack. An openng at the ground floor and openngs
on the roof allow the stack effect. Table 1 gves geometrc characterstcs of the buldng and Fgure 2 shows the buldng model n Sketchup. Table 1 Buldng geometry characterstcs. Wndow to Wall Rato (WWR) 35% Net condtoned area 2734 m 2 Volume (50m x 15m x 15m) 11250 m 3 5 15 5 5 50 20 10 20 Fgure 1 Buldng plan vew. Fgure 2 Buldng model geometry. Each wndow frame has two panes, the left one for natural ventlaton and the rght one for daylghtng. Overhangs have been added to the south façade to reduce the solar gans durng summer. Automatcally controlled wndows have been added to the south and north façade. Vents between the hallways and the offces allow ar movement even f the doors are closed. Modellng method We modelled two commonly used passve ventlaton strateges usng the mult-zone arflow network model wth the EnergyPlus buldng energy smulaton program: nght stack drven cross ventlaton, to reduce coolng needs (Strategy A), and wnd drven cross ventlaton durng the day, to mprove thermal comfort (Strategy B). In Strategy A, we assume wndows are closed durng offce hours, and that coolng loads are met usng mechancal coolng. Wndows and vents open autonomously f the ndoor temperature exceeds 26 C, allowng nght-tme free coolng through the starwell passve stack. The modelled strategy vares the openng area dependng on the nsde-outsde temperature dfference, n order to reduce the possbly large fluctuatons n temperature. We modelled starwell passve stack as four vertcally stacked zones connected to each other through horzontal openngs and to the offce floors through vents. In Strategy B, we assume wndows are operated durng normal offce hours by occupants based on ther thermal comfort. When wndows are n use, ventlaton coolng s provded va naturally drven cross ventlaton. The modelled control strategy s based on CEN 12521 adaptve comfort model and allows wndow openng durng the day f the ndoor operatve temperature s greater than the adaptve comfort temperature. We assumed n ths case that the buldng s n free-runnng mode wth no mechancal heatng or coolng. For both smulaton scenaros we used wnd pressure coeffcents from the AIVC dataset for semsheltered low-rse rectangular buldngs (Lddament M.W., 1986). We modelled unntentonal nfltraton by applyng Equvalent ar Leakage Areas (ELAs) evenly over the exteror envelope surface area of the buldng model. We appled four dfferent ELAs derved from the four classfcatons of buldng envelope performance specfed n the KlmaHaus buld regulatons. We calculated ELAs from the maxmum ACH at 50 Pascals, as defned n KlmaHaus, usng translaton equatons defned n ASHRAE (TC, 2001). Wndow poston and dmenson were not consdered varables. We developed hourly lghtng schedules to represent the use of artfcal lghtng. We generated these schedules by frst performng a buldng smulaton run usng the daylghtng controls opton n EnergyPlus. Ths opton uses the antcpated avalablty of daylght comng through the wndows, to moderate electrc lghtng on needed bass. Performng buldng smulatons usng ths opton s computatonally costly, and so ths opton was only used to generate our electrc lght usage schedules. Smulatons performed for our parametrc analyss used EnergyPlus s FullExteror shadng mode whch assumes no daylghtng modeled; resultng n sgnfcantly qucker smulaton run tmes. We used 15 mnute smulaton run tme steps based on analyss by (Zha J., 2011) that demonstrated 15 mnute tme steps were suffcent to the couplng of arflow and thermal models. Parameter selecton and varable range assessment Our senstvty analyss took nto account only uncertan buldng propertes due to lack of nformaton durng early desgn stage. Input parameters that were prmarly smulaton envronments parameters (.e. convecton coeffcents, zone ar heat balance algorthm) were not vared n our analyss, as these type of parameters would typcally reman constant throughout the desgn process. We assgned value ranges for the envelope thermal nsulaton, ar tghtness (ELA) and densty based on a
Average durnal temperature [K] Av. global horz. radaton [Wh/m²] combnaton of buldng regulatons (D.Lgs.192/05), energy performance requrements (KlmaHaus energy certfcaton) and techncal feasblty. For nstance, the upper bound U-value was the U-value requred by the local buldng regulaton and the lower bound was set by takng nto account the techncal and economc feasblty of the envelope constructon. Table 2 Input parameter varables F1 F2 DESCRIPTION MIN MAX Exteror wall nsulaton thckness [m] Exteror roof nsulaton thckness [m] Exteror wndow U-value [W/m²K] Exteror wndow Solar Heat Gan Coeffcent (SHGC) 0.1 0.25 0.16 0.28 0.5 1.7 0.3 0.9 F5 Exteror wall densty [kg/m 2 ] 230 430 F6 Slab densty [kg/m 2 ] 150 415 F7 Overhang depth [m] 0.3 1.5 F8 Insde reveal depth [m] 0 0.24 F9 People fracton radant 0.3 0.6 F10 Lghts fracton radant 0.18 0.72 F11 Effectve Leakage Area 0.5 2 Wndow openng factor 0.4 1 Wndow dscharge coeffcent 0.3 0.9 F14 Vent dscharge coeffcent 0.3 0.9 F15 Number of people per Zone 7 16 Lghtng Watts per Zone Floor Area [W/m²] Electrc equpment Watts per Zone Floor Area [W/m²] Wnd velocty profle: Exponent α, Boundary layer thckness δ [m] 5 20 5 20 α=0.10 δ=210 α=0.33 δ=460 We vared electrcal equpment gans and lghtng gans between the most effcent equpment avalable on the market (LED lghtng and new electrcal equpment) and a reference one based on the EU techncal background report on ndoor lghtng (BRE, 2011). We vared the number of people n the zone assumng two dfferent offce layouts, open space (7.14 m²/p) and sngle offce (12.5 m²/p) (SIA 2024-2006). Ths resulted n average daly total nternal gans of between 12 W/m² and 40 W/m². We vared the people fracton radant (ASHRAE, 2009) and the lghts fracton radant, takng nto account dfferent lumnare confguratons (IES, 1993). We vared the shadng overhangs, nsde reveal depth and solar heat gan coeffcents as per Table 2. Varaton n these three parameters mpacted the nternal solar heat gans. In addton, we also vared the wndow and vent openng factors and ther dscharge coeffcents (Karava P., 2004). Varaton n these parameters was based on typcal wndow performance, representatve of a range of dfferent wndow types, operaton and wnd drecton. We consdered dfferent wnd velocty profles based on the ratonale that the locaton or orentaton of the buldng mght be subject to change durng early desgn stage, whch we consdered lkely to mpact coolng loads. Table 2 lsts all the consdered varables and ther ranges. Because the prncple objectve of Strategy A s to provde space coolng, we used the coolng loads as our metrc to assess the performance. In Strategy B no coolng system was modelled, therefore our metrc of performance was the number of comfortable occuped hours. We calculated the comfort hours based on the three categores of adaptve comfort descrbed n the European Standard EN15251-2007. Clmate dependency Pror work by (Zha J., 2011) ndcated that smulated buldng performance s sgnfcantly mpacted by the use of locally measured weather, as compared to weather staton data. Zha recommends the use of local weather data when avalable, partcularly for buldngs wth hgh solar gans. We performed parametrc analyss usng three dfferent weather fles: Bolzano, Palermo and San Francsco (DOE, 2011). 10 8 6 4 2 0 apr may jun jul aug sep oct dt_bolzano dt_palermo R_SAN FRANCISCO Fgure 3 Day-nght temperature dfference and solar radaton of the three selected locatons. 600 500 400 300 200 100 0 dt_san FRANCISCO R_BOLZANO R_PALERMO
Fgure 3 compares average day-nght temperature dfferences and solar radaton n the three locatons. Bolzano represents a typcal contnental clmate wth large durnal swng and low wnd speed (see Fgure 4). Palermo represents a typcal Medterranean clmate wth hot summers and low wnd breezes (see Fgure 5). San Francsco has mld summers and hgher wnd speeds (see Fgure 6). Fgure 4 Wnd rose for the Bolzano clmate. Fgure 5 Wnd rose for the Palermo clmate. Fgure 6 Wnd rose for the San Francsco clmate. Lmtatons Our buldng model assumptons should be consdered frst before applyng these study results to real buldng natural ventlaton desgn. In EnergyPlus, thermal zones are consdered well mxed zones, assumed to have unform temperature and pressure varyng hydrostatcally. Therefore temperature stratfcaton s not modelled wthn an arflow network zone. As long as the zone celng heght s around 3 m, ths assumpton s consdered reasonable. The starwell passve stack modellng method s based on the approach used n pror studes (Hensen J., 2002), but has not been ndependently expermentally valdated by our team. The convecton model used n the analyss s not optmal for use wth passve coolng strateges because convecton coeffcents are ndrectly predefned by the adaptve convecton algorthm (Beausolel-Morrson I., 2000). As n an early desgn stage no other tools are avalable, our use of the adaptve convecton algorthm can be consdered vald. Our analyss was performed on a rectangular buldng wth a sngle orentaton, and so we were able to make use of publshed wnd pressure coeffcents. The results of the analyss can be consdered generalzable to other rectangular buldngs, but could be napproprate for alternatve buldng geometres or orentatons. Bulk arflow model does not consder the nternal space layout. Internal walls and furnture may affect the natural ventlaton performance. Desgners should consder the mpact ths may have on lmtng cross ventlaton durng the early desgn stages. Senstvty analyss method We performed a senstvty analyss usng the Elementary Effects method descrbed n (Saltell A., 2008). Ths method determnes whch nput varables have neglgble, lnear or nonlnear effect on the objectve wth a relatvely small number of samples (combnatons of nput values). Rather than testng all possble combnatons of the nput parameters, the EE method selects representatve combnatons of nput parameters to test. Groups of combnatons of nput parameters are called trajectores. Typcally the number of combnatons n each trajectory s equal to the number of test parameters plus 1. Wthn a trajectory, each combnaton of test parameters dffers from the prevous by changng only one parameter each tme. For a detaled explanaton of the parameter selecton process see (Saltell A., 2008). The EE method subdvdes the varable ranges of each nput parameter nto equal ntervals of equal sze. The boundares of these ntervals are called levels. For our parametrc analyss, we used four levels (three ntervals), based on work by Saltell that showed the four levels have equal probablty of beng selected. When applyng the EE method, the dscrete probablty dstrbuton for each factor can be user defned. We have selected unform dstrbutons for all nput parameters, because we consder each of the values to be equally lkely n an early desgn stage phase.
When two combnatons of sequental nput parameters wthn a trajectory are smulated, only one varable wll dffer between them; the dfference n the output between these two runs s used to calculate a metrc called the elementary effect (EE). The elementary effect (EE) s defned for the th parameter, on the k th nput analysed as: EE y x,...,x 1 1, x Δ, x Δ 1,...,x k y(x) where X, = 1,...k represents the nput parameters and Y the output results for the selected parameters combnaton. p s the number of levels and Δ s equal to [p/ 2 (p-1)]. The EE of a factor depends also on the values of the other factors. The senstvty measures proposed by (Saltell A., 2008) are the mean (Eq. 2), the standard devaton (Eq. 3) and the mean of the absolute values of the elementary effects (Eq. 4). t1 (1) r 1 μ EE, t (2) r r 1 2 σ EE μ, t (3) r 1 t1 r * 1 μ EE, t (4) r t1 μ* quantfes the nfluence of the factor on the objectve. Parameter rankng s based on these values. Hgh values of σ demonstrate that the factor nteracts wth other varables and has a nonlnear effect on the objectve. Low values of μ, assocated wth hgh μ* values, ndcates that there s no drect correlaton between ths nput value and the output value. Whether ths factor has a negatve or postve mpact on the output depends on value of the other sgnfcant factors. If μ s equal to μ*, an ncrease of the factor corresponds to an ncrease n the output. If μ s equal to μ* n magntude but they have opposte sgns, an ncrease of the factor corresponds to a decrease n the output. We wrote parameter selecton code n Python that generates 200 trajectores consstng of 19 ponts (the number of nput factors plus one), draws 10 trajectores from the 200 avalable 500 tmes and selects the 10 that are farthest among each other. Ths procedure guarantees a good exploraton of the whole desgn space. Then, the algorthm computes 10 EEs for each nput parameters. Fnally, the senstvty statstcs are calculated. We run smulatons wth the selected parameter combnatons by means of jeplus (Dr Y Zhang, 2011), an EnergyPlus shell for parametrc studes. The frst elementary effects analyss nvolved 1440 smulatons and showed that the solar and nternal gans are the most nfluental parameters. Therefore, we performed a second elementary effects analyss (840 smulatons) fxng nternal and solar gans to the lowest reasonable amount to better determne the nfluence of the other parameters. Table 3 lsts the values that have been fxed for ths second round of parametrc analyss. Table 3 Input parameter values that have been fxed for the second analyss DESCRIPTION Exteror Wndow Solar Heat Gan Coeffcent VALUE 0.6 F7 Overhang depth [m] 0.5 F15 Number of people per Zone 10 Lghtng Watts per Zone Floor Area [W/m²] Electrc equpment Watts per Zone Floor Area [W/m²] RESULTS AND DISCUSSION The graphs n Fgure 7 to Fgure 10 compare the statstcal ndcators computed by the frst Elementary Effect analyss. The nfluence percentages are based on the μ* results. The graphs show smlar tendences n all the consdered clmate condtons and ventlaton strateges. The parameters affectng solar and nternal gans (, F15-) have a domnant nfluence on natural ventlaton performance. In Fgure 8 and Fgure 10, the hgh σ values for parameters F15- for the San Francsco weather suggests that the mpact the nternal loads have on comfort s strongly nfluenced by the selected values of the other (non-load) parameters. The postve μ means that an ncrease of nternal gans mproves comfort condtons for our Strategy B scenaro because of the lower outdoor temperatures n San Francsco. On the contrary, n Palermo comfort can be mproved by decreasng both nternal and solar gans. From Fgure 9 we see that the wndow dscharge coeffcent () for Palermo has a notably hgher relatve mpact on the number of comfort hours, compared to the other two clmates. We theorze that ths s because the forces that drve natural ventlaton (wnd speed, ndoor/outdoor temperature dfference) are smaller n Palermo wth moderate weather and low average wnd speeds. 5 5
F1 F2 F5 F6 F7 F8 F9 F10 F11 F14 F15 Influence on nr of comfort hours Standard devaton (σ) F1 F2 F5 F6 F7 F8 F9 F10 F11 F14 F15 Influence on coolng need Standard devaton (σ) 40% 35% 30% 25% 20% 15% 10% 5% 0% Factors Fgure 7 I Elementary Effect analyss results: nfluence on coolng need for Strategy A. 20 F7 F7 F7 10 F15 F15 0-10 0 10 20 30 Mean (μ) Fgure 8 I Elementary Effect analyss results: mean and standard devaton of coolng need for Strategy A. 25% 20% 15% 10% 5% 0% Factors Fgure 9 I Elementary Effect analyss results: nfluence on number of comfort hours for Strategy B. 80 70 60 50 40 30 F14 F7 F7 0-600 -400-200 0 200 400 Mean (μ) Fgure 10 I Elementary Effect analyss results: mean and standard devaton of comfort hours number for Strategy B. The graphs n Fgure 11 to Fgure 14 compare the statstc ndcators computed by the second round Elementary Effect analyss, where parameters affectng nternal and solar gans are fxed. Results show more evdent clmate dependences. Wnd velocty profle affects natural ventlaton performance for 50% n San Francsco weather. Ths s because both wnd speed and frequences are hgher n San Francsco rather than n Bolzano or Palermo. In wndy locatons wnd velocty profles parameters have to be carefully estmated dependng also on buldng surroundng area. Comparng Fgure 11 to Fgure 13, t s evdent that n Palermo clmate wndow dscharge coeffcent () and openng factors () have more effect on cross ventlaton performance than on passve nght coolng. In the Bolzano clmate, nght coolng performance s more dependent on envelope characterstcs (F1 F8) and on fracton radant of nternal gans (F9 - F10). An ncrease of the exteror wall densty (F5) translates to an ncrease of the thermal mass of the buldng, whch n turn mproves nght coolng performance because of Bolzano s large durnal temperature swngs. The effect of exteror roof densty (F6) s not evdent n the data, because t affects only zone temperatures n the upper floors of the buldng. From Fgure 11 and 12 we see that the Effectve Leakage Area (F11) has more nfluence on nght coolng performances n Bolzano and San Francsco than n Palermo. Both n Bolzano and San Francsco μ s equal to μ* for F11, but they have opposte sgn. Ths means that ncreasng the Effectve Leakage Area wll decrease the coolng needs n these two ctes. In the Palermo clmate, low values of μ are assocated wth hgh μ* values. Ths means that there s not a drect correlaton between ELA and the coolng need. Coolng need can ether ncrease or decrease wth ELA, dependng on the values of our other nput parameters. 70000 60000 50000 40000 30000 20000 10000
Standard devaton (σ) Influence on coolng need Standard devaton (σ) Influence on nr of comfort hours The postve mean value of, and for Palermo n Fgure 14 show how wndow U-value, openng factor and dscharge coeffcent has a greater effect on cross ventlaton performance than on nght coolng performance n that clmate. Fgure 12 show that n all the analysed clmates, an ncrease of wall nsulaton thckness (F1) would cause an ncrease of coolng need for Strategy A. Fgure 12 and 14 show that, n the San Francsco case, ncreasng wndow and vent dscharge coeffcent ( F14) wll decrease thermal comfort performances but reduce coolng need. We propose that s lkely because of the lower outdoor temperatures and the hgh wnd speed n the cty. When wndows are open durng the day, the ndoor temperature cools down quckly below the comfort temperature level. In the Bolzano clmate, these parameters have less mpact on coolng need and agan whether they ncrease or decrease coolng need depends on the value of the other parameters. Fgure 12 hghlghts the hgher parameter nteracton (hgher standard devatons) n the Bolzano weather case compared to the other two clmate types. 50% 40% 30% 20% 10% 0% F1 F2 F5 F6 F8 F9 F10 F11 F14 Factors Fgure 11 II Elementary Effect analyss results: nfluence on coolng need for Strategy A. San Francsco Bolzano Palermo -10-8 -6-4 -2 0 2 Mean (μ) Fgure 12 II Elementary Effect analyss results: mean and standard devaton of coolng need for Strategy A. 12 10 F2 F6 F5 F9 F9 F1 2 F11 F10 F6 F2 F1 F9 0 F1 8 6 4 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Fgure 13 II Elementary Effect analyss results: nfluence on number of comfort hours for Strategy B. Fgure 14 II Elementary Effect analyss results: mean and standard devaton of comfort hours number for Strategy B. Wnd velocty profle nfluence seems drectly correlated to how wndy the locaton s. The fact that these results were hghly dependent on the weather data used hghlghts the mportance of a relable weather fle even n early desgn stages. CONCLUSION F1 F2 F5 F6 F8 F9 F10 F11 F14 Factors 8000 6000 F6 F14 F8 4000 2000 0 0-300 -200-100 0 100 200 300 400 Mean (μ) San Francsco Bolzano Palermo The results presented n ths paper underlne the most mportant parameters and ther effect on nght coolng performances and natural ventlaton strateges n three dfferent clmate types. Our senstvty analyss showed a domnant nfluence of parameters affectng solar and nternal gans on natural ventlaton performance n all the consdered clmates. An accurate assessment of those parameters could reduce sgnfcantly the results uncertanty. Smulaton results showed that arflow network parameters had the most sgnfcant mpact on thermal comfort; these nclude the wndow openng factors and dscharge coeffcents. Arflow parameters nfluence ventlaton flow rates whch drectly mpact comfort due by lowng ndoor ar temperatures. In case of hgh wnd speeds, wndow openng factors and dscharge coeffcents are less sgnfcant than wnd pressure coeffcents. 14000 12000 10000
In clmates wth large durnal swngs n temperature, envelope characterstcs have greater mpact on both thermal comfort and coolng needs. The graphs provde quanttatve measures of the parameters effect on natural ventlaton performances and can be used to support natural ventlaton desgn usng buldng smulaton. Generalzng these results to other buldngs should be lmted to non-resdental buldngs wth regular geometry n smlar clmate types and wnd condtons. ACKNOWLEDGEMENTS The author would lke to thank the jeplus developers for the effcent support provded. A specal thanks goes to Sergo Tarantno for hs knd support n smulaton runnng. REFERENCES ASHRAE Handbook Fundamentals 2009. Amercan Socety of Heatng, Refrgeratng and Ar-Condtonng Engneers. Beausolel-Morrson I 2000. The adaptve couplng of heat and ar flow modelng wthn dynamc whole-buldng smulatons, PhD Thess. Unversty of Strathclyde, Glasgow. BRE 2001. Green Publc Procurement - Indoor lghtng - Brussels : European Commsson, DG EnvronmentC1, BU 9. Breesch H., Janssens A. 2010. Performance evaluaton of passve coolng n offce buldngs based on uncertanty and senstvty analyss, Energy and Buldngs. De Wt S. 2001. Uncertanty n predctons of thermal comfort n buldngs, PhD Thess. Techncal Unversty of Delft. DOE EnergyPlus Energy Smulaton Software: Weather Data - U.S. Department of Energy, 2011. [Onlne] //http://apps1.eere.energy.gov/bu ldngs/energyplus/weatherdata_about.cfm?cfid =5007126&CFTOKEN=a43f11bcc0665041- ACC2CE53-5056-BC19-15947478BCC4B5AC Domınguez-Munoz F. Cejudo-Lopez J.M., Carrllo- Andres A. 2010. Uncertanty n peak coolng load calculatons, Energy & Buldngs. Dr Y Zhang jeplus An EnergyPlus batch shell for parametrc studes 2012. [Onlne] // http://www.esd.dmu.ac.uk/~yzhang/jeplus/docs _html/users%20manual%20ver1.1.html. Hensen J. Bartak M., Drkal F. 2002. Modelng and smulaton of a double-skn facade system, ASHRAE transactons - Vol. 108. Hopfe C. 2009. Uncertanty and senstvty analyss n buldng performance smulaton for decson support and desgn optmzaton, PhD thess. Technsche Unverstet Endhoven IES 1993. Illumnatng Engneerng Socety of North Amerca, Lghtng Handbook: Reference & Applcaton - New York : 8th edton, 1993. Karava P. Stathopoulos T., Athents A.K. 2004. Wnd Drven Flow through Openngs A Revew of Dscharge Coeffcents, Internatonal Journal of ventlaton, Vol. 3. Lddament M.W. 1986. Ar nfltraton calculaton technques, AIVC. Saltell A. Ratto M., Andres T., Campolongo F., Carbon J., Gatell D., Sasana M., Tarantola S. 2008. Global senstvty analyss. The prmer, John Wley & Sons, Ltd. TC ASHRAE ASHRAE Handbook Fundamentals 2001. Equaton 33-34 - Vol. SI edton. Zha J. Krart M., Johnson M.H. 2011. Assess and mplement natural and hybrd ventlaton models n whole-buldng energy smulatons, Energy and Buldngs, TRP-1456.