Assessment of thermal comfort and energy savings in a field study on adaptive comfort with application for mixed mode offices

Assessment of thermal comfort and energy savings in 1 a field study on adaptive comfort with application for 2 mixed mode offices 3 4 Abstract 5 The study of the thermal comfort of the occupants of a building represents an important challenge, due to 6 its close relation with energy efficiency. Facing the application of set-point temperatures, the adaptive 7 comfort model proposes the linking of the comfort temperature to the outdoor temperature which would 8 potentially reduce the use of the HVAC system. Although there are studies that propose experimental 9 adaptive models, few verify their effectiveness. In the current study an adaptive comfort algorithm for 10 hybrid buildings is experimentally validated based on a 17-month field study in office buildings in Spain. 11 The implementation of the algorithm in the HVAC control system, both during the cooling and the 12 heating period, allowed for the evaluation of the energy consumption, obtaining savings of 27.5% and 13 11.4% respectively. The percentage of thermal sensation votes in comfort evolved from 94% (prior to 14 implementing the comfort algorithm) to 87.5% (once implemented) for the summer season and from 15 79.5% to 81.6% for the winter season. The results demonstrate that the adaptive model is effective for the 16 optimization of HVAC systems, and that it is possible to achieve energy savings without impairing the 17 comfort of its occupants for the type of climate and buildings considered. 18 Keywords: thermal comfort; mixed mode; energy savings; baseline 19 1. Introduction 20 The thermal comfort (TC)-energy efficiency (EE) dilemma represents a major challenge in the operation 21 and management of buildings [1]. In the area of thermal comfort, many field studies have been carried out 22 in buildings of different nature and in zones with different climate [2]. Firstly, most of them were based 23 on the Predicted Mean Vote until numerous subsequent studies demonstrated that this steady state 24 approach failed as it was discussed in Peeters, L. et al. [3] based on several references that support such 25 sentence. People, by nature, are able to adapt to the changing conditions of the thermal environment, 26 which forms the basis of the adaptive thermal comfort approach: if a change occurs such as to produce 27 discomfort, people react in ways which tend to restore their comfort. The indoor temperature where most 28 people are comfortable is known as neutral or comfort temperature. Nicol and Humphreys [4] proposed 29 that this temperature was closely correlated with the outdoor temperature and they also suggested that an 30 algorithm could be defined to determine the optimum indoor temperature as a linear function of the mean 31 outdoor temperature for free running (FR) buildings. 32 The results of field studies on adaptive models have important implications for energy consumption [2]. 33 Applying the adaptive comfort temperature as a room temperature set-point potentially reduces the use of 34 the HVAC system. The savings are mainly due to the acceptance of higher indoor temperatures than those 35 recommended in summer periods and lower than those recommended in winter periods [5]. In situations 36 or places where the use of the HVAC system is unavoidable, a greater acceptability range of the thermal 37 environment will lead to lower energy consumption [6,7]. 38 In the area of energy efficiency, the application of energy conservation measures (ECMs) in buildings can 39 lead to a substantial reduction in energy consumption [8,9] and although to date, several studies have 40 faced the implementation of energy conservation measures, most of these measures do not focus on the 41 set-point temperature of HVAC systems and cover other aspects [2,10,11]. 42 Therefore, it is common for TC-related problems to be addressed separately to EE-related problems, so 43 that, in the literature, studies can be found in the two fields but there is a smaller amount of research 44 integrating both concepts. Although in [12], both objectives were simultaneously contemplated obtaining 45 an experimental adaptive comfort algorithm (ACA) and quantifying up to 30% of energy savings without 46 impairing the comfort of the occupants, few follow-up studies have followed this approach. Those studies 47 which are based on adaptive comfort approach, in the case of implementing the obtained algorithm, opt 48 for either the verification of the energy saving [13,14] or the verification of the comfort of the users [15]. 49 There are also studies that quantify the energy savings based on the change of the set-point temperature in 50 a static way [16,17] and there are a large number of references based on modelling and simulation in 51 which it is necessary to emphasize that the own methods used implicitly carries some uncertainty since it 52 is not possible to incorporate all the factors that are contemplated in the field studies and that can affect 53 the obtained results [5,18,19]. 54 Another issue to be highlighted in relation to energy efficiency is the quantification of energy savings. 55 Initially one of the major obstacles was the lack of basis to establish a baseline energy consumption 56 (hereinafter referred to as baseline), which is necessary to determine improvements in energy efficiency. 57 Due to the importance of quantifying the savings, numerous efforts have been made to develop standard 58 protocols to verify it [20]. In 1997 the Efficiency Valuation Organization (EVO) published the 59 International Performance Measurement and Verification Protocol (IPMVP) [21] where standardized 60 methods for the measurement and verification of savings were developed and in 2002 ASHRAE 61 published its guidelines for energy measurement and saving demand [22]. Both organizations are 62 considered to be the main international benchmarks for measuring and verifying energy savings [23]. 63 One of the fundamental aspects that they face is the standardization of the evaluation and verification of 64 savings based on the basis that these can’t be measured directly, since the savings represent a decrease 65 compared to a previous situation that does not occur simultaneously. The savings should therefore be 66 determined by comparing the measured consumption after the implementation of an improvement 67 (verification period) and the prevision of the energy consumption of the baseline (obtained in the model 68 period) before it [24,25]. Likewise, Reichl and Kollmann [26] highlighted the need for a formalization of 69 the baseline development process. 70 Several approaches to determine baseline consumption can now be found in the literature [25,27,28]. 71 Although one of the most commonly used approaches for baseline determination is the direct application 72 of regression methods [25], it implies the need for a subsequent calibration and given that the uncertainty 73 of the savings measured using such method is a widely discussed issue today [29,30] the model proposed 74 in the current field study is based on the application of transfer functions. There are several reasons that 75 justify the use of transfer function models versus generalized linear models. The relationship is almost 76 instantaneous and established a priori. Also, the input variables influence the output variables but not vice 77 versa and it is not necessary to calibrate the model afterwards. Although currently models based on 78 transfer functions are used in all scientific fields to evaluate dynamic responses, the first precedent of this 79 method is in the field of construction [31]. 80 Due to the fact that an important development in EE-TC research should be the quantification of energy 81 savings and the evaluation of the real impact on occupants’ satisfaction, the objective of the current study 82 is empirically verifying that the adaptive control algorithms lead to equivalent comfort conditions with 83 reduction in energy consumption comparing with a fixed set point temperature. 84 The following sections detail the proposed methodology for it, based on adaptive thermal comfort 85 approach, including the analysis of the results (both in terms of thermal comfort and energy consumption) 86 and the main conclusions. 87 2. Methodology 88 Four main phases can be identified in the proposed methodology. Phase 1 involves a field study on 89 thermal comfort and the methods to obtain an experimental adaptive comfort algorithm. Phase 2 of 90 integration comprise the implementation of the ACA in the control system of a building. Phase 3 involves 91 the measurements of the energy savings. Phase 4 involves the validation of the ACA in terms of thermal 92 comfort and energy savings. Although the attainment of the experimental adaptive comfort algorithm is 93 not developed in the present article (phase 1), the essential information related to it is shown in order to 94 easily analyse and understand all the results. 95 2.1 Field study (phase 1) 96 2.1.1 Evaluation of thermal comfort 97 Phase 1 involves a field study based on the adaptive thermal comfort approach [12,32,33]. For this 98 purpose, 11 office spaces were selected with 54 workers in three non-residential buildings and located in 99 the Southwest region of Spain, in Seville (37°N, 5°W). Most of the rooms had similar dimensions and 100 they were homogeneous in terms of occupancy and use. All the spaces dated from the 90’s, had double 101 glass windows and blinds that could be opened and closed manually with NE/SE, NE and E orientation. 102 Two of the three buildings were classified as mixed mode or hybrid (MM) buildings and operated by 103 switching from the natural ventilation mode to the conditioned/heated mode [11]. 104 The climate in Seville is characterized by variable rainfall, dry and very hot summers and mild and humid 105 winters. Table 1 shows the mean and standard deviation of the average monthly outdoor temperature 106 (𝑇𝑚𝑒𝑎𝑛), the mean and standard deviation of the maximum monthly outdoor temperature (𝑇𝑚𝑎𝑥) and the 107 mean and standard deviation of the average monthly outdoor relative humidity (𝐻𝑅𝑚𝑒𝑎𝑛) during 2016. 108 Table 1. Outdoor climate 109 Variables Jan Feb. March. Apr. May June July Aug. Sept. Oct. Nov. Dec. 𝑇𝑚𝑒𝑎𝑛(°𝐶) m 12.7 12.9 13.0 16.5 19.8 25.8 29.4 29.5 25.8 21.3 14.3 12.8 s.d 2.1 2.6 1.8 2.3 2.7 2.6 1.6 1.3 3.5 2.3 2.8 1.9 𝑇𝑚𝑎𝑥(°𝐶) m 16.9 17.8 19.8 22.2 25.8 33.7 37.3 37.5 33.5 27.1 19.2 17.3 s.d 1.9 2.6 2.0 3.1 4.6 3.5 2.8 1.8 4.6 3.7 3.3 1.9 𝐻𝑅𝑚𝑒𝑎𝑛(%) m 81.4 70.2 65.1 66.7 58.9 43.0 41.8 41.5 48.7 66.6 74.1 80.9 s.d 8.6 12.2 11.2 12.0 18.6 10.3 10.8 8.2 11.7 11.6 11.7 6.7 The complete field study was carried out for one year and five months, from October 2015 to March 110 2017. About 6.376 thermal sensation data, more than 1.000.000 data sets of sensors and almost 100.000 111 sets of system operating HVAC data were collected during that time. To measure the indoor thermal 112 environment experienced by occupants, standalone data loggers were used (including air temperature, 113 relative humidity, globe temperature, air velocity, surface temperature, C02 concentration and luminosity) 114 at 15 minutes intervals throughout the whole monitoring period (17 months). The instruments were placed 115 on users' desks or as close to them as possible and away from external heat sources. Outdoor temperature 116 was measured by a weather station available in the buildings. 117 In order to evaluate the thermal comfort of the occupants, a longitudinal survey and two additional 118 weekly questionnaires were elaborated. For the thermal sensation vote (TSV) two thermal sensation 119 scales were used based on ASHRAE scale. Additionally, the Nicol scale of five points of thermal 120 preference (TP) was used as well as a binary scale to evaluate the acceptability of the indoor environment. 121 Table 2 shows the scales used in the field study as well as the implementation period of each one. 122 Table 2. Scales in the field study 123 Scales Period Thermal sensation_I warm (w), slightly warm (sw), neutral (n), slightly cool (sc), cool (c) 12 months (October 2015-October 2016) Thermal sensation_II hot (h), warm (w), slightly warm (sw), neutral (n), slightly cool (sc), cool (c), cold (cd) 5 months (October 2016-March 2017) Thermal preference (Nicol scale) much cooler (mc), a bit cooler (bc), no change (nc), a bit warmer (bw), much warmer (mw) 17 months (October 2015-March 2017) Acceptability (binary scale) unacceptable (uc), acceptable (ac) 17 months (October 2015-March 2017) The Griffith method was used to calculate the comfort temperature or neutral temperature proposed by 124 EN15251 [34] and for evaluating the relationship between the comfort indoor temperature and the 125 outdoor, the running mean temperature (𝑇𝑟𝑚) was used. Eq. (1) represents the relationship between the 126 comfort temperature and the running mean temperature in the form of an adaptive comfort algorithm 127 based on the experimental data. For the better understanding of the current study, the procedure for 128 obtaining the proposed ACA for hybrid buildings is not explained in detail and can be consulted in [35]. 129 𝑇𝑐𝑜𝑚𝑓𝑜𝑟𝑡_𝑀𝑀 =0.24 𝑇𝑟𝑚 +19.3 (n=3739, R2 =0.41, p<0.001) (1) 130 2.1.2 Evaluation of energy consumption 131 In order to evaluate the sensitive heat capacity delivered by the HVAC system, Eq. (2) was used. 132 𝑄󰇗=𝑚󰇗 𝐶𝑝 𝑎𝑖𝑟 |∆𝑇| (2) 133 Where 𝑄󰇗 is the sensitive heat capacity delivered, 𝑚󰇗 the airflow of the fan-coil, 𝐶𝑝 𝑎𝑖𝑟 the specific heat of 134 air and |∆𝑇| the drop of the air temperature of the fan-coil, that is to say, its difference in air temperature. 135 The drop of the air temperature was measured by two sensors installed for this purpose, one in the air 136 inlet and the other one in the air outlet of each of the fan coils in the rooms under study. The air velocity 137 of each fan coils was monitored by the Building Management System and the fan coils’ technical 138 specifications were consulted to know the proportional airflow rate. 139 The energy consumption was calculated based on Eq. (3). 140 𝐶 =𝑄󰇗𝜁𝑠𝑒𝑎𝑠𝑜𝑛𝑎𝑙 ⁄ (3) 141 Where 𝐶 is the energy consumption, 𝑄󰇗 the sensitive heat capacity delivered by the HVAC system and 142 𝜁𝑠𝑒𝑎𝑠𝑜𝑛𝑎𝑙 the seasonal efficiency (𝑆𝐸𝐸𝑅 or 𝑆𝐶𝑂𝑃 for the cooling or heating period respectively). Focusing 143 the Eq. (3) for heating and cooling periods Eq. (4) was obtained. 144 𝐶𝑜𝑜𝑙𝑖𝑛𝑔: 𝐶 =𝑄󰇗𝑆𝐸𝐸𝑅 ⁄; 𝐻𝑒𝑎𝑡𝑖𝑛𝑔: 𝐶 = 𝑄󰇗𝑆𝐶𝑂𝑃 ⁄ (4) 145 Eq. (3) and Eq. (4) show that reductions in demand and consumption are proportional for the whole 146 period (both in the cooling season and in the heating season). 147 2.2 Integration (phase 2) 148 2.2.1 Building management system 149 In phase 2 the comfort algorithm obtained (Eq. (1)) was integrated in the building management system. 150 The implementation of the ACA and the verification of energy savings and thermal comfort were carried 151 out in one of the three buildings considered in phase 1, specifically one of the MM buildings where eight 152 offices were selected with a daily occupation both in the morning and in the afternoon and always 153 occupied by the same users. The choice of the building was due to the fact that there was a possibility of 154 integrating the algorithm into its control system. 155 The building had a centralized four-pipe operation HVAC system with a chiller and a heat pump that 156 provided heating and air conditioning throughout the building. All the variables related to the operation of 157 the HVAC system were centrally managed through a single platform that integrated all the data and 158 allowed the establishment of logical rules. The temperature was also centrally set by a fixed base set-point 159 for both the warm season and the cold season. 160 Each room in the building had one or more fan-coils for heating and air conditioning on demand. The fan- 161 coils were controlled by the users of each room using a thermostat located onsite. The variables under the 162 occupants' control were fan-coil (on/off) status, the driving force and the environmental temperature 163 within a range of +/- 3 degrees relative to the base set-point set in the system. 164 2.2.2 Integration of the adaptive comfort algorithm 165 The implementation of the adaptive comfort algorithm was carried out both during a winter season and 166 during a summer season. Most important milestones are shown in the Table 3. 167 168 Table 3. Milestones in the study 169 Milestone Date M.1 Beginning of the first campaign (set-point period) October 2015 M.2 Implementation of ACA-summer season (ACA period) July 2016 (14/07/2016) M.3 Beginning of the second campaign (set-point period) October 2016 M.4 Implementation of ACA-winter season (ACA period) February 2017 (06/02/2017) M.5 Ending of the field study March 2017 During the period hereinafter referred to as "set-point", the centrally established set-point was maintained, 170 while during the period hereinafter referred to as "ACA", an indoor temperature according to the adaptive 171 comfort algorithm which was obtained experimentally (Eq. (1)) was applied. The environmental variables 172 of the rooms were monitored as well as the thermal perception of the users, which was obtained through 173 surveys on their thermal satisfaction. The actual energy consumption of the fan-coils was also quantified 174 by sensors installed for this purpose. 175 The integration of the algorithm into the building system was carried out by the use of rules and logic 176 gates in the control module available in the building management system. 177 2.3 Measurement and validation (phase 3) 178 In the current study, the energy savings were evaluated once the adaptive control algorithm (ECM) was 179 implemented, being the approach for baseline determination based on the application of transfer 180 functions. 181 Eq. (5) represents the generic model for the determination of the baseline based on transfer functions [36]: 182 ( ) ( ) ( ) 10  == −+−= n i i m i ii itfditYatf (5) 183 Where 𝑓(𝑡) is the target variable, 𝑌𝑖 are the independent variables or excitations, 𝑎𝑖 are the adjustment 184 coefficients of each variable 𝑌𝑖 at the current moment and in the past and 𝑑𝑖 show the relation of the target 185 variable with the past instants, that is to say, the dynamic inertia. 186 As can be seen in all the manuals for the evaluation of building behaviour [37], the evolution of the 187 indoor temperature and the consumption are linked. In order to know one, the other is required and vice 188 versa. The proposed model, in contrast to tendencies of published material, combined two inverse 189 characterization models. 190 Based on Eq. (6), the particularized model for the characterization of the consumption and for the 191 determination of the baseline is defined in Eq. (6) [37]. 192 𝐶𝐼(𝑡)=∑𝑎𝑖 𝑐·∆𝑇(𝑡−𝑖) 𝑚 𝑖=0 +∑𝑑𝑖 𝑐·𝐶𝐼(𝑡−𝑖) 𝑛 𝑖=1 (6) 193 Where 𝐶𝐼(𝑡) is the hourly consumption. We opted for an hourly model as it is more precise than the daily 194 models. Likewise, the hourly model allowed us to adjust quite accurately the consumption during the 195 model period. 196 ∆𝑇(𝑡)= 𝑇𝐹𝑅(𝑡)−𝑇𝐼𝑁𝑇(𝑡) is the difference between average temperature of indoor air when the system is 197 off (and the building operates as a free running building) and the average temperature of the indoor air 198 when the HVAC system is on. This variable showed the effect of the HVAC system in the temperature of 199 the room. 200 𝑚 and 𝑛 depend on the inertia of the model and represent the way the current instant is affected by the 201 previous instants. 202 𝑎𝑖 𝑐 y 𝑑𝑖 𝑐 are coefficients of the consumption model unknown a priori. They were identified during the 203 model period (based on the experimental data) applying the room air weighting factor procedure of 204 ASHRAE [36-38]. 205 The periods of cooling and heating were distinguished, obtaining a different model for each of them. 206 Table 4 shows the coefficients and the 𝑅2 for both periods. 207 Table 4. Coefficients of the consumption model and the free running temperature model 208 Variables Cooling period Heating period 𝑎12 -4.5 1.5 𝑎11 7.9 -0.2 𝑎10 -2.0 -4.3 𝑎9 -2.5 4.6 𝑎8 1.5 -3.2 𝑎7 0.6 3.2 𝑎6 3.9 -3.0 𝑎5 -10.2 3.9 𝑎4 7.0 -5.2 𝑎3 -3.3 3.0 𝑎2 6.6 2.4 𝑎1 -6.0 -5.7 𝑎0 1.2 2.7 𝑑1 0.5 0.5 𝑅2 0.91 0.84 Eq. (7) (model for the free running temperature) is the particularization of Eq. (5) for temperatures. 209 𝑇𝐹𝑅(𝑡)=∑𝑎1𝑖 𝑇𝐹𝑅 ·𝑇𝐸𝑋𝑇(𝑡−𝑖) 𝑚 𝑖=0 +∑𝑎2𝑖 𝑇𝐹𝑅 ·𝑅𝐴𝐷(𝑡−𝑖) 𝑚 𝑖=𝑜 +∑𝑑𝑖 𝑇𝐹𝑅 ·𝑇𝐹𝑅(𝑡−𝑖) 𝑛 𝑖=1 + 𝐾 (7) 210 Where 𝑇𝐸𝑋𝑇 is the average outdoor temperature and 𝑅𝐴𝐷 is the average global horizontal irradiance. 211 These variables were considered model variables and allowed baseline identification during the reference 212 period. Likewise, the exploitation of the model as baseline entailed knowing both climatic conditions and 213 indoor temperatures. 214 It should be emphasized that although the effect of irradiance is usually condensed with the effect of the 215 outdoor temperature for a more balanced temperature, in Mediterranean climates the effect of irradiance 216 can be especially critical. In the current work it was equally weighted to the outdoor temperature in order 217 to obtain a more realistic model. 218 𝐾 is related to internal gains in the room and represents the increase of the temperature due to internal 219 sources. These internal sources were assumed constant and invariant so that they could be considered as a 220 fixed parameter in the reference period for obtaining the baseline since the spaces considered, being 221 tertiary buildings, presented a very stable use (there was an average period of the day in which the HVAC 222 system was in used). 223 𝑎1𝑖 𝑇𝐹𝑅, 𝑎2𝑖 𝑇𝐹𝑅 y 𝑑𝑖 𝑇𝐹𝑅 are the coefficients particularized for the free running temperature model. They were 224 calculated using the same procedure for calculating the coefficients of the consumption model. 225 In relation to the indices of both models (Eq. 6 and Eq. 7) it should be pointed out that, although 226 ASHRAE [37] suggests fixing 𝑛 = 1 and 𝑚 = 3, Ciulla et al [31] propose using higher values of 𝑚 and 𝑛 227 due to the fact that on an hourly basis the variables that affect indoor temperature and consumption (sun 228 protection status, window closure/opening) are unknown. The lack of awareness of these variables is 229 balanced out by increasing the number of previous instants considered in the model. 230 Due to the above, for both the consumption model and the free running temperature model, the values 231 𝑛 = 1 and 𝑚 = 12 were used. So that although the proposed model is based on transfer functions, such 232 increase in the indexes makes it shift away towards ARMAX models [39]. 233 3. Verification and analysis of results (phase 4) and discussion 234 The fourth phase contemplated both, the validation of the comfort level of users and the quantification of 235 energy savings, once the adaptive comfort algorithm was implemented. For this, the data of the 236 environmental variables monitoring and the thermal sensation of the occupants were analysed. In terms of 237 thermal comfort, the evolution of the thermal sensation and thermal preference data, the thermal 238 acceptability data and the evolution of the indoor temperature in the room were considered. In relation to 239 energy efficiency, the actual measures once the ACA was implemented were considered in comparison 240 with the values predicted by the baseline. 241 As the comfort algorithm was implemented during the summer season and during the winter season, both 242 comfort and energy performance are shown independently for the cooling period and for the heating 243 period. Although the baseline was referenced on an hourly basis, both the comfort level and the savings 244 6.376 during the 17 months in which the field study was carried out. The following main conclusions can 362 be drawn: 363 - A unique adaptive comfort algorithm was previously obtained for MM buildings located in an area 364 characterized by mild winters and very hot summers. 365 -The acceptability of the occupants in terms of percentage of thermal sensation votes was empirically 366 verified for a MM office building regarding the integration of the ACA in its HVAC system. During the 367 heating period it rose slightly from 79.5% to 81.6%, being statistically equal the percentage of votes in 368 comfort. During the cooling period, it decreased slightly from 94% to 87.5%, being a slightly difference 369 statistically significant (p<0.001). It can be concluded that such percentage remained similar values 370 before and after the inclusion of the ACA in the HVAC system. 371 -The energy savings were quantified for the same MM office building by comparing the values predicted 372 by the baseline and the real values measured once the ACA was included in the HVAC system. The 373 results showed an energy savings of 27.5% during the cooling period and 11.4% during the heating 374 period. It was slightly higher for the cooling period but energy savings were also identified during the 375 heating season. 376 -The results highlight the validity of an adaptive comfort algorithm for MM office buildings and show 377 that it improves the HVAC system in terms of energy saving, while maintaining the comfort of the 378 occupants. The proposed algorithm was validated in terms of comfort level and energy savings 379 considering the experiment conditions (local climate, workers and analysed buildings), so it would be 380 advisable to confirm such results taking into account a larger number of buildings and a wider sample of 381 workers. Nevertheless it is also important to highlight that the obtained results are promising and similar 382 achievements are expected even considering another type of building and climate. 383 Acknowledgements 384 The authors would like to thank all the participants in the field study as well as the offices management 385 for allowing access to their workspace and participating in the surveys. This research was supported 386 financially by a Grant-in-Aid (TEP-7247) from the “Excellence Projects” Program of the Andalusia 387 Government. 388 References 389 [1] R. Becker, I. Goldberger, M. Paciuk, Improving energy performance of school buildings while 390 ensuring indoor air quality ventilation, Build. Environ. 42 (2007) 3261–3276. 391 doi:10.1016/j.buildenv.2006.08.016. 392 [2] M. Pellegrino, M. Simonetti, G. Chiesa, Reducing thermal discomfort and energy consumption of 393 Indian residential buildings: Model validation by in-field measurements and simulation of low- 394 cost interventions, Energy Build. 113 (2016) 145–158. doi:10.1016/j.enbuild.2015.12.015. 395 [3] L. Peeters, R. de Dear, J. Hensen, W. D’haeseleer, Thermal comfort in residential buildings: 396 Comfort values and scales for building energy simulation, Appl. Energy. 86 (2009) 772–780. 397 doi:10.1016/j.apenergy.2008.07.011. 398 [4] F. Nicol, M. Humphreys, Derivation of the adaptive equations for thermal comfort in free-running 399 buildings in European standard EN15251, Build. Environ. 45 (2010) 11–17. 400 doi:10.1016/j.buildenv.2008.12.013. 401 [5] P.T. Bhaskoro, S.I.U.H. Gilani, M.S. Aris, Simulation of energy saving potential of a centralized 402 HVAC system in an academic building using adaptive cooling technique, Energy Convers. 403 Manag. 75 (2013) 617–628. doi:10.1016/j.enconman.2013.06.054. 404 [6] L. Yang, H. Yan, J.C. Lam, Thermal comfort and building energy consumption implications - A 405 review, Appl. Energy. 115 (2014) 164–173. doi:10.1016/j.apenergy.2013.10.062. 406 [7] L. Yang, W. Zheng, Y. Mao, J.C. Lam, Y. Zhai, Thermal Adaptive Models in Built Environment 407 and Its Energy Implications in Eastern China, Energy Procedia. 75 (2015) 1413–1418. 408 doi:10.1016/j.egypro.2015.07.237. 409 [8] ISO-14040 Environmental Management – Life cycle assessment – Principles and framework, 410 (2006). 411 [9] ISO-14044 Environmental Management – Life cycle assessment – Requirements and guidelines, 412 (2006). 413 [10] D. Rowe, A Study of a Mixed Mode Environment in 25 Cellular Offices at the University of 414 Sydney, Int. J. Vent. 1 (2003) 53–64. doi:10.1080/14733315.2003.11683644. 415 [11] J.C. Salcido, A.A. Raheem, R.R.A. Issa, From simulation to monitoring: Evaluating the potential 416 of mixed-mode ventilation (MMV) systems for integrating natural ventilation in office buildings 417 through a comprehensive literature review, Energy Build. 127 (2016) 1008–1018. 418 doi:10.1016/j.enbuild.2016.06.054. 419 [12] K.J. McCartney, F.J. Nicol, Developing an adaptive control algorithm for Europe, Energy Build. 420 34 (2002) 623–635. doi:10.1016/S0378-7788(02)00013-0. 421 [13] F. Nicol, S. Roaf, Pioneering new indoor temperature standards: The Pakistan project, Energy 422 Build. 23 (1996) 169–174. doi:10.1016/0378-7788(95)00941-8. 423 [14] K.W.H. Mui, W.T.D. Chan, Adaptive comfort temperature model of air-conditioned building in 424 Hong Kong, Build. Environ. 38 (2003) 837–852. doi:10.1016/S0360-1323(03)00020-9. 425 [15] G.Y. Yun, J.H. Lee, K. Steemers, Extending the applicability of the adaptive comfort model to 426 the control of air-conditioning systems, Build. Environ. 105 (2016) 13–23. 427 doi:10.1016/j.buildenv.2016.05.027. 428 [16] R. Zmeureanu, A. Doramajian, Thermally acceptable temperature drifts can reduce the energy 429 consumption for cooling in office buildings, Build. Environ. 27 (1992) 469–481. 430 doi:10.1016/0360-1323(92)90045-Q. 431 [17] A.C. Roussac, J. Steinfeld, R.J. de Dear, A preliminary evaluation of two strategies for raising 432 indoor air temperature setpoints in office buildings, Archit. Sci. Rev. 54 (2011) 148–156. 433 doi:10.1080/00038628.2011.582390. 434 [18] A. Kusiak, G. Xu, Z. Zhang, Minimization of energy consumption in HVAC systems with data- 435 driven models and an interior-point method, Energy Convers. Manag. 85 (2014) 146–153. 436 doi:10.1016/j.enconman.2014.05.053. 437 [19] S. Ezzeldin, S.J. Rees, The potential for office buildings with mixed-mode ventilation and low 438 energy cooling systems in arid climates, Energy Build. 65 (2013) 368–381. 439 doi:10.1016/j.enbuild.2013.06.004. 440 [20] J. Kelly Kissock, C. Eger, Measuring industrial energy savings, Appl. Energy. 85 (2008) 347– 441 361. doi:10.1016/j.apenergy.2007.06.020. 442 [21] Efficiency Valuation Organization (EVO), Concepts and options for determining energy and 443 water savings. International performance measurement and verification protocol, 1 (2012). 444 [22] ASHRAE, American Society of Heating, Refrigeration and Air Conditioning Engineers, 445 Measurement of Energy and Demand Savings, Guideline, Atlanta, Georgia, 2002. 446 [23] F. Rossi, D. Velázquez, I. Monedero, F. Biscarri, Artificial neural networks and physical 447 modeling for determination of baseline consumption of CHP plants, Expert Syst. Appl. 41 (2014) 448 4658–4669. doi:10.1016/j.eswa.2014.02.001. 449 [24] F. Rossi, D. Velázquez, A methodology for energy savings verification in industry with 450 application for a CHP (combined heat and power) plant, Energy. 89 (2015) 528–544. 451 doi:10.1016/j.energy.2015.06.016. 452 [25] A.Z. Dalgleish, L.J. Grobler, Measurement and verification of a motor sequencing controller on a 453 conveyor belt, Energy. 28 (2003) 913–927. doi:10.1016/S0360-5442(03)00034-3. 454 [26] J. Reichl, A. Kollmann, The baseline in bottom-up energy efficiency and saving calculations - A 455 concept for its formalisation and a discussion of relevant options, Appl. Energy. 88 (2011) 422– 456 431. doi:10.1016/j.apenergy.2010.03.008. 457 [27] D. Velázquez, R. González-Falcón, L. Pérez-Lombard, L. Marina Gallego, I. Monedero, F. 458 Biscarri, Development of an energy management system for a naphtha reforming plant: A data 459 mining approach, Energy Convers. Manag. 67 (2013) 217–225. 460 doi:10.1016/j.enconman.2012.11.016. 461 [28] S. Motlaghi, F. Jalali, M.N. Ahmadabadi, An expert system design for a crude oil distillation 462 column with the neural networks model and the process optimization using genetic algorithm 463 framework, Expert Syst. Appl. 35 (2008) 1540–1545. doi:10.1016/j.eswa.2007.08.105. 464 [29] C.D. Kissock K, Agami T, Fletcher D, Effect of short data periods on the annual prediction 465 accuracy of temperature-dependent regression models of commercial building energy use, Int. 466 Sol. Eng. Conf., (1993) 455–463. 467 [30] T.A. Reddy, J.K. Kissock, D.K. Ruch, Uncertainty in Baseline Regression Modeling and in 468 Determination of Retrofit Savings, J. Sol. Energy Eng. 120 (1998) 185–192. 469 doi:doi:10.1115/1.2888068. 470 [31] G. Ciulla, V. Lo Brano, A. Orioli, A criterion for the assessment of the reliability of ASHRAE 471 conduction transfer function coefficients, Energy Build. 42 (2010) 1426–1436. 472 doi:10.1016/j.enbuild.2010.03.012. 473 [32] I. Fato, F. Martellotta, C. Chiancarella, Thermal Comfort in the Climatic Conditions of Southern 474 Italy., ASHRAE Trans. 110 PART I (2004) 578–593. 475 [33] M. Indraganti, R. Ooka, H.B. Rijal, Thermal comfort in offices in summer: Findings from a field 476 study under the “setsuden” conditions in Tokyo, Japan, Build. Environ. 61 (2013) 114–132. 477 doi:10.1016/j.buildenv.2012.12.008. 478 [34] CEN. Indoor environmental input parameters for design and assessment of energy performance of 479 buildings: addressing indoor air quality, thermal environment, lighting and acoustics. EN 15251. 480 Brussels: European Committee for Standardization; 2007. 481 [35] E. Barbadilla-Martín, J.M. Salmerón Lissén, J. Guadix Martín, P. Aparicio-Ruiz, L. Brotas, Field 482 study on adaptive thermal comfort in mixed mode office buildings in southwestern area of Spain, 483 Build. Environ. 123 (2017) 163–175. doi:10.1016/j.buildenv.2017.06.042. 484 [36] D.G. Stephenson, G.P. Mitalas, Calculation of heat conduction transfer functions for multilayers 485 slabs, ASHRAE Trans. 77 (1971) 117–126. 486 [37] ASHRAE, Handbook, Fundamentals, 1997. 487 [38] T.J.R. Cooper, 1986 ASHRAE handbook: Refrigeration systems and applications: ASHRAE, 488 Inc., USA 1986, ISSN 82 64 3223, US$100, Int. J. Refrig. 10 (1987) 243. 489 [39] E. Hickey, D.G. Loomis, H. Mohammadi, Forecasting hourly electricity prices using ARMAX- 490 GARCH models: An application to MISO hubs, Energy Econ. 34 (2012) 307–315. 491 doi:10.1016/j.eneco.2011.11.011. 492 493