On the assessment of the energy performance and environmental behaviour of social housing stock for the adjustment between simulated and measured data: The case of mild winters in the Mediterranean climate of southern Europe.

1 1On the assessment of the energy performance and environmental behaviour of 2social housing stock for the adjustment between simulated and measured data: the 3case of mild winters in the Mediterranean climate of southern Europe. 4Rocío Escandón a,*, Rafael Suárez a, Juan José Sendra a 5a Instituto Universitario de Arquitectura y Ciencias de la Construcción, Escuela Técnica Superior de 6 Arquitectura, Universidad de Sevilla, Av. de Reina Mercedes 2, Seville (41012), Spain. 7Abstract 8 Current European energy policies stress the potential of housing stock retrofitting to reduce global energy 9 consumption. In order to implement efficient measures, it is essential to know its real in-use energy 10 behaviour. In southern Europe, social housing represents an important percentage of the residential stock 11 built before the implementation of the first energy regulations. However, there are few studies specifically 12 analysing their energy behaviour. Standardized use and occupancy patterns are not usually suited to 13 Mediterranean social housing, and this mostly results in estimated consumption exceeding real 14 consumption. 15 This research aims to quantify the thermal comfort and energy consumption of the social housing stock 16 during the characteristically mild winters in a Mediterranean climate, based on the monitoring of 17 representative case studies from southern Spain. The results show that the dwellings analysed are far from 18 conforming to adaptive comfort standard EN-15251 and yet, their limited local heating systems are rarely 19 turned on, reducing the expected energy consumption. In addition, real use and occupancy patterns are 20 defined for the case studies, allowing the development of energy simulation models that are better suited 21 to the real behaviour of this social housing stock. 22 Keywords: social housing stock, monitoring, thermal comfort, energy consumption, energy models 23 calibration. * Corresponding author. E-mail address: [email protected] 2 24 1. Introduction 25 Over the last few decades, European building regulations have been establishing a common framework 26 for the encouragement of energy efficiency [1, 2], promoting the retrofitting of housing stock. In the case 27 of southern Europe, residential stock built prior to the first regulations which globally limit the energy 28 demand of buildings (1976-1979) represents between 63 and 76% of the total housing stock [3, 4, 5]. 29 Most of these residential buildings do not incorporate any specific measures for thermal insulation in their 30 envelopes, and are therefore obsolete from an energetic point of view. 31 32 Social housing represents an important percentage of the residential stock in southern Europe, with 33 specific socioeconomic characteristics that entail particular needs and a use pattern different from the 34 standardized one. Field research on social housing in southern Europe [6, 7, 8, 9] shows that these 35 buildings have much lower energy consumption than that estimated by the national energy assessment 36 procedures derived from Directive 2002/91/EEC and its updates [10], based on general premises of 37 intensity and habits of use of buildings and energy, not applicable in the case of social housing. An 38 example of this is that the estimated average heating consumption of Spanish multi-family dwellings 39 (with an average area of 88.7 m2) in the Mediterranean climate is 3673 kWh per year per dwelling [11], 40 while for low income households (declared annual household income below 15,000 €) in southern Europe 41 it is around 1386 kWh per year [6]. 42 43 There is extensive research analysing the divergence between real consumption and consumption 44 estimated through simulation, either by software recognized by the scientific community or the official 45 software of the different countries [12,13,14,15]. This ‘performance gap’ has generally been attributed to 46 user behaviour and, to a lesser extent, to a poor identification of the constructive characteristics of 47 buildings [16,17]. In order to reduce the error rate due to the poor identification of the constructive 48 characteristics, it would suffice to carry out the relevant tests in the building, such as air permeability and 49 infra-red thermography tests or U-measurements [12,18]. However, the error associated to user pattern 50 definition is much more difficult to identify and reduce. 51 52 Many of the studies about occupant behaviour focus on retrofitted dwellings (low-energy) in central and 53 northern Europe (climates with severe winters), where real consumption is frequently much higher than 3 54 estimated consumption due to the Rebound effect [19, 20]. However, in many other cases, the opposite 55 occurs: real consumption is much lower than that estimated, because users spend far less time in the 56 dwellings than the time established by the standardized use patterns of the different countries, known as 57 the Prebound effect [13, 20]. 58 59 In the case of social housing, the divergence between real and estimated consumption is even greater due 60 to an additional variable: when users occupy their dwellings, do they do so in conditions of thermal 61 comfort? The standardized use patterns assume that users live in thermal comfort conditions, but in social 62 housing in southern Europe, although it belongs to a mild winter climate zone, it is usual to forgo comfort 63 and live in fuel poverty conditions [6, 7]. In addition, in the specific case of southern Spain, only 5% of 64 multi-family dwellings built before 1979 have a centralized heating system, and 13% have a local heating 65 system [3]. These data demonstrate that this housing stock does not have thermal conditioning means to 66 ensure the maintenance of thermal comfort conditions. 67 68 Due to the fuel poverty found in this social housing stock, investments in energy retrofitting (usually 69 focused on improving the insulation of its envelope) [21, 22, 23] are often ineffective from an energy 70 saving point of view with excessive payback periods. Although it is very difficult to reduce energy 71 consumption when it is so low, this does not mean that retrofitting is not beneficial, but simply means that 72 instead the final aim ought to be the improvement of thermal comfort conditions in the dwellings, where 73 the profit margin is very high [24]. 74 75 The main objective of this research is to quantify the real conditions of thermal comfort in the specific 76 context of social housing in southern Europe (mild winter climate), based on the adaptive model 77 established by standard EN-15251 [25] and energy consumption. To this end, a specific methodology 78 based on in-use monitoring has been developed and applied to three case studies representative of the 79 typology, constructive system and climate of southern Spain. Through this monitoring, real use and 80 occupancy patterns are defined in order to develop energy simulation models adjusted to the real 81 behaviour of this housing stock, significantly reducing the ‘performance gap’ between real and estimated 82 consumption. The conclusions obtained are expected to be the potential starting point for a future change 4 83 in the standardized use and occupation patterns, and the energy retrofitting policies of social housing 84 stock in the Mediterranean climate. 85 86 2. Methodology 87 In this work the methodology used has been developed based on previous research in order to further 88 examine the characterization of the current energy conditions of social housing stock in southern Europe 89 [26]. To do so, a detailed energy and environmental assessment of the dwellings was carried out through 90 monitoring and subsequent energy simulation. 91 2.1. Monitoring 92 Different monitoring methodologies have been validated by the scientific community. Two 93 methodologies are distinguished focusing on the measurement of environmental parameters [27]: spot 94 measurements taken only once or long-term measurements taken at specific intervals, from minutes to 95 hours. The type of measurement depends on the type of data required for the investigation. 96 97 In general, spot measurements are used when the aim is the evaluation of thermal comfort through 98 surveys and it is necessary to link these with the internal and external environmental conditions of that 99 point in time, or when it is not possible to carry out long-term measurements due to economic or intrusion 100 constraints. Long-term measurements are fundamental when the purpose of the monitoring is to 101 determine the behaviour of the building over a given period of time such as a season or a full year. In 102 addition, these types of measures are necessary to evaluate thermal comfort following an adaptive model, 103 one of the main aims of this research. Thus, this monitoring methodology includes long-term 104 measurements (temperature, relative humidity, CO2 level, and energy consumption) which are 105 complemented with spot tests (air permeability and infra-red thermography). 106 2.1.1. Air permeability test and Infra-red thermography 107 Depressurisation tests were carried out in each case study in order to verify the airtightness of the 108 envelope of the dwellings, according to norm UNE EN-13829 [28]. The Blower Door equipment used in 109 this test was installed in the entrance door of the dwellings and controlled from inside the residential unit. 5 110 111 This information was complemented with the capture of images of dwelling envelopes using a 112 thermographic camera, following norm UNE EN-13187 [29]. The aim of this test is to search for thermal 113 behaviour patterns of the envelopes in the buildings under study. 114 2.1.2. In situ ambient measurements 115 The in situ data collection of environmental variables is essential for the evaluation of the environmental 116 behaviour of the case studies. For this reason, this research has monitored air temperature, relative 117 humidity and CO2 levels inside the dwellings. Two WOHLER CDL 210 indoor data-loggers were placed 118 in each dwelling (one in the living room and the other in the main bedroom) to measure the variables 119 every 30 minutes for a full year. 120 121 Data provided by three meteorological stations belonging to the Spanish State Meteorological Agency 122 [30] were used to analyse external environmental variables. The variables of temperature, relative 123 humidity, solar radiation, wind speed and direction, and precipitation were measured every 30 minutes 124 over a year. 125 126 In addition to the analysis of environmental variables in winter this research also studies the level of 127 thermal comfort in the dwellings according to the adaptive model established by EN-15251 [25]. This 128 model of adaptive thermal comfort analysis focuses on the interaction between people and buildings 129 (through indoor temperature) and on their expectations (based on outdoor temperature) [31]. In order to 130 obtain the outdoor reference temperature (TeR), the daily weighted average was calculated according to 131 equation 1. For the calculation of the optimum operating temperature (Top), equation 2, established in EN- 132 15251, was used. Case studies belong to category III, associated with existing buildings, which means an 133 acceptability band of +/- 4 ºC in relation to the calculated Top. 134 (1) TeR = (1 - P- * Ted-1 + P * TeR-1 135 where: 136 TeR: running mean temperature for today 137 Ted-1: daily mean external temperature for previous day; 138 TeR-1: running mean temperature for the previous day; 139 P; is a constant between 0 and 1. Recommended to use 0.8. 6 140 (2) Top = 0.33 * TeR + 18.8 141 142 2.1.3. Energy consumption measurements 143 Another key aspect for the evaluation of the behaviour of the case studies is the in situ data collection of 144 energy variables. As is usually the case with social housing in southern Spain, none of these case studies 145 have gas-fuelled heating systems, and therefore the research focuses on electricity consumption. The 146 detailed energy consumption (sub-metering) was monitored by a general consumption meter in the 147 electrical panel and several individual meters in the sockets of some appliances (mainly local heating and 148 cooling systems). Consumption was measured in kWh every 15 minutes over a year. This helps to 149 establish relationships between comfort level, indoor and outdoor conditions, and user behaviour. In 150 addition, in order to complete and contrast the monitored consumption, historical data of general 151 electrical consumption of the case studies were compiled using past electricity bills. 152 2.1.4. User patterns 153 In order to avoid the usual divergence between real and estimated consumption (Prebound effect), 154 monitoring data was used to define user patterns. A mixed methodology was used in this research to 155 capture the technical and social aspects from user practices qualitatively and quantitatively [32]. In order 156 to explain and validate these aspects the qualitative data obtained from the user surveys and the user 157 patterns obtained from the quantitative data measured in the dwellings (temperature, CO2 concentration 158 and electric consumption) were cross-referenced. 159 160 In order to analyse the quantitative data, the hourly average values for the entire winter period were used 161 [33]: temperatures in degrees centigrade (ºC), CO2 concentration in parts per million (ppm), heating local 162 systems operation on a scale of 0 to 1 (0 means off and 1 on) and general electrical consumption in 163 kilowatt-hour (kWh). This research focused on the analysis of weekdays, which is when patterns are 164 usually repeated the most, whereas at weekends behaviour is usually more irregular. 165 166 7 167 2.2. Energy simulation 168 For the simulation of the environmental and energy behaviour of the case studies, the methodology 169 requires two initial tasks: construction of the energy model to reproduce the rooms studied and an energy 170 model adjustment process. 171 2.2.1. Energy model construction 172 The energy model was developed with software DesignBuilder (v.4.7.0.027), which uses EnergyPlus 173 [34], a simulation program recognized by the US DOE [35]. 174 175 Models recreating the whole dwelling and its boundary conditions were generated for each of the case 176 studies monitored. The climate data used for the energy simulation were provided by three meteorological 177 stations belonging to the Spanish State Meteorological Agency [30], whose data were previously 178 validated comparing them with spot measurements taken outside the case studies. The constructive 179 definition of the envelope (section 3.2) and the use and occupation patterns of the case studies (section 180 3.3) were also taken into account for the development of the model. In addition to occupation habits, the 181 models also include lighting patterns, use of local heating systems, natural ventilation (window opening), 182 and use of sun protection (shutters, awnings…). 183 184 Another aspect that helps to reduce the uncertainty of the model is the introduction of the rate of 185 infiltration measured in the air permeability test. 186 2.2.2. Energy model adjustment 187 As Cipriano et al. [36] correctly point out, the literature review shows the absence of a generally accepted 188 methodology for the calibration of energy building models. However, some authors define calibration 189 methodologies based on the manual adjustment of the models [37, 18], while others base them on a multi- 190 stage guided procedure [36, 38]. 191 This is a complex task, involving many parameters with a significant degree of uncertainty. Many of the 192 studies on calibration focus on the importance of in situ measurements as a tool for the adjustment of 193 energy simulations [36, 38]. Therefore, as described in the previous section, all the information collected 194 in situ in this research was incorporated into the energy models. However, there are standard criteria to 8 195 determine when a model can be considered to be calibrated, focusing on how the model adjusts to 196 measured data and analysing the degree of error [39, 40, 41]. 197 198 Usually, the literature about the calibration of energy model focuses on the adjustment of the heating 199 energy consumption [36]. However, this would make no sense in these case studies, since the heating 200 energy consumption is very insignificant in the social housing profile in southern Spain (the dwellings are 201 usually in conditions of free evolution). Therefore, this research focuses on the manual adjustment of 202 environmental variables during the winter period, specifically in the interior temperature (as this is the 203 variable involved in thermal comfort analysis). The results of energy simulations and monitoring were 204 compared both graphically (for a representative week) and statistically (for the entire winter period). The 205 graphs provide information about the time periods in which the two datasets diverge, while the statistical 206 indices provide a numerical index to assess how closely the simulation matches the measured data. 207 208 ASHRAE Guideline 14-2002 [39] was followed in the statistical validation of the model, establishing two 209 error indicators, the Mean Bias Error (MBE) and the Coefficient of Variation of the Root Mean Square 210 Error (CVRMSE) values, according to equations (3) and (4): 211 (3) MBE = Ni i=1(Mi Si) Ni i=1Mi 212 where: 213 Mi: measured data at instance n; 214 Si: simulated data at instance n; 215 Ni: count of the number of dates used in the calibration. 216 (4) CVRMSE = Ni i=1 (  )2 Ni 1 Ni Ni i=1Mi 217 where: 218 Mi: measured data at instance n; 219 Si: simulated data at instance n; 220 Ni: count of the number of dates used in the calibration 221 222 ASHRAE Guideline 14 considers a building model calibrated when hourly MBE values fall within ±10% 223 and hourly CVRMSE values fall below 30%. The MBE provide an indication of errors averaged to the 9 224 mean of measured values but may be influenced by offsetting errors, while the CVRMSE index is a 225 measure of accumulated error normalized to the mean of the measured values. 226 227 3. Case studies: social housing 228 Whenever a detailed analysis is carried out, involving long campaigns of monitoring and evaluation of the 229 thermal envelope of buildings, it is complicated to have a large sample of case studies, due to financial 230 and time constraints [27]. For this reason, based on a previous analysis of the social housing stock of 231 southern Spain [42], the statistically more representative morphological and constructive building 232 typologies were determined and some neighbourhoods were selected by the Andalusian Government as 233 cases of interest for its energy retrofitting plan [43]. Three dwellings belonging to these cases of interest 234 were selected as case studies in this research. The main characteristics of the climate in which they are 235 located, morphological typology, constructive characteristics of its envelope and the user patterns will be 236 described below. 237 3.1. Location and climate 238 The three case studies are located in places with a Mediterranean climate (figure 1a), specifically three 239 Spanish cities with the lowest values in the scale of winter climate severity: Seville (winter climate zone 240 B), Huelva and Malaga (winter climate zone A) (figure 1b). Table 1 summarizes the main climate 241 characteristics of these locations. 242 (a) (b) Figure 1. Mediterranean climate map: (a) in Europe, (b) in southern Spain [44]. 243 16 342 hourly temperature, relative humidity and CO2 concentration evolution during the coldest month of the 343 winter period. In the temperature graphs an adaptive comfort band has been included following standard 344 EN-15251. The CO2 graphs also included a reference limit below which indoor air quality is considered 345 good (1200 ppm), following the FSIAQ (Finnish Classification of indoor climate 2000 [47]) standard for 346 category S3 (a less demanding category which may be associated with existing buildings). 347 348 In case 1 (Seville), during the winter period monitored (December 2014 - February 2015) the coldest 349 month was January. During this period (figure 5), with mild outdoor temperatures ranging from 5 ºC to 350 18 °C, inside the dwelling the temperatures were always around 14-15 °C, and so always below the lower 351 limit set by adaptive thermal comfort standard EN-15251 (around 18 °C). In particular, the average 352 January interior temperature was 15 °C in the living room and 14.2 °C in the main bedroom. The 353 behaviour of the two monitored rooms was very similar, with the exception of some times of night when 354 a local heating system was used in the bedroom. As it is a very shaded dwelling, there is not much solar 355 gain during the day repeatedly making indoor temperatures fall below outdoor ones. 356 357 Analysing other environmental variables (figure 6), an internal relative humidity between 60 % and 80 % 358 was found, compared to an external one varying from 40 % to 90 %. The interior CO2 concentration 359 shows two very different ways of use of the rooms monitored. While the living room is routinely kept 360 below 1250 ppm (with the exception of high occupancy periods) due to sporadic use, the bedroom 361 reaches 5000 ppm for a high percentage of nights because of more continued use. 362 363 In case 2 (Malaga), during the winter period monitored (December 2015 - February 2016) the coldest 364 month was February. In this month (figure 7), with warm outdoor temperatures ranging from 10 ºC to 20 365 ºC, the temperatures inside the dwelling were always around 18-19 ºC, and almost always below the 366 lower limit set by the adaptive thermal comfort standard EN-15251 (around 20 °C in this case). The 367 average interior temperature for February was 18.1 ºC in the main living room and 18.8 ºC in the main 368 bedroom. The behaviour of the two monitored rooms was very similar, with the exception of some 369 periods where a local heating system was used in the main living room. This dwelling maintains a very 370 stable temperature throughout the whole day. 371 17 372 373 374 Figure 5. Case 1 - Seville: hourly temperature evolution during January. 375 376 Figure 6. Case 1 - Seville: hourly relative humidity and CO2 concentration evolution during January. 377 378 Figure 7. Case 2 - Malaga: hourly temperature evolution during February. 18 379 380 Figure 8. Case 2 - Malaga: hourly relative humidity and CO2 concentration evolution during February. 381 382 Figure 9. Case 3 - Huelva: hourly temperature evolution during January. 383 384 Figure 10. Case 3 - Huelva: hourly relative humidity and CO2 concentration evolution during January. 385 19 386 An internal relative humidity between 50 % and 80 % was found (figure 8), compared to an external one 387 ranging from 30 % to 95 %. The interior CO2 concentration shows a very similar way of use for the two 388 rooms monitored, probably because the doors of all the rooms are always open linking them. Both are 389 routinely kept below 1250 ppm due to low occupancy of these rooms. Another key aspect for the good 390 quality of the interior air in this case study is the continuous operation of the natural ventilation system. 391 This system keeps a low continuous ventilation rate through the kitchen window and extracted by the 392 vents in the bathrooms. 393 394 In case study 3 (Huelva), during the monitored winter period (December 2016 - February 2017) the 395 coldest month was January. During this month (figure 9) the external temperatures were warm, between 396 10 ºC and 20 °C, while inside the dwelling the temperatures were measured at around 13-14 °C, far below 397 the lower limit set by adaptive thermal comfort standard EN-15251 (around 20 °C in this case). In 398 particular, the January average interior temperature was 14.7 °C in the living room and 13 °C in the main 399 bedroom. The two monitored rooms exhibit almost parallel behaviour, with the exception of night-time, 400 when the temperature decrease in the bedroom is greater. 401 402 Analysing other environmental variables (figure 10), an internal relative humidity between 50 % and 85 403 % was found, compared to an external one varying from 30 % to 70 %. The interior CO2 concentration 404 shows two slightly different ways of using the monitored rooms. During the day the doors of all rooms 405 are kept open which homogenizes the quality of the air. However, at night the bedroom reaches higher 406 CO2 concentrations, up to 2000 ppm, while the living room stays below 1250 ppm. In general, the interior 407 air maintains good quality due to the fact that the users ventilate the dwelling naturally every morning 408 (from 8.00 to 11.00 h, as seen from the survey). 409 4.2. Comfort analysis 410 The thermal comfort level in the three case studies during the winter period has been analysed according 411 to the adaptive model established by standard EN-15251 [26]. In addition, the percentage of hours of 412 discomfort and the average deviation of the interior temperatures with respect to the comfort band have 413 been contrasted with the percentage of hours of use of the local heating systems in each dwelling. 414 20 Period December 2014 January 2015 February 2015 Winter 2014/15 % hours of discomfort Living room 99.7 100 100 99.9 Bedroom 1 100 100 99.5 99.9 Average deviation Indoor temperature – Comfort band (ºC) Living room -3.4 -3.7 -2.7 -3.3 Bedroom 1 -3.9 -4.2 -3.9 -3.8 % hours of use of local heating systems Portable air heater 5.4 4.4 3.3 4.4 Bedroom 2 heat pump 9.6 14.7 4.6 9.9 415 (a) Period December 2015 January 2016 February 2016 Winter 2015/16 % hours of discomfort Living room 1 92.5 87.5 99.4 92.9 Bedroom 1 * 100 95.4 96.1 Average deviation Indoor temperature – Comfort band (ºC) Living room 1 -1.3 -1.6 -2.0 -1.6 Bedroom 1 * -1.3 -1.3 -1.3 % hours of use of local heating systems Living room 1 heat pump 12.9 15.2 18.2 15.3 Living room 2 heat pump 13.6 17.7 20.8 17.3 * Loss of monitoring data due to equipment failure. 416 (b) Period December 2016 January 2017 February 2017 Winter 2016/17 % hours of discomfort Living room 100 100 100 100 Bedroom 1 98.7 100 100 99.5 Average deviation Indoor temperature – Comfort band (ºC) Living room -3.6 -4.6 -3.0 -3.8 Bedroom 1 -4.8 -5.8 -4.2 -5.0 % hours of use of local heating systems Living room heat pump 7.1 22 2.8 10.9 Bedroom 1 heat pump 0.1 0.3 0.1 0.2 417 (c) 418 Figure 11. Evaluation of thermal comfort and use of heating systems during the winter period: (a) case 1 - 419 Seville, (b) case 2 - Malaga, (c) case 3 - Huelva. 420 421 In cases 1 and 3 (Seville and Huelva), a high percentage of hours of discomfort was detected throughout 422 the winter period in the two rooms monitored, around 100 % (figure 11a, 11c). In case 3, despite having 423 higher external temperatures than in case 1, interior temperatures even further removed from the comfort 424 band are observed. This is mainly due to heat losses of case 3 through the roof, as it is on the top floor. In 425 both cases, in spite of being in a clear situation of discomfort, the percentage of hours of use of local 426 heating systems is very low. During the coldest period (January) this percentage is less than 15 % in 21 427 Seville and 22 % in Huelva. In case 3 the use of the local heating system in the main bedroom is almost 428 zero although this is the room furthest away from the comfort band. 429 430 In case 2 (Malaga), as the climate is milder than in case 1 (but similar to case 3), the percentage of hours 431 of discomfort is slightly lower than in the other cases but it is still high. This percentage was above 90 % 432 during the whole winter period in the two rooms monitored (figure 11b). Although the percentage of 433 hours of discomfort does not differ much from the other cases, the graph shows that the average deviation 434 of the interior temperature with respect to the comfort band is much lower in case 2 (-1.6 ºC in the living 435 room and -1.3 °C in the bedroom) than in cases 1 (-3.3 °C in the living room and -3.8 °C in the bedroom) 436 and 3 (-3.8 °C in the living room and -5.0 °C in the bedroom). However, in this case the percentage of 437 hours of use of local heating systems is higher than in case 1 and similar to the case 3 living room during 438 the colder periods, reaching 20 % in February. In spite of this, the percentage of hours of use of the 439 heating systems does not nearly cover the percentage of hours of discomfort. 440 4.3. Energy consumptions 441 The electricity consumption of the three case studies has been analysed comprehensively and in detail 442 during the winter period, focusing on the consumption of the local heating systems. In order to evaluate 443 the heating consumption, the addition of the consumption of all local heating systems (kWh) has been 444 divided by the heated area, which in this case is the total area of the dwelling except for the kitchen and 445 bathrooms (the doors of the rooms are always kept open). In the three case studies, very low levels of 446 total consumption and no significant percentages of the use of local heating systems were detected. 447 448 In cases 1 and 2, it is worth noting that the Hot Water production system is electric. In case 1 (Seville) 449 specifically the water heater shows very significant consumption (figure 12a), around 24 %, well above 450 heating consumption, around 13 %. Also in this case, the high consumption of two computers, around 19 451 %, is significant since the users usually work at home. In case 2 (Malaga) (figure 12b), a balanced 452 distribution of consumption can be observed, in which heating accounts for around 20 % of total 453 consumption. In case 3 (Huelva) (figure 12c), the living room heating system consumption is notable as it 454 accounts for 29 % of the total consumption of the dwelling. In this case, the global consumption of all 455 devices without individual meters (‘others’ in figure 12) represents a more significant percentage than in 22 456 the other cases (52 %), due to the fact that it has less electric equipment (Hot Water is heated by gas) and 457 the consumption from lighting and cooking acquires greater relevance. 458 459 Looking at the total consumption data, it is apparent that case 1 (occupied by a young couple who work at 460 home) consumes the same as case 2 (occupied by a family of 4 members) and almost double that of case 461 3 (occupied by a retired couple). The economic level of the users of case 1 is slightly higher than the 462 others. When analysing the heating consumption, case 2 consumes the highest amount in relation to the 463 space heated, as well as taking up the highest percentage of use of the heating systems (figure 11). (a) Period Total consumption (kWh) Heating consumption (kWh/m2 heated) December 2014 355.89 1.12 January 2015 394.80 1.15 February 2015 305.84 0.47 Winter 2014/15 1056.53 2.75 (b) Period Total consumption (kWh) Heating consumption (kWh/m2 heated) December 2015 268.01 1.46 January 2016 327.23 1.57 February 2016 410.44 1.42 Winter 2015/16 1005.68 4.45 (c) Period Total consumption (kWh) Heating consumption (kWh/m2 heated) December 2016 165.87 0.77 January 2017 206.46 1.76 February 2017 133.68 0.38 Winter 2016/17 506.01 2.91 464 Figure 12. Evaluation of electric consumption during the winter period: (a) case 1 - Seville, (b) case 2 - 465 Malaga, (c) case 3 - Huelva. 23 466 Table 5. Statistical validation of the energy models. MBE Living room MBE Bedroom CV Living room CV Bedroom Case 1 (Seville) -5.33 % -5.06 % 10.14 % 11.27 % Case 2 (Malaga) -0.23 % -3.17 % 4.08 % 4.33 % Case 3 (Huelva) -4.78 % -1.73 % 8.22 % 5.61 % ASHRAE Standards MBE < 10 % CV(RMSE) < 30 % 467 468 469 (a) 470 471 (b) 472 473 (c) 474 Figure 13. Graphic validation of the energy models: (a) case 1 - Seville, (b) case 2 - Malaga, 475 (c) case 3 - Huelva. 24 476 4.4. Energy model adjustment 477 Following data collection for all case studies, the energy models were constructed and subsequently 478 simulated to verify the adjustment to the measured data. First, the results of energy simulations and 479 monitoring for a typical week were compared in graphs. According to the results obtained, the energy 480 model achieves the adjustment of variables on which there is some degree of uncertainty (due to the 481 impossibility of measurement), such as the natural ventilation rate through opening windows and the 482 occupants’ metabolic rate. The final result is an adequate graphic adjustment of the energy models (figure 483 13), which took into account the degree of error of the measurement equipment (+/- 0.5 ºC), represented 484 in the graph by error bars. 485 486 In addition, the statistical validation of the energy models has been verified during the whole winter 487 period, following the indicators established by the ASHRAE [35] (table 5). The adjustment between the 488 energy models and the measured data shows that the monitoring of the buildings is essential to obtaining 489 a valid adjustment, and so greatly reducing the number of uncertain variables. Therefore, for the energy 490 evaluation of retrofitting measures it is essential to have a calibrated simulation model, even more so in 491 case studies with a user pattern so far removed from the standards. 492 493 5. Conclusions 494 In southern Europe, there is extensive social housing stock with a notable need for retrofitting due to 495 energy obsolescence. Energy simulation incorporating standardized user patterns cannot be used to 496 ascertain the most appropriate measures for these particular dwellings to avoid the Prebound effect after 497 retrofitting. Given that the case studies evaluated demonstrate a variety of user profiles, retrofitting 498 decisions should not be simplified by using a single user profile. National regulations should define 499 different user profiles, including the number and age of users, the intensity of use of the dwellings, and 500 the climatic zone in which it is located (to consider the different criteria for thermal comfort). 501 502 After monitoring the environmental and energy behaviour of some case studies (during the winter 503 period), the following results have been obtained: 25 504 Despite the mild climate conditions during the harshest winter months, the average indoor 505 temperatures measured are low, oscillating between 13 ºC (Huelva) and 18.8 ºC (Malaga). 506 The low values of the interior temperature mean that for most of the time the dwellings are in 507 discomfort conditions, and the percentage of hours outside the comfort band ranges from 92.9 % 508 (Malaga) to 100 % (Huelva). 509 In spite of these inadequate indoor thermal conditions, users generally reject the use of their 510 local heating systems, mainly for financial reasons. The percentage of hours of use of the local 511 heating systems ranges from 9.9 % (Seville) to 17.3 % (Malaga). 512 Despite the warm Mediterranean winters, the low energy performance of the pre-1980 social housing 513 envelope (without any kind of thermal insulation) causes indoor thermal conditions far removed from 514 comfort. This, added to the lack of global and efficient heating systems and to financial constraints of 515 users, means that this housing stock is completely exposed to the risk of falling into a situation of fuel 516 poverty. The current energy policies must take into account the environmental behaviour of social 517 housing in southern Spain described in this research, and ensure that energy retrofitting grants focus on 518 passive measures that entail the improvement of thermal comfort instead of energy savings. 519 520 This research also demonstrates that the monitoring of in-use dwellings permits enough information to be 521 obtained to generate energy simulation models that minimize the ‘performance gap’ between real and 522 estimated behaviour. This is essential if retrofitting decisions are to be based on the results of energy 523 simulations. 524 525 Acknowledgements 526 This research was partially funded by the V Internal Research Plan of the Universidad de Sevilla and the 527 TEP-130 research group. 528 529 Part of this research was carried out within the research projects: ‘Intervención en barriadas residenciales 530 obsoletas: Manual de buenas prácticas’ (ref. 2013-0000006939), funded by the Andalusian Regional 531 Government; and ‘REFAVIV: Rehabilitación energética de las fachadas de viviendas sociales 532 deterioradas aplicando productos innovadores nacionales (DIT) y europeos (DITE)’ (ref. BIA2012-