The methodology consists of four steps (Figure 1). First, we determined the characteristics of the selected buildings based on measured parameters, building regulations and commonly employed Spanish databases. A monitoring campaign was carried out to obtain the actual building parameters of both dwellings, regarding the building envelope (thermal transmission and infiltration), energy consumption (total and for individual electrical equipment), subjective thermal comfort (occupants’ thermal votes), objective thermal comfort (temperature, relative humidity and CO2 level) and heating and ventilation occupancy practices.

As a second step, we developed the energy model of only one of the buildings, since our objective is to investigate the effect of different occupancy profiles. The software used for the building simulations is Design builder, which uses Energy+ for the calculations [19]. The monitored data was used as input data in the basic model. Four sub-models (M1, M2, M3 and M4) were developed based on the basic model, in which different occupancy profiles and household size were determined based on the monitored case studies, and based on building regulations.

The third step was to carry out a parametric analysis to determine the impact of the different occupancy profiles on energy demand.

Finally, the results of the simulation were obtained, analyzed and the conclusions drawn.

Two apartments were selected in the neighbourhood Ciudad de Los Angeles in South Madrid. The buildings, built in 1972, belong to the typical construction carried out between 1950 and 1980 in Spain [20]. The apartments are located in different buildings within a complex of eight apartment blocks with the same construction type. Seven of the eight buildings have been renovated (Figure 2).

One apartment in one of the renovated buildings, and one apartment in the not renovated building were selected to compare their energy performance and to evaluate the effectiveness of the renovation strategy applied. Monitoring equipment was installed in both the renovated (CP17) and the not renovated (CP18) apartments.

The apartments were specifically selected with similar building’s characteristics in order to compare and evaluate the effect of different occupancy patterns, without other confounding variables. Both apartments have the same general characteristics, such as construction process, orientation, layout both within the apartment and within the building. In addition, the apartments have the same number and typology of occupants.

The apartments have a useful living area of 63 m2 (Table 1), North-South orientation and are located in a middle level within the building. In the renovated building, the original masonry facade has been retrofitted with a system consisting on a four centimetres layer of expanded polystyrene (SATE system). In addition, the single layer glazing (3 mm) and aluminium frames have been updated by adding an external window with aluminium frame and Climalit 4/6/4 glazing without solar treatment. Both dwellings have central heating with a gas boiler and radiators in all rooms. Table 1 shows the main characteristics of the buildings and the users.

The dwellings were also selected based on the characteristics of the users. Previous work has shown that demography, lifestyle and comfort preferences affect the energy performance of buildings [21]. Apartments with similar occupants’ profiles were selected to determine the influence of other occupancy variables. However, these is out of the scope if this paper.

The monitoring campaign consisted on three types of data collection: 1) measurements on the building envelope; 2) actual electricity consumption; and 3) occupants’ behaviour (interviews and questionnaires). The measurements on the building envelope were used to define the building physical parameters, while the actual electric energy consumption and occupants’ behaviour were used to define the occupancy patterns.

There are two main approaches to study the relationship between occupants’ behaviour and energy efficiency of buildings. The first approach consists on the application of questionnaire surveys and interviews to the residents of the dwellings, and it is therefore of a qualitative nature. The second approach consists on on-site monitoring campaigns and it is quantitative and more objective in nature [27]. In this study a mix-methods approach has been used to determine the occupancy profiles of the case studies.

Each occupancy profile was generated based on the qualitative analysis of the electricity measured in the flat (quantitative data), and the interviews with the residents (qualitative data). Two profiles were developed per dwelling: a 24 hours profile summarising the weekly occupancy, and a 48 hours profile showing the occupancy during the weekends. A study in the United Kingdom shown no difference on occupancy between weekdays and weekends [13], however, in these case studies, large differences were seen during the weekends. For both profiles, the occupancy per hour has been defined according to the mean occupancy per hour for the monitoring period during weekdays, and during weekends.

Figure 3 shows the occupants’ presence at home on a typical weekly day for dwellings CP17 (top) and CP18 (bottom). The figures to the left show the total electrical consumption in the dwellings, codified into 0 = no consumption at all, and 1 = some consumption (the electricity use of the refrigerator is not included in this figure). For this research these values are translated into absent from home (no electricity consumption) and present at home (some electricity consumption). The figures show, per hour, the average presence at home on a typical week day (average from Monday to Friday for the whole monitoring period).

The figures to the right, show the occupancy profiles resulting from the triangulation of the electricity consumption data (figures to the right) with the interviews carried out with the residents. In addition, the following criteria was used:

After 22:00 hours the electric energy consumption decreases drastically, suggesting that the residents have gone to sleep.

After 9:00 hours, the electricity consumption increases, indicating that the residents start with their daily routines.

These criteria have been corroborated with the information provided by the interviews carried out with the residents.

Figure 4 and Figure 5 show the same profiles as described before, for the weekends.

Figure 6 shows the occupancy profile defined based on the Spanish normative (Documento Básico de Ahorro de Energía del Código Técnico de Edificación Español [28]). The objective of this profile is to compare it with the data obtained from the case studies. The normative defines occupancy profiles based on the type of building, occupation density (low, medium and high) and occupancy period (8, 12, 16 and 24 hours). Only one profile has been defined for residential buildings, while 12 profiles have been defined for non-residential buildings.

To investigate the effect of the different occupancy profiles and household composition (density) on the heating demand, the simulations were carried out only based on the characteristics measured in the renovated building (CP17). The remaining data entered in the model have been obtained from the current regulations and recognized Spanish databases.

Building simulations were carried out variating the occupancy profile and the occupancy density. In this way, the number of variables was limited only to those affected by occupancy.

Figure 7 show the framework in which the submodels are based, indicating the factors investigated per generated submodels.

Figure 8 shows the comparison between the occupancy profiles developed with the monitoring data in the dwellings CP17 and CP18, with the profile obtained from the Spanish Building Code. The profiles from the monitoring case studies show similarities. Both households consists of single elderly women with similar daily routines. The difference between the households is observed during the morning (between 8:00 and 14:00 hours), in which the resident of CP17 works part-time according to the interviews. The resident of CP18 also leaves the house in the morning, but for a shorter period (11:00-13:00 hours).

In comparison, the profile obtained from the Spanish Building Code (CTE) represents a household consisting of a nuclear family. In the profile, this type of household is absent from 8:00 to 15:00 hours; during the afternoon, only a few households’ members are back; and during the night, everybody is back home.

Figure 8 also shows the comparison of profiles for weekend days. On the one hand, according to CTE, residential buildings have a full occupancy during the whole weekends. On the other hand, the case studies shown very different profiles. The resident of CP17 keeps a similar profile to that followed on weekdays; while the resident of CP18 is usually absent during the weekends and until Monday mornings.

Table 2 shows the heating demand resulting from each of the simulation models (in kWh/year), the internal heat gains caused by occupancy (in kWh/year), and the relation between the heating demand and the occupancy-related heat gains.

The yearly heating demand range for the simulated dwelling is between 44.7 and 53.7 kWh/m2. The yearly heating period is considered from October 1 to March 31. The highest heating demand is seen in the simulation ran without occupants, which is used as a basis for the comparisons.

In the simulations carried out with the case studies profiles (CP17 and CP18), we see a difference of 1.2 kWh/year on the occupancy heat gains, since the resident of dwelling CP18 is absent on the weekends (22.6 % of the time). The results also show that the difference between the heat gains in the CP18 profile in comparison to the heat gains in the CTE low-density profile, is only 2.1 %. This difference corresponds to the difference on occupancy profiles (i.e. time when the occupants are home). The heat gains in the CTE high-density profile are twice as high as in the other profiles, increasing the importance of the occupancy in the heating demand.