**4. Discussion**

#### *4.1. Benchmark Analysis*

Dividing campus facilities into categories allows us to compare buildings based on their utility and to identify the category that has the lowest and largest energy consumption by GFA. The category that scored the highest EUI average in this case study is the research and science category by an average metric score of 119.5 kWh/m2, which is inferior to the laboratory intensity median given by CBECS 2003 data by Oak Ridge National Lab and the Department of Energy, which is 226 kWh/m<sup>2</sup> for the Mediterranean climate [9]. In another comparative study, which was conducted at the regional scale in the state of California (Mediterranean climate), they showed that some laboratories scored the highest energy intensity with a value of 909.5 kWh/m2, and it is four times bigger than the state average energy intensity [29–31]. The results in our study have proven that this category can provide a wide range of energy intensity, which makes it worthy of more in-depth study. The observations indicate that the longer operation hours inside the facilities of the research and science category, its heavy plug load materials (like ultra-low freezers and incubators), and other laboratory equipment are the reason behind the high intensity. In addition, the high number of computers used are the reason behind the high intensity. The second most intense category is the library category, which accounts for only 7% of the total GFA but contributes to 15% in of total EC. The buildings in this category peak during the months of the preparation of exams, especially in the summer session, when their EC becomes three times higher, unlike the teaching and seminary rooms, which have a slight increase during the same period when the teaching days are relatively lower. Our portfolio have an EUI inferior than 50 kWh/m<sup>2</sup> (Figure 6), which is the equivalent of one third of the k–12 schools in hot and humid zones. This brings back the question of which is more energy intense—schools or universities—and how much can the weather parameter contribute to the increase of electricity consumption. Thus, there are several parameters other than the weather to take into consideration, such as occupancy rate, number of COM, plug load, and operating time, that are responsible for the EC gap between different categories. On the other hand, simulation techniques represent one of the efficient alternatives to evaluate the energy performance of a building regardless of its utility. This method was used to develop a benchmark analysis based on models of equipment and system performance, which proved that plug loads and HVAC are some of the biggest influencers of high energy consumption in laboratory buildings.

#### *4.2. Correlation Analysis*

Despite the complexity of EC in university campuses, we were able to demonstrate that outdoor temperature and number of occupants positively correlate with the overall energy use, which confirmed our choice of variables. Nevertheless, another study that developed a simulation of the building occupants' decision-making and information communication process found that the network size has no significant impact on the EC [32]. Still, that result needs to be confirmed in non-residential building, especially in cases like schools and universities, where the number of occupants changes substantially over the year. On the other hand, a study that was conducted on 10 universities in the US and confirmed that EC correlates highly with outdoor temperature [33]. Weather variations can easily change cooling and heating use by 20–30% [34]. The estimation of the occupancy rate for each building remains challenging, especially in this case study because of the irregular patterns of the student movement inside the campus, which is not only related to the classes or other scheduled activities. Nonetheless, some studies used CO2 measuring, relative humidity, and acoustic sensors to estimate occupancy [35–37]; however, those techniques are hard to implement in our case of study because of the several components that the buildings incorporate. On the other hand, many studies [38,39] have focused on the behavior of occupants rather than the size of active occupants. In order to evaluate the energy saving potential, one study developed an occupancy model of individuals moving in and out in o ffices [40], while another study found an alternative to analyze occupancy patterns using physical-statistical approaches to improve energy demand forecasting [35,41]. Still, identifying how occupant's behavior influences EC is complex because of the stochastic nature of individual actions [37].

The five buildings (31,30,29,24,32) from our portfolio scored a higher energy intensity than the university in a Mediterranean climate median. Installing solar panels on the roof as a backup is highly recommended since the campus is located on southern coast of Spain, where the yearly sum of global irradiation is over 1900 kWh/m<sup>2</sup> [41]. Therefore, e fforts and investments have to prioritize these facilities because they are driving the high consumptions and there is a considerable gain potential to achieve from their high energy intensity.

If a model with the two explicative variables (N and CDD&HCC) is established and a multiple regression is conducted, Equation (4) is obtained.

$$\text{EC (Kwh)} = 1050.375608 \,\text{N} + 1172.322578 \,\text{CDM\&HCC (°C)} - 9320518.243 \tag{4}$$

with R<sup>2</sup> = 0.83. If the model obtained is plotted (see Figure 11), the range of expected energy consumption can be found according to the parameters of number of occupants (N) and CDD and HCC in ◦C. Therefore, a threshold of 8 GWh has been estimated as the energy consumption limit to be achieved for 15,000 persons and for a CDD and HCC of 1350 ◦C, both factors being the average of the last eight years.

**Figure 11.** Energy consumption (EC) model obtained for the explicative variables: N (number of occupants) and CDD and HCC (◦C).
