*Correlation Approach*

There is a lack of data for most of the electric components and the physical characteristics of the university buildings (building materials, building geometric sizes). Thus, the given data set of observations gives us the opportunity to establish a statistical approach that allows us to measure the relationship between two variables by defining there correlation coe fficient, which will provide us with a straightforward interpretation of the two variables on the overall electricity consumption on a yearly basis. This method relies on historic values of overall energy consumption and background knowledge of the input variables that influence perception. In this case study, we define the first explanatory variable occupancy rate as the total number of students, professors, and administration sta ff for the academic year. If we ge<sup>t</sup> a weak correlation, we proceed by dividing the number of occupants into two groups, students and sta ff (professors and administration sta ff), and then test them separately. The second explanatory variable is the weather explanatory variable, defined as the sum of the heating and cooling degree-days during one academic year; its unit is in degrees Celsius.

Heating and cooling degree-days (HDD and CDD) are defined as the di fferences between the average daily outdoor temperature T(o/d) and corresponding base temperatures Tb [26]. The base temperature for heating and cooling is di fferent from place to place. It also depends on the type of building (household, administration, hospital). In this case study, the CDD temperature base is (T(b, CDD) = 28 ◦C) and the HDD temperature base is ( T(b, HDD) = 14 ◦C). These assumptions are based on a survey conducted inside the campus [24,25]. Note that the temperature base values indicate the outside temperature, and there is usually a minimum of two degrees of di fference between the inside and outside temperatures. We sum up both variables on a sequence of every academic year so that it can be lined up with the quantity of interest (EC).

$$\text{CDD} = T(\mathbf{o}/d) - T(\mathbf{b}, \mathbf{C} \mathbf{D} \mathbf{D}) \quad \text{if} \quad T(\mathbf{b}, \mathbf{C} \mathbf{D} \mathbf{D}) > T(\mathbf{o}/d) \quad \text{then} \quad \mathbf{C} \mathbf{D} \mathbf{D} = \mathbf{0} \tag{1}$$

$$\text{HDD} = \; \text{T(b,HDD)} - \; \text{T(o/d)} \quad \text{if} \quad \text{T(b,HDD)} < \text{T(o/d)} \quad \text{then} \quad \text{HDD} = 0 \tag{2}$$

The extensive weather data set will be used in this paper to develop and validate statistical models. The complex nature of EC inside buildings and the lack of the data of most components of buildings drive us to a black box model that relies on a simple input and output system [27]. The statistical model of the linear regression is set up according to the Formula (3):

$$\text{Ei} = b \times (\text{Xi}) + a \tag{3}$$

where *Ei* is the annual energy consumption corresponding to the academic year *i*, the input *Xi* is the explanatory variable, *b* is the slope, *a* is the y-intercept [28].
