**E**ff**ect on the Thermal Properties of Mortar Blocks by Using Recycled Glass and Its Application for Social Dwellings**

#### **Vicente Flores-Alés <sup>1</sup> , Alexis Pérez-Fargallo <sup>2</sup> , Jesús A. Pulido Arcas <sup>3</sup> and Carlos Rubio-Bellido 1,\***


Received: 25 September 2020; Accepted: 29 October 2020; Published: 31 October 2020

**Abstract:** Including recycled waste material in cement mixes, as substitutes for natural aggregates, has resulted in diverse research projects, normally focused on mechanical capacities. In the case of recycled glass as an aggregate, this provides a noticeable improvement in thermal properties, depending on its dosage. This idea raises possible construction solutions that reduce the environmental impact and improves thermal behavior. For this research, an extended building typology that is susceptible to experiencing the risk of energy poverty has been chosen. The typology is typical for social housing, built using mortar blocks with crushed glass. First, the basic thermophysical properties of the mortars were determined by laboratory tests; after that, the dynamic thermal properties of representative constructive solutions using these mortars were simulated in seven representative climate zones in Chile. An analysis methodology based on periodic thermal transmittance, adaptive comfort levels and energy demand was run for the 21 proposed models. In addition, the results show that thermal comfort hours increases significantly in thermal zones 1, 2, 3 and 6; from 23 h up to 199 h during a year. It is in these zones where the distance with respect to the neutral temperature of the m50 solution reduces that of the m25 solution by half; i.e., in zone 1, from −429 ◦C with the m25 solution to −864 ◦C with the m50. This research intends to be a starting point to generate an analysis methodology for construction solutions in the built environment, from the point of view of thermal comfort.

**Keywords:** crushed glass; periodic thermal transmittance; energy demand; adaptive comfort; social housing

#### **1. Introduction**

The average glass recycling rate in the European Union (28 member countries) has reached a 76% threshold for the first time. This means that more than 12.4 million tons of glass were collected throughout the European Union in 2017, 2% more than in 2016 [1,2]. The proportion of recycled glass in the US is currently estimated to be around 35%. There is very little reliable data available for other countries. In the case of China, the recycling rate for container glass is currently still below 20% and in South Africa it is over 41% [3]. It is important to highlight the impact of domestic recycling in a recovery chain of simple and safe containers. Glass is inert and maintains its inherent properties regardless of how many times it has been recycled. If suitably collected, it can be recycled ad infinitum in a closed circuit, hence repeatedly using this waste will help preserve the natural resources of the Earth, minimizing landfill spaces and saving energy and money [4].

#### *1.1. E*ff*ect of Crushed Glass on Thermal and Mechanical Properties of Mortars*

Cement mixes have traditionally been researched as construction products, in particular incorporating inert aggregates with sufficient resistance capacity to be substitutes of the virgin aggregate, with the ultimate goal of reducing the environmental impact associated with gravel and sand extraction. The chemical composition of glass mainly has a formless siliceous nature, which makes it compatible with natural aggregates, although when a very small size particle is used (<20 μm) [5], it has a reactive nature induced by the high alkalinity of cement [6]. The alkalinity of glass causes the breakdown of the matrix favoring, by its formless nature, the formation of calcium silicate hydrate, improving the cementitious performance [7]. The fine particles of glass have a high specific surface [8] and therefore favor high pozzolanic reaction kinetics due to the strong reaction between the alkali in the cement and the reactive silica in the glass [9]. It must also be considered that glass can intervene in alkali aggregate reactions, although the participation of recycled material in this reaction depends on the particle size, with an expansive phenomena of sizes above 1 mm being favored [10]. In fact, the research made shows that fine waste glass has a mitigating capacity of the ASR [11]. Finally, it is worth mentioning that there are incipient studies which address the capacity that incorporating glass into cement mortar provides to improve the resistance to the penetration of chlorides [12,13], and to develop bactericidal features [14,15].

Regarding resistance, several researchers have shown that crushed glass increases compression resistance [16,17], although the data has a significant spread in the results, obtaining optimal improvement percentages of between 10% and 30% compared to the reference mortar, depending on mortar type and the maximum size of the glass [18]. Regardless of the aggregate substitution, the dose conditions, and especially the w/c ratio, are determining factors in the mechanical behavior of the product [19]. In addition, Castro and de Brito (2013), regarding the mechanical behavior, showed a general improvement in terms of resistance to the carbonatation of concretes that contain glass waste (size <4 mm) as a natural fine aggregate [20].

Likewise, previous research has highlighted the capacity of glass aggregate as a substitute for sand to substantially reduce the thermal conductivity of mortar [21], demonstrating that energy savings can be achieved when using a glass aggregate instead of the sand alternative [22]. Sikora et al. have shown that substituting fine sand by WG can improve the thermal properties of cement mortars, while maintaining an acceptable mechanical strength [23].

#### *1.2. Low-Cost Materials and Its Application to Energy-E*ffi*cient Dwellings*

In the last few years, the building industry has been striving to reduce its energy consumption, and one of its strategies for this focuses on using recycled products as a construction material. In this sense, since crushed glass has beneficial effects on the thermal conductivity of mortars, is easy to obtain and has a low price, it has emerged as a viable option. Since crushed glass has a controversial effect on the mechanical properties, its use in structural elements is out of the question, being rather applied to coatings and finishing without structural function, but where thermal insulation becomes crucial to reduce the energy demands of the building.

Research on mortar blocks brings the opportunity of introducing an affordable material that also allows for a better insulation, as suggested by other authors [24], and finds a specific application in projects with limited financial resources, such as those comprising social dwellings. The use of recycled glass has wider implications, it is an environmentally friendly material, an affordable constructive solution and is technically feasible, even in countries where manpower has limited technical skills [25]. Previous research by the authors has focused on clarifying the feasibility of such materials on the basis of its chemical, mineralogical, physical, thermal and mechanical properties, and former studies claim that a percentage of recycled glass between 25% and 50% results in mortars with a lower thermal conductivity and higher density that, at the same time, have mechanical capacities comparable to mortars with natural aggregates [26,27].

Up to date, the great majority of studies in the field, as well as building codes, use the static thermal transmittance as a proxy for assessing the insulation capacity of a given material. Nevertheless, this approach ignores the complex interplay among other variables, such as the thermal inertia [28], which, in combination with the former, exerts a remarkable influence on the energy performance of buildings [29]. Static thermal transmittance relies on a simplification, where the temperature gap between the interior and the exterior of the building is constant, disregarding the effect that warm and cold climates with wide thermal oscillations might have on the energy demand of the building [30]. Recently, the UNE-EN ISO 13786 Standard [31] has taken the lead in incorporating the so-called dynamic thermal properties of building materials. In brief, this document assumes that heat flow between both sides of the envelope depends on the dynamic variation of the temperature gap through time. Starting with the calculation procedure as per the UNE-EN ISO 6946 [32], this standard considers both the static and the dynamic thermal transmittance.

Being this a novel approach, research on this area is still limited, but some researchers have already shed light on this, suggesting that periodic thermal transmittance can lead to a reduction in the energy demand of the building [33]; the authors have also made a contribution in this field, claiming that not static thermal transmittance, but thermal inertia, is the driver to improve indoor thermal comfort in social dwellings located in the Central-South area of Chile [34]. However, research in this field is still scarce and fragmented. Plenty of studies deal with the development of a new construction material with improved static thermal transmittance that, after determination of its properties by laboratory test, is applied to common constructive solutions. In the case of dynamic thermal properties, the process is more complex and there is still a research gap in considering studies that comprise the whole process: Determination of the properties by laboratory tests, implementation of the material in constructive solutions, and analysis of the effect on different aspects of energy demand and building comfort.

This study aims at filling-in this research gap by presenting a study comprising all these steps. In turn, the results of this study are expected to be applied to the design and construction of social dwellings in Chile. Subsidized housing always represents a challenge for designers and builders, as it needs to balance the constraints of a limited construction budget with the need of providing the lower strata of society with decent living standards. Chile is a representative case study for two reasons: First, this country encompasses a great variety of climates, including hot dry deserts in the North and cold steppes in the South. Second, this country has had a continuous and solid program of social dwellings, with 3.6 million subsidies granted between 1.964 and 2.015, and an estimated investment of 19 billion Euros since 1.990 [35]. At the present time, Supreme Decrees 01 and 49 establish the basic technical standards for social dwellings [36–39], which consist of predefined typologies with standardized constructive solutions: Built surfaces are between 36 and 55 m2 and usually have a living-dining area, a kitchen, a bathroom and two or three bedrooms. Considerations of energy efficiency were introduced only after the enactment of the General Urbanism and Constructions Ordinance (OGUC, in Spanish) in 2007 [40], which was the first legislation that established the limits for the U-values of the external envelope in Chilean buildings. After this, the government has put much effort into improving the benchmark for energy efficiency by releasing technical guidelines that, although not mandatory, have started to impregnate professional practice in the country: The Standards for the Sustainable Construction of Housing, published in 2014, raised the benchmark for thermal envelopes but is still not mandatory [41].

This background describes a country that, in spite of an increasing awareness about energy efficiency in buildings, still has a long way to go. Chilean researchers have clarified how low insulated houses can have unacceptable low temperatures in winter, as low as 14 ◦C [42], which, in turn, may lead to a higher prevalence of respiratory illnesses [43]. The thermal adaptation of users in central-southern Chile has its own particularities, but currently, from the adaptive thermal comfort models included in the standards, the model that is part of ASHRAE 55-2017 is the one with the most similarities to those of the users [42,44].

In sum, there is a need for comprehensive research on how recycled glass incorporated into construction blocks can improve both static and thermal properties of the thermal envelope, and on how these constructive solutions may find an application in social dwellings in Chile.

The article is organized in three sections. First, the methodology used in the research is described, analyzing the following aspects: (i) types of construction solutions considered and stationary transmittance; (ii) considerations about the periodic thermal transmittance calculation of UNE-EN ISO 13786; (iii) definition of the case study and the thermal modeling; (iv) analysis of the studied climate zones; and (v) the approach of the comfort and energy demand analysis. Second, the results are presented and discussed. Finally, the main conclusions of the results obtained in the study are summarized.

#### **2. Materials and Methods**

#### *2.1. Thermopyshical Properties of the Material*

In previous works, the potential viability of the materials being studied has been clarified [26,27,45], starting from the evaluation of the chemical, mineralogical, physical, thermal and mechanical characteristics, as well as the correlation with thermal conductivity coefficients [46]. The thermal conductivity results have confirmed that the doses with 25% and 50% of glass aggregate have thermal conductivity coefficients that are noticeably lower than those of the reference material, and also higher densities, maintaining a sufficient mechanical capacity compared with natural aggregate mortars (Table 1).

**Table 1.** Sand and glass composition of mortars used in the study, density of the end product (ρ), and thermal conductivity (δ) at 30 ◦C.


The particles were classified according to their size as per UNE EN 933-1:2012 standard (0.063, 0.125, 0.250, 0.50, 1.00, 2.00 mm) [30] and continuous particle size with maximum compactness were prepared in accordance with Fuller's curve (Figures 1 and 2) [28]. Thermal conductivity was determined by differential scanning calorimetry (MDSC) at 30 ◦C, thermal diffusivity (cm2/s) and thermal conductivity (W/mK) were measured on Linseis measuring equipment (LFA 1600) and a DSC Q20-TA [35].

**Figure 1.** Aggregate particle size of crushed glass. Reprinted from Flores-Alés, V., Alducin-Ochoa, J. M., Martín-del-Río, J. J., Torres-González, M., and Jiménez-Bayarri, V. (2020). Physical-mechanical behaviour and transformations at high temperature in a cement mortar with waste glass as aggregate. Journal of Building Engineering, 29, 101158.

**Figure 2.** Crushed glass and mortar based on recycled glass as aggregate.

#### *2.2. Constructive Solutions and Static Thermal Properties*

This study considered three constructive solutions, which are commonplace in the construction of dwellings. Mortar blocks with a thickness of 11.5 cm constitute the core of the external walls; they are coated by a layer of cement mortar on the outside and plaster on the inside. The three solutions differ in the percentage of crushed glass used to elaborate the mortar blocks (mR, m25 and m50) which, in turn, modified their static properties: Thermal conductivity, gross density and specific heat capacity (Table 2). It is remarkable how higher percentages of crushed glass brings lower thermal conductivities and higher densities, while maintaining a nearly constant specific heat capacity.


**Table 2.** 15 cm brick thermal transmittance: mR, m25, m50.

#### *2.3. Assumption of Periodic Thermal Transmittance and Dynamic Thermal Properties*

The UNE-EN ISO 13786 standard defines the analytical calculation procedure and the parameters related to the dynamic thermal behavior of building envelopes [31]. The theoretical basis of this was primarily established by Carslaw and Jaeger [47], who analyzed the sinusoidal ratio between the heat flow and the indoor and outdoor temperatures. The oscillation period (T) of the temperatures can be an hour, a day or a year, and this research assumes that the sinusoidal variation has a period of 1 day, which corresponds to the daily oscillations of the external temperatures [31]. After this, three static properties of all the considered materials must be known: thermal conductivity (λ), density (ρ), and specific thermal capacity (c). In this case, they were already determined by laboratory tests. Thermal bridges do not have to be considered due to their low impact on dynamic thermal properties [31].

The calculation procedure uses the thermal transference matrices for each one of the homogeneous layers of the wall (*Zmn*). The elements of a transference matrix are defined as (Equation (1)):

$$\begin{aligned} Z\_{mn} &= \begin{pmatrix} Z\_{11} & Z\_{12} \\ Z\_{21} & Z\_{22} \end{pmatrix} \\ Z\_{11} &= Z\_{22} = \cosh(\xi)\cos(\xi) + j\operatorname{sech}(\xi)\operatorname{sen}(\xi) \\ Z\_{12} &= -\frac{\delta}{\Delta} \{\operatorname{sech}(\xi)\cos(\xi) + \cosh(\xi)\operatorname{sen}(\xi) + j\operatorname{\{\cosh(\xi)\operatorname{sen}(\xi) - \operatorname{sech}(\xi)\operatorname{\cos}(\xi)\}\} \\ Z\_{21} &= -\frac{\lambda}{5} \{\operatorname{\sinh(\xi)\cos(\xi) - \cosh(\xi)\operatorname{\sin}(\xi) + j\operatorname{\{\cosh(\xi)\sin(\xi) + \sinh(\xi)\operatorname{\cos}(\xi)\}\} \end{aligned} \tag{1}$$

where δ (m) is the periodic penetration depth of a thermal wave on the layer's material (Equation (2)), and ξ (dimensionless) is the ratio between *d* and δ (Equation (3)).

$$
\delta = \sqrt{\frac{\lambda T}{\pi \rho c}} \tag{2}
$$

$$
\xi = \frac{d}{\delta} \tag{3}
$$

The transference matrix of a wall (*Z*) is defined as the multiplication of the matrices of the different layers (*Zi*) from the outside (*i* = *N*) to the inside (*i* = 1):

$$Z = \begin{pmatrix} Z\_{11} & Z\_{12} \\ Z\_{21} & Z\_{22} \end{pmatrix} = \prod\_{i=N}^{1} Z\_i \tag{4}$$

The different periodic variables can be determined by operating the elements of the matrix. The periodic variables defined in UNE-EN ISO 13786, and considered in the study, are: (i) periodic thermal transmittance (|*Y*12|) (W/(m2K)), which is the module of the complex number defined as the complex amplitude of the heat flow density through the indoor component's surface, divided by the complex amplitude of the temperature of the outdoor area, when the indoor temperature is constant (Equation (5)); (ii) time shift periodic thermal transmittance (ϕ) (h), which is the period of time between the maximum amplitude of a cause and the maximum amplitude of its effect, related to the periodic thermal transmittance (Equation (6)); (iii) decrement factor (f) (dimensionless), which is the quotient between the periodic thermal transmittance module and the U-value (Equation (7)); (iv) internal thermal admittance (|*Y*11|) (W/(m2K)), which is the complex number module defined as the complex amplitude of the heat flow density through the surface of the component adjoining the indoor area, divided by the complex amplitude of the temperature in the same area when the indoor temperature is kept constant (Equation (8)); (v) time shift internal side (ϕ11) (h), which is the period of time between the maximum amplitude of a cause and the maximum amplitude of its effect, related to the internal thermal admittance (Equation (9)); (vi) external thermal admittance (|*Y*22|) (W/(m2K)), which is the complex number module defined as the complex amplitude of the heat flow density through the surface of the component adjoining the outdoor area, divided by the complex amplitude of the temperature in the same area when the outdoor temperature remains constant (Equation (10)); and (vii) time shift external side (ϕ22) (h), which is the period of time between the maximum amplitude of a cause and the maximum amplitude of its effect related to the external thermal admittance (Equation (11)).

$$Y\_{12} = -\frac{1}{Z\_{12}}\tag{5}$$

$$\varphi = \frac{T}{2\pi} \arg(Z\_{12})\tag{6}$$

$$f = \frac{|\mathcal{Y}\_{12}|}{\mathcal{U}}\tag{7}$$

$$Y\_{11} = -\frac{Z\_{11}}{Z\_{12}}\tag{8}$$

$$
\varphi\_{11} = \frac{T}{2\pi} \arg(\mathcal{Y}\_{11}) \tag{9}
$$

$$Y\_{22} = -\frac{Z\_{22}}{Z\_{12}}\tag{10}$$

$$q\_{22} = \frac{T}{2\pi} \arg(\mathbf{Y}\_{22}) \tag{11}$$

#### *2.4. Case Study and Thermal Model*

Figure 3 shows housing model for the study.

A representative case-study was selected from the Project Bank of the Bio-Bio Region's Housing and Urbanization Service (SERVIU, in Spanish) [37]; a detached house with a higher surface of thermal envelope (external walls, roof and slabs) would have, a priori, a worse energy performance in this climate [38]. The prototype was modelled in the EnergyPlus® simulation software [36] and parametric simulations were done considering the thermal constructive properties of Table 3. The base-case scenario considered U-values as per the Chilean Building code, and the thermal properties of three different walls were introduced, using the three different mortars considered (Table 2). Ventilation rates were adjusted as per the minimum values recommended by the Chilean standard. In addition, the values for internal loads, such as occupation, lighting, equipment and ventilation, were considered (Table 4), as well as their schedules (Figure 4).

**Figure 3.** Housing model for the study.

**Table 3.** Thermal constructive properties of the case studies with the three bricks.


\* Depends on the climate zone considered.


**Table 4.** Internal heat loads for the models.

**Figure 4.** Occupation, lighting and equipment and ventilation schedules. Source: [34].

#### *2.5. Climate Zones*

Chile's climate varies greatly, covering the climatic variants B (arid and semi-arid), C (template) and E (cold) of the Köppen–Geiger classification. According to the current standard, Chile is divided into 7 thermal zones considering the annual heating degree days (Table 5) [48]. The Thermal Regulation (RT, in Spanish) for housing came into force in 2000. In a first stage, maximum thermal transmittance requirements were defined for roofs and, in a second stage in 2007, requirements were established for walls, ventilated floors and windows [49], which are mainly based on heating degree days. A representative city has been chosen for each of the 7 thermal zones, which also allows for an easy classification as per Köppen–Geiger (Table 5). An EPW weather file was considered for each city to model the external conditions.


**Table 5.** Selected locations from Chilean Climatic Zoning.

#### *2.6. Thermal Comfort and Energy Analysis*

The influence of using mortar bricks with glass in housing will be determined starting from the simulation results from EnergyPlus®, using two thermal comfort indicators and four energy demand indicators. The comfort indicators will be obtained by simulating the dwelling in the seven climates with the three brick types (mR, m25 and m50), in free oscillation during the entire year. With the results of the hourly operational temperatures in free oscillation, the number of annual hours where the operational temperatures are within the adaptive thermal comfort (ATC) limits of the model defined in the ASHRAE 55-2017 [44] standard, will be quantified, as will the distance in hourly operational temperature degrees to the thermal neutrality to quantify the reduction or increase of extremely hot or cold temperatures. In this case, a thermal acceptability limit of 80%, as per ASHRAE 55-2017, was considered; Chile, as many other countries, still does not have its own standard for adaptive thermal comfort, thus international documents are adopted. ATC models were originally developed

for office buildings [50], whereas they find also application in residential buildings, considering that occupants may change their clothes and operate windows to achieve thermal comfort [51]. Besides, previous studies also support the fact that adaptive comfort finds an application in naturally cooled houses, and finds applicability in social dwellings [42].

The ASHRAE adaptive model is governed by Equation 12, which defined the neutral temperature inside the building (*Tn*) as a function of the *T*pma(out); a range of ±3.5 ◦C gives an acceptability of 80% and ±2.5 ◦C gives a 90%.

$$T\_n = 0.31 \times T\_{\text{pma(out)}} + 17.8 \tag{12}$$

*T*pma(out) is a weighted average of the mean external temperatures of the previous 7 days (Equation (13)). *T*e(d−1) is the average outdoor temperature of the previous day, *T*e(d−2), the average outdoor temperature of two days prior and so on and so forth; and α is a constant that depends on the thermal oscillation of the local climates, assuming α = 0.8 [44].

$$T\_{\rm pma(out)} = (1 - \mathfrak{a}) \times \left( T\_{\rm e(d-1)} + \mathfrak{a} \times T\_{\rm e(d-2)} + \mathfrak{a}^2 \times T\_{\rm e(d-3)} + \mathfrak{a}^3 \times T\_{\rm e(d-4)} + \cdots \right) \tag{13}$$

(*T*pma(out)) must be within 10.0 ◦C and 33.5◦C so that this standard find application. If (*T*pma(out)) falls outside those limits, the neutral temperature will be a constant, as the standard assumes that when it is too cold or too hot, active cooling or heating becomes necessary, thus internal temperatures are decoupled from the external oscillations. In that case, this study assumed a heating setpoint temperature of 20 ◦C and a cooling setpoint of 26 ◦C, as per EN 16798 standard, Category II [52]. If necessary, the dwelling would have to resort to heating or cooling devices and therefore the energy demand was also recorded. Those variables were simulated for the 3 constructive solutions (Table 2) and the 7 climate zones of Chile (Table 5).

#### **3. Results and Discussion**

#### *3.1. Periodical Thermal Properties*

Taking the data as a base for the static thermal transmittance (Table 2), and following the calculation procedure described in Section 2.2, the periodic thermal properties of the 3 constructive solutions considered were calculated. Thermal conductivity (λ), density (ρ) and specific heat capacity (c) were already known for each solution; the calculation period for the thermal oscillation was 24 h for all cases, which is the recommended value for daily meteorological variations and temperature setback. The dynamic thermal properties were calculated (Table 6) using the tool to calculate thermal mass [53].


**Table 6.** Calculation results according to EN ISO 13786.

Static thermal transmittance could be reduced by around 13% on using recycled glass aggregates, and similar effects can be seen in the dynamic thermal properties, though the discussion there is more complex. Dynamic thermal transmittance was reduced by around 22%; the decrement factor was reduced from 0.86 to 0.78 (−10%), and the time shift was increased by one hour, which means that the thermal oscillation amplitude is reduced conjointly with a delay in the transmission of heat

from the outside to the inside. This could be expected because of the greater heat capacity of the proposed material.

However, to fully grasp the real implications of these properties, additional data were needed. Since dynamic thermal properties are highly dependent on temperature oscillation, which, in turn, is a function of the local climate, it was deemed necessary to clarify how these solutions would work in all the climate zones of Chile. For this purpose, external temperature variations were simulated during a 24-h cycle, by approximating the oscillation to a cosine function in the form of:

$$t(h) = t\_{avg} + t\_{amp} \times \left(\cos(t - t\_{\text{max}})\right) \tag{14}$$

where *tavg* is the daily average temperature, *tamp* is the daily temperature amplitude, *t* is time in hours and *tmax* is the time of the day when the outdoor temperature reaches its maximum. As a result, this function delivers an output, *t(h)*, which is the hourly external temperature for 24 h. Additional data were needed to calculate the temperature oscillation inside the building: The thermal resistance of the external air layer and the static thermal transmittance are obtained from Table 2; the external thermal admittance and the time shift for the external side are obtained from Table 6. A calculation routine was written in Matlab®, where the properties of the m50 solution (Table 6) and the climate data for each location (Table 5) were input, giving as a result the indoor and outdoor temperature oscillation for the coldest and hottest months of the year. No HVAC systems were considered, so the mere effect of the walls on the indoor environment could be clarified. Data were depicted graphically; the x axis was extended to 36 h and the scales of the y axis were unified for an easier comparison (Figure 5).

**Figure 5.** *Cont*.

ǻǼ

**Figure 5.** Thermal oscillation of m50 solution, in winter and summer, for representative cities of the seven climate zones of Chile.

Two aspects of the results from these simulations can be commented on: Time shift and decrement factor. In general, this solution works better in climates with large thermal oscillations between day and night, such as Lonquimay or Santiago. During the cold season, indoor temperatures are never lower than outdoor temperatures; while on the contrary, during the hot season, thermal inversion is observed, and the walls can mitigate the low temperatures during the first few hours in the morning and also help in weathering the peak in temperature during the middle of the day. In regions with low thermal oscillation, whether we are talking about cold warm (Antofagasta) or cold climates (Punta Arenas), the effect is not so evident, although indoor temperatures are always a couple of degrees above outdoor temperatures during the cold season.

#### *3.2. Thermal Comfort and Energy Saving Analysis*

Dynamic simulations have been developed to establish the impact of glass aggregate mortars on the thermal behavior of social housing in the different thermal zones. The substitution of conventional mortar bricks for glass aggregate mortars can signify anywhere between a 1% and 14% reduction in the enclosure's thermal transmittance, maintaining the same total thickness and insulation (Table 7). This reduction in thermal transmittance, along with a higher thermal inertia, means that the use of said material in zones Z1, Z2, Z3 and Z6, results in an increase in thermal comfort hours when the dwelling operates under free oscillation. These are higher for the m50 mortar, ranging from 65 to 199 h, while they range from 23 and 103 for the m25 mortar. Therefore, the increase in recycled glass percentage has a direct relationship on the increase in hours in comfort for climates where the increase in thermal inertia this material provides, can be taken advantage of.


**Table 7.** Hours in thermal comfort and distance in respect to the neutral temperature by zones considering the mortar used.

There are thermal zones, such as Z4 and Z5, where it is necessary to exceed 25% in the substitution of aggregate for glass so that there is a positive impact; however, this impact is not as important as in zones Z1, Z2, Z3 and Z6. For zone 7 (Punta Arenas), none of the glass aggregate mortars have an increase of comfort hours due to their low daily oscillation during the coldest months, just as was commented in Section 3.1. However, in zone 1 (Antofagasta), there is a significant increase in hours, despite this not being a climate with major thermal oscillations. This is due to the proximity of outdoor hours to the comfort limits (See Table 5).

The distance regarding neutral temperature marked by the ASHRAE adaptive comfort model has also seen important reductions in zones 1, 2, 3 and 6, with the m50 mortar representing almost double the reduction compared to the m25 in these zones. Zone 7, just as in the difference in the results of comfort hours, does not see a reduction in this indicator, as the distance increases by 401 ◦C for m50 and 531 ◦C for m25. Zones Z4 and Z5 only have a reduction of the distance when the m50 mortar is used, and end up being −183 ◦C and −122 ◦C, respectively (See Table 7).

These results are associated with a higher thermal comfort and lower overheating issues in dwellings for zones 1, 2, 3 and 6 when these operate under free oscillation. Said results can also assume a reduction in respect to heating and cooling consumptions, as well as in demand peaks due to a lower thermal oscillation inside the dwelling, linked with a higher thermal inertia of the enclosures. In Table 8, it can be seen that heating peak demands can be reduced by between 0.1% and 7.6%, with zones Z1 and Z2 providing a higher reduction (2.6–7.6%). The same occurs with the heating demand, with reductions of between 3.4% and 15.3%. In these zones, the use of an m50 mortar assumes a 50% increase in consumption reduction compared to the m25, with Z1 passing from 2667 to 2559 KWh/year and Z2 from 7047 to 6557 KWh/year.

The cooling demand and peaks also see important reductions in zones Z1, Z2 and Z3 (in the rest of the zones there is no cooling demand) (See Table 8). The reductions of the cooling peak oscillate between 4.8% and 12.2%, falling in the most favorable zone, Z1, from 4.14 to 3.64 kWh. The reductions for cooling demand are between 7.3% and 9.8%, with reductions seeing a noticeably similar percentage among the three zones. However, the kWh reduction oscillates between 150 and 366 KWh/year, with the zones providing a better performance, with glass aggregate mortar being in zones 2 and 3, respectively.


**Table 8.** Heating and cooling demand and peak by zones, depending on mortar used.

It can be seen by these results that the thermal and energy benefits of using glass aggregate mortars are closely tied to the percentage of glass incorporated, the thermal zone and the dwelling's type of operation. The m50 mortar can imply an increase in the hours of comfort and lower indoor thermal oscillations to avoid overheating in zones 1, 2, 3 and 6. When the dwelling operates with air-conditioning systems, the substantial differences of using these mortars would be in zones 1, 2 and 3, with zone 2 producing the highest saving, with a total reduction of 857 kWh/year, bearing in mind both heating and cooling. However, it should also be remarked that this material might have a controversial effect in the coldest zones, giving as a result a slight increment in the heating demand.

#### **4. Conclusions**

In this research, mortar blocks with doses of 25% (m25) and 50% (m50) of recycled glass aggregate were analyzed from a thermal point of view. This assumes a tangential vision to the traditional analysis based on mechanical behavior. The approach to the analysis focuses on social housing, using 21 models located in the 7 thermal representative zones of Chile. Using a methodology based on analyzing periodic thermal transmittance, adaptive comfort levels and energy demand, the thermal analysis is addressed holistically, making it possible to extrapolate this methodology to different construction solutions.

On considering thermal transmittance results, it can be pointed out that with the m50 solutions, this is reduced by 22%, the decrement factor is reduced by 10%, and the time shift increases by approximately one hour. Meanwhile, locations with a higher thermal amplitude between day and night like 6 (Lonquimay) and 3 (Santiago), both in winter and in summer, see significant reductions when compared with traditional solutions without a recycled glass aggregate.

Regarding thermal comfort, we can conclude that an increase is detected in the hours of comfort in all zones except in zone 7 (Punta Arenas), increasing 199 h in zone 1 (Antofagasta) with the m50 solution. The extension effect of hours of comfort and the reduction of distance, compared to the neutral temperature, is accentuated in zones 1 (Antofagasta), 2 (Valparaiso), 3 (Santiago) and 6 (Lonquimay). The m50 solution increases hours of comfort by almost 50% when compared with m25 and reduces the distance in respect to the neutral temperature by half.

Regarding energy demands, the most relevant finding is the reduction of the peak demand in all thermal zones, from 0.1% (heating) with m25 in zone 7, to 12.2% (cooling) with m50 in zone 1. The annual energy demand reduction is seen in zones 1 (Antofagasta), 2 (Valparaiso) and 3 (Santiago), with up to 15%. On this occasion, the m50 reduces the demand by around 50%, compared to the m25.

This research is a starting point to be valued in decision-making when it comes to implementing a construction solution with recycled materials in a country. In future work, the simulation results should be validated with actual prototypes. Further research is needed to generate an implementation methodology of new construction solutions which consider energy behavior, bearing in mind the energy poverty levels of social housing and that this cannot be done in any other way than with low environmental impact solutions.

**Author Contributions:** All the authors contributed equally to the present research. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research as well as The APC were funded by National Agency for Research and Development (ANID) of Chile, grant number 1200551.

**Acknowledgments:** The authors would also like to acknowledge that this paper is part of the project "Conicyt Fondecyt Regular 1200551 - Energy poverty prediction based on social housing architectural design in the central and central-southern zones of Chile: an innovative index to analyze and reduce the risk of energy poverty" funded by the National Agency for Research and Development (ANID). In addition, we would like to acknowledge to the research group "Confort ambiental y pobreza energética (+CO-PE)" of the University of the Bío-Bío for supporting this research. Authors also wish to express their gratitude to the General Research Center at the University of Seville (CITIUS).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Prediction of Cooling Energy Consumption in Hotel Building Using Machine Learning Techniques**

#### **Marek Borowski \* and Klaudia Zwoli ´nska**

Faculty of Mining and Geoengineering, AGH University of Science and Technology, 30-059 Kraków, Poland; kzwolinska@agh.edu.pl

**\*** Correspondence: borowski@agh.edu.pl; Tel.: +48-12-6172068

Received: 22 October 2020; Accepted: 24 November 2020; Published: 26 November 2020 -

**Abstract:** The diversification of energy sources in buildings and the interdependence as well as communication between HVAC installations in the building have resulted in the growing interest in energy load prediction systems that enable proper management of energy resources. In addition, energy storage and the creation of energy buffers are also important in terms of proper resource management, for which it is necessary to correctly determine energy consumption over time. It is obvious that the consumption of cooling energy depends on meteorological conditions. Knowing the parameters of the outside air and the number of users, it is, therefore, possible to determine the hourly energy consumption of a cooling system in a building with some accuracy. The article presents models of cooling energy prediction in summer for a hotel building in southern Poland. The paper presents two methods that are often used for energy prediction: neural networks and support vector machines. Meteorological data, time data, and occupancy level were used as input parameters. Based on the collected input and output data, various configurations were tested to identify the model with the best accuracy. As the analysis showed, higher prediction accuracy was obtained thanks to the use of neural networks. The best of the proposed models was characterized by the *WAPE* and *CV* coefficients of 19.93% and 27.03%, respectively.

**Keywords:** energy consumption; heating and cooling system; optimization and management; energy use prediction; neural network; support vector machine

#### **1. Introduction**

Nowadays, people spend the majority of time indoors, which leads to increased costs associated with maintaining comfort conditions in buildings. Internal installations such as heating, cooling, and ventilation systems therefore play a key role and thus constitute the main source of costs. Hence, the combination of economic and environmental factors is an important task for manufacturers and designers, contributing to the development of new solutions to provide comfort conditions for small operating and investment outlays. In 2018, the industry sector accounted for about 32% of total global energy consumption. The transport sector accounted for 28% of the energy use. The building construction and operations accounts for the largest share of global final energy use—36%, of which 8% was connected with operation of the non-residential building, 22% with residential building operation and 6% with the construction industry [1]. Global final energy consumption in buildings [2] in 2018 increased 1% from 2017, and about 7% since 2010. According to the Energy efficiency indicators Report published by IEA, nearly 50% of building consumption is related to space heating and 4% with cooling. Increasing energy consumption in the European Union led to appearance of the general regulation in this field—Directive 201013/EU of the European Parliament and of the Council on the energy performance of buildings [3]. According to Article No. 8, Member States should set system requirements for the purpose of optimizing the energy use of technical systems in the buildings. Regulations cover at least

heating, hot water, air-conditioning, and large ventilation systems. Furthermore, Member States may encourage the use of active control systems such as automation, control, and monitoring systems that aim to save energy.

In general, methods for estimating energy use have two purposes: design or optimization of the building and HVAC systems (forward modeling), and calculating retrofit savings or implementing model predictive control in existing buildings (data-driven modeling). Behavior of the system is described by a mathematical model, which includes input variables, system structure (physical description), and output variables (reaction to the input variables). Data-driven modeling may be divided into three groups: "Black-Box" (Empirical), Calibrated Simulation, and Gray-Box Approach, which differ in data requirements, time, and effort required to develop the appropriate models. In the first method, a simple or multivariate regression model that describes a relationship between measured energy use and the various input parameters is constructed. A Calibrated Simulation Approach uses a simulation computer program to evaluate existing buildings' energy consumption and then calibrates the physical input parameters to the program. The Gray-Box Approach formulates a physical model and identifies important parameters by statistical analysis. This method is a mixture of physics-based and data-driven methods, and it could be implemented for fault detection and diagnosis (FDD) and online control [4,5].

Building energy consumption prediction is crucial to appropriate energy management, thus improving energy efficiency of systems and performance of the buildings. Generally, for building energy consumption prediction, two techniques are used: statistical methods and artificial intelligence methods. In recent years, artificial intelligence methods have become very popular. This technique is often applied to the prediction of energy consumption due to good, accurate prediction results [6]. Among the most popular data-driven prediction models using empirical approach modeling are artificial neural networks (ANNs) and support vector machines (SVM). One of the popular techniques is also decision tree (DT) and random forest (RF), which generates multiple decision trees that operate as an ensemble [7]. To improve their solutions, many authors use various methods mentioned above and choose the results of the best one [8–10]. The literature contains numerous interesting solutions using different methods. Table 1 illustrates applications of certain algorithms in literature.



**Table1.**Areviewofpredictivemodelsforbuildingenergyconsumption.

Jovanovi´c et al. [22] presented a prediction of heating energy using various neural networks: feed forward backpropagation neural network (FFNN), radial basis function network (RBFN), and adaptive neuro-fuzzy interference system (ANFIS). The subject of analysis was the university buildings in Norway. The authors have predicted building energy use based on the input feature: mean daily outside temperature, mean daily wind speed, total daily solar radiation, minimum daily temperature, maximum daily temperature, relative humidity, day of the week, month of the year, and heating consumption of the previous day. Data for the working days in the cold period for three years was used. For model FFNN, all input variables mentioned above are used; for model RBFN, seven most influencing parameters, and for model ANFIS, only three of them. The results showed that all three models have very good agreement with measured values. Sha et al. [16] presented a simplified energy prediction method based on the three input features: degree-day, day type, and month type. Their study adopted three machine learning algorithms: MLR, SVR, and ANN. The results showed that ANN and SVR methods have better performance than the MLR model. The authors mentioned that all of the methods do not have sufficient quality in heating energy prediction, due to the size of the training dataset. Manjarres et al. [20] proposed to implement the HVAC energy management system in a separate part of the office building in Spain. Solution includes a two-way communication system, enhanced database management system, and a set of machine learning algorithms based on random forest (RF) regression techniques. The proposed optimizer included information such as: indoor and outdoor temperatures, relative humidity, and occupancy level. The simulations took into account different modes of operation of HVAC systems. Data were collected from 63 days in summer and 46 days in winter, which was the basis during the simulation phase. The solution assumes the ON/OFF operation of the HVAC system and the operation of the mechanical ventilation system in accordance with the proposed solutions, which is to ensure minimization of energy consumption while maintaining the assumed temperature inside the rooms. The authors estimate that the implementation of the described system will contribute to the reduction of heat demand by 48% and 39% for cooling consumption. Ahmad et al. [21] presented a comparison between feed-forward back-propagation artificial neural network and random forest for HVAC electricity consumption of a hotel building in Madrid. Results showed that ANN performed marginally better than RF. Generally, both models have comparable quality of prediction and could be implemented in the building system.

In order to increase the accuracy of forecasting, as well as to expand the possibility of implementing solutions, many authors decide to modify classic prediction models or combine several approaches. The biggest problem of these patterns is the nonlinearity of relationships. Zhong et al. [18] proposed using a novel vector field-based support vector regression method. The purpose of the method is to define the optimal feature space by modifying the input data space. The resulting algorithm is then used to build a predictive model. The implementation of such a solution in an office building in China gave very good results compared to commonly used methods. The algorithm proposed in this article is used to determine the refrigeration load forecasting model based on the integrated data set. The input parameters are both external and internal. Casteleiro-Roca et al. [15] described an intelligent hybrid model to predict the short-term energy demand in a hotel, including three techniques: clustering, MLP, and LS-SVR. It was used for predicting the power load of the building for each hour in a 24-h horizon. The authors identified three input variables: the energy demand in the previous 24 h, the mean temperature of the previous day, and the occupancy rate of the hotel. The obtained results were compared with conventional forecast techniques based on ARIMAX modeling and a method based on tree models. The hybrid appeared to have better accuracy (lower mean absolute error) than the above-mentioned models.

Artificial Intelligence Algorithms are also widely used for electricity consumption forecasting, including district public consumption [23]. Güngör et al. [24] presented electricity consumption prediction for a variety of households using different prediction algorithms (Holt-Winters, ARIMA, LSTM i TESLA). Results showed that TESLA performance is better than other prediction methods. The authors have used five different classifiers (Logistic Regression, Stochastic Gradient Descent,

K-Nearest Neighbors, Random Forest, Support Vector Machine). Another example of such algorithms is the forecast of meteorological conditions for the needs of the HVAC system proposed by I¸sik and Inalli. A back-propagation neural network perceptron model with seven inputs was used to predict temperature, solar radiation, and relative humidity. A comparison with the ANFIS model showed that the ANN model has better accuracy in terms of forecasting meteorological data for HVAC. [25]

In this study, cooling energy consumption predictive models based on an artificial neural network and support vector machine algorithms are presented. The paper includes a statistical analysis of historical data of hotel building. The cooling load and proposed variables during the summer season are considered. The main contribution of this paper is the development of high-accuracy predictive models and comparison neural network-based model with the support vectors approach. The structure of this study was organized as follows. Firstly, an introduction to energy prediction is presented. A short review of the prediction method was included. Section 2 provided the case of the study and used methods description, including the details of data collection, input variables, and proposed models. In the Results section, the main results obtained from models are presented. In these sections, real energy consumption and predicted load are compared and analyzed. Finally, conclusions and remarks are given in Section 5.

#### **2. Materials and Methods**

#### *2.1. Case of Study*

The Turówka hotel building located in Wieliczka near Kraków (the south-central part of Poland) is used as a case of study for this paper. The five-story building is a reconstruction of a historic salt store. The hotel has a floor area of 5525.00 m<sup>2</sup> and a volume of 19,300 m3. Facilities include 50 double rooms, a hotel bar, a restaurant, a drink bar, a conference room, and a pool. A detailed description of the building with an analysis of the cooling and heating load was described in the previous article [26]. Based on the analysis of the summer period in the hotel building, it was found that the cooling system is responsible for 50–60% of the total energy load. Additionally, the analysis showed a clear relationship between the outside temperature and the cooling load; therefore, the presented predictive model applies to the cooling energy prediction during the summer. The data used in the research mainly include three parts: meteorological data, load data, and data related to operational conditions. The input variables used for the predictive models for the calculation of the cooling demands include weather conditions, occupancy level in the hotel, hour, and day of the week. The cooling energy load data is provided by meters systems installed in the building cooling system i.e., feed and return of the high and, depending on demand, the low parameter of the refrigerant. Data transmitted via a serial communications protocol—MODBUS RTU—is stored in a recording system. The measurement system consists of MULTICAL heat meters by Kamstrup and flow sensors submitted to a type of approval according to EN 1434 [27], which includes the 2400-h measurement stability test of the flow sensors. The meteorological data used in this paper are obtained from the National Research Institute—Polish Institute of Meteorology and Water Management; only the outside temperature was measured directly at the hotel area. The data used in predictive models are an hourly time series collected in a summer season in Poland, where the cooling load was observed. The hourly meteorological data were calculated as an average value based on the measurement with sampling time set to 10 min. Data directly from the analyzed object, i.e., outdoor temperature and cooling energy consumption, were collected with a sampling time of 1 min. The data used for the models were calculated as an arithmetic average of measurement from an hourly period of time. For this study, the measurement season was selected from 15 May to 15 September 2019. The framework of the case study is shown in Figure 1.

**Figure 1.** The framework of the case study.

#### *2.2. Methodology*

#### 2.2.1. Methodology of Artificial Neural Networks

There are many types of artificial neural networks (ANN) including simple feed forward networks, recurrent neural networks, and spiking neural networks, RBFNs. One of the most commonly used models is the multi-layer back-propagation neural network (BPNN). The BPNN architecture includes three types of the layers: an input layer (variable), an output layer (predicted value), and a hidden layer. A basic processing unit in this model is a neuron. A schematic diagram of an artificial neural network structure which consists of all three layers is shown in Figure 2.

For Multi-layer Perceptron (MLP), neurons in the input layer distribute the input signals *xi* to neurons in the hidden layer. Each neuron *j* in the hidden layer sums up its input signals *xi* after weighting them with the strengths of the respective connections *wji* from the input layer and computes its output *yj* as a function *f* of the sum [28]:

$$y\_j = f(\sum w\_{ji} x\_i) \tag{1}$$

In this study, the Statistica Artificial Neural Network Package was used. A partition of the data into training (70%), validation (15%), and test (15%) is carried out.

**Figure 2.** An illustration of a typical ANN topology.

#### 2.2.2. Methodology of Support Vector Machines

Support Vector Machine is a supervised learning method increasingly used in solving nonlinear problems. This method is commonly used for classification, regression, and clustering. One of the main features of this model is the lack of local minima. It consists of two main parts: universal linear learning algorithm and a specific kernel that calculates the inner product of input points in feature space [29].

The universal linear function is as Equation (2):

$$\mathbf{f}(\mathbf{x}) = \boldsymbol{w}^{\mathrm{T}} \ \boldsymbol{\varphi}(\mathbf{x}) + \boldsymbol{b} \tag{2}$$

where f(x) means the forecasting values and the coefficients *w* and *b* are adjustable.

The main aim of SVM is to find the optimal hyperplane between classes, with the maximal margin. The margin is defined as the distance between the closest point in each class and hyperplane. For this purpose, the ε-insensitive loss function is used. Minimizing the overall errors is expressed by Equation (3) [30]:

$$\min\_{w,b,\xi^\*,\xi} R(w,\xi^\*,\xi) = \frac{1}{2}w^Tw + \mathbb{C}\sum\_{i=1}^N \left(\xi\_i^\* + \xi\_i\right) \tag{3}$$

with the constraints:

$$\varepsilon\_i \mathbf{y}\_i - \varepsilon \mathbf{w}^T \boldsymbol{\varphi}(\mathbf{x}\_i) - \mathbf{b} \le \boldsymbol{\varepsilon} + \boldsymbol{\xi}\_i^\* \tag{4}$$

$$
\varepsilon - y\_i + w^T q(\mathbf{x}\_i) + b \le \varepsilon + \xi\_i \tag{5}
$$

$$
\mathbb{L}\_{i'}^\* \mathcal{L}\_i \ge 0 \tag{6}
$$

The method of operation of the support vector machine is shown in Figure 3.

There are four types of kernel function linear, polynomial, radial basis function (RBF), and sigmoidal function [29]. The most used kernel functions are the Gaussian RBF with a function of kernel defined by Equation (7):

$$\mathbf{K}(\mathbf{x}\_{i}, \mathbf{x}\_{\circ}) = \exp(-\gamma |\mathbf{x}\_{i} - \mathbf{x}\_{\circ}|^{2}) \tag{7}$$

where xi and xj are vectors in the input space and γ is kernel parameter. An equivalent definition involves a σ parameter, where γ = 1/2σ2.

**Figure 3.** An illustration of a typical ANN topology.

The forecasting accuracy of the SVM model is affected by hyperparameters. In the ε-SVR, three proper parameters need to be determined: C, ε, and the kernel parameter (γ). Parameter ε represents the width of the ε-insensitive loss function and can affect the number of support vectors in the model. The higher ε value results in fewer support vectors and more flat estimates. Parameter C is related to model complexity and the degree of the deviations larger than ε which are tolerated. In models with high C parameter values, the main objective is to minimize the empirical risk only, without regard to model complexity part in the optimization formulation. Parameters σ and γ describe the Gaussian function width [31]. In this paper, the radial basis function RBF function is used as a kernel function to estimate the cooling load of the hotel building. A grid-search technique was applied to find the optimal parameter values. The kernel parameter γ was selected in priori based on the literature. According to the Limsvm-2.6 [32], the kernel parameter is defined as γ = 1/n, where n means the number of input variables. For d-dimensional problems, Cherkassy and Ma [31] noticed that the width parameter σ depends on the number of input variables d in accordance with the formula σd~(0.2–0.5). On this basis, four values of γ parameters were found: 0.1, 0.3, 0.5, 0.7. The ten-fold cross-validation was applied to reduce the error of the model. The dataset was randomly divided into two parts: learning samples (75%) and testing samples (25%). The feasible ranges of the parameters are set as follows: C ∈ [0.5, 150] and ε ∈ [0.01, 0.5].

#### *2.3. Model Evaluation Index*

The performance of the proposed models is evaluated by the mean absolute error (*MAE*), root mean square error (*RMSE*), weighted absolute percentage error (*WAPE*), coefficient of variance (*CV*), and coefficient of determination (*r*) [11,12,18,33]. The indicators are calculated as follows:

$$MAE\ \ = \frac{1}{n} \sum\_{i=1}^{n} |E\_A - E\_P|\tag{8}$$

$$RMSE \;= \sqrt{\frac{1}{n} \sum\_{i=1}^{n} (E\_A - E\_P)^2} \tag{9}$$

$$WAPE\ = \frac{\sum\_{i=1}^{n} |E\_A - E\_P|}{\sum\_{i=1}^{n} E\_A} \tag{10}$$

$$CV = \frac{\sqrt{\frac{1}{n}\sum\_{i=1}^{n}(E\_A - E\_P)^2}}{\overline{E\_A}} \tag{11}$$

$$r = \frac{cov(E\_{A'}E\_P)}{\sigma\_{E\_A}\sigma\_{E\_P}}\tag{12}$$

where *n* denates the entire number of observations, *EA* is the actual value, *EA* denotes the mean of actual values, and *EP* represents the predicted value.

#### **3. Results**

#### *3.1. Preliminary Statistical Analysis*

The first step in the analysis was a preliminary statistical analysis of the dataset. From the available data, nine parameters were selected that could potentially have an impact on cooling energy consumption. Apart from the most obvious values, such as temperature, air humidity, wind speed, and relative humidity, it was decided to use time variables such as the hour and day of the week as well as the occupancy level. Table 2 presents the information summary of the data used in this work.

**Table 2.** Descriptive statistics for input and output variables.


In the beginning, the relationship between cooling energy consumption and the values of the variables was determined. Graphs for each of the variables are presented in Figure 4.

**Figure 4.** Relationship between cooling energy consumption and selected variables.

The Pearson correlation coefficient was the key parameter determining the acceptance of a given variable for analysis. Coefficients for each between the studied variables and the predicted output were provided. The results are summarized in Figure 5.

**Figure 5.** Matrix of the Pearson correlation coefficients.

Variables with low correlation coefficients can be significant in more complex predictive models. Based on the analysis of the correlation coefficient, the rejection of two parameters was initially planned: day and occupancy level. Due to the low correlation coefficients of the variables which, according to the authors, may have a significant impact on the predicted values, a more detailed analysis of the cooling load variability overtime was performed. Figure 6 shows plots of variation over time for the raw data collected with a sampling time of 1 min, and also after taking the hourly mean which was then accepted for prediction models.

As shown in Figure 6, there is a clear relationship between the demand for cooling energy and the time it was recorded. This relationship is not linear, hence the low correlation coefficient in the previous analysis step. The adoption of averaged values for further analysis may of course affect the accuracy of prediction, due to the unstable variability of the demand value; however, due to practical aspects, it was decided that the hourly variables were more efficient and could give sufficient effects in the prediction models. Despite the low correlation coefficients for time parameters and the occupancy level, it was decided that these parameters may be of great importance for a hotel facility due to the variability of the energy load depending on the use by guests.

**Figure 6.** Hourly variation of cooling energy consumption: (**a**) for the sample days (30 May, 30 June, 30 July, 30 August) based on data sampled every minute; (**b**) for example days (30 May, 30 June, 30 July, 30 August) on the basis of hourly average data; (**c**) mean values grouped by hour and categorized by month based on data sampled every minute.

#### *3.2. Prediction Models*

#### 3.2.1. Artificial Neural Networks

As mentioned in Section 2.2.1, the Statistica Artificial Neural Network Package was used to prepare predictive models. The data was divided into three groups. In addition, 70% of the data was used for the design of the network, 15% was used for validation, and 15% for testing. The hidden layer activation-functions, output layer activation-function, and the number of hidden neurons are selected using the methodology based on statistical tests and least-squares estimation. Models were created by combining different types of activation functions and a different number of hidden neurons. The selection of the function was made from identity, logistic, hyperbolic tangent, and exponential function. To find the optimum number of hidden neurons, various numbers of neurons were examined. The various combination of activation functions and numbers of hidden neurons as described above were tested. Sum-of-squares was selected as the error function during the network training process. Training Algorithm BFGS (Broyden–Fletcher–Goldfarb–Shanno) was chosen for this work. Five models with the best accuracy were selected and described in Table 3.


**Table 3.** Network configurations tested.

Figure 7 shows the schemes of regressions for all data sets according to each of the MLP models. The plots explain the correlation between the real values and the MLP model output. The solid line in each plot represents the best linear fit between the output and target values.

**Figure 7.** Comparison between real cooling energy consumption and prediction of SVM model: (**a**) for the model MLP-1; (**b**) for the model MLP-2; (**c**) for the model MLP-3; (**d**) for the model MLP-4; (**e**) for the model MLP-5.

In Table 4, sensitivity coefficients are presented. The analysis describes the change in the system's outputs due to variations in the parameters that affect the system. Performing the sensitivity analysis consists of controlling how the network error behaves in the case of fluctuations of the independent variables. For each input variable, its values are converted to the mean (from the training set) so that it does not contribute any information to the model. After supplying such modified data to the network input, the final prediction error is checked. A larger error value means that the model depends on the proposed variable. The higher the value of the sensitivity analysis coefficients, the greater the importance of a given variable for a good fit of the model.


**Table 4.** Sensitivity analysis of inputs.

According to the statistical analysis and the analysis of the sensitivity of neural networks, a clear impact on energy consumption is noticeable in the case of parameters such as relative humidity, maximum wind speed, occupancy level, hour, and day of the week

#### 3.2.2. Support Vectors Machines

As mentioned above, ranges of gamma, C, and epsilon parameters were defined and tested in the search for optimal values. Below, in Table 5, models with given characteristic parameters are defined which were characterized by the best fit during the tests.


**Table 5.** Network configurations tested.

In Figure 8, the relationship between the real energy consumptions and the predicted values was plotted. The line of a theoretical perfect 1:1 match line was marked in red.

**Figure 8.** Comparison between real cooling energy consumption and prediction of SVM model: (**a**) for SVM 1 model; (**b**) for SVM 2 model; (**c**) for SVM 3 model; (**d**) for SVM 4 model; performance of models.

Based on the coefficients defined in Section 2.3, selected models of neural networks were compared. As previously mentioned, five matching indexes were selected for this purpose: *MAE*, *RMSE*, *WAPE*, *CV*, and correlation coefficient (*r*). The results for cooling consumption are presented in Table 6.


**Table 6.** Performance evaluation of different ANN models for cooling energy consumption.

Similarly, for selected SVM models, the coefficient values are summarized in Table 7.

**Table 7.** Performance evaluation of different SVM models for cooling energy consumption.


Figure 9 shows comparison of the real values and predicted results of the MLP and SVM models with the high accuracy; it is MLP 2 and an SVM 4 model.

**Figure 9.** Prediction performance during cooling season of the model with highest accuracy (**a**) neural network MLP 2; (**b**) Support Vector Machine SVM 4.

The figure shows the real values of hourly consumption of cooling energy in the analyzed period with a dashed gray line. The blue color marks the results obtained with the MLP 2 model, which received the best results among the proposed ANN networks, and the red color indicates SVM 4, analogically the best among the presented SVM models. In both of the proposed models, the worst effects present at predicting extreme values, where the cooling load is very high or very low. Comparing the models with each other, it is noticeable that neural networks perform better in this matter, and the difference between the real load and the predicted value is significantly smaller.

#### **4. Discussion**

The analysis of the cooling energy load in the hotel building showed that, based on several parameters characterizing the external conditions, building using conditions and time, it is possible to estimate the overall cooling load of the building. The analysis used nine parameters, including hour, day of the week, average temperature, occupancy level, wind direction, average wind speed, maximum wind speed, total hourly precipitation, and relative humidity. The external temperature has an obvious influence on energy consumption, which determines the demand for cooling energy. As shown by the preliminary analysis, relative humidity, wind speed, and direction can also be considered significant factors. Despite the low correlation coefficients of the time parameters (hour and day of the week) and the occupancy level, it was decided that, in the general model, they may be significant; therefore, they were not omitted in further analysis. Based on the assumed parameters, five models of neural networks based on the MLP algorithm and four SVM models with different configurations of parameters defining

the models were proposed. Five model evaluation indicators were used to check the quality of the models shown: the mean absolute error (*MAE*), root mean square error (*RMSE*), weighted absolute percentage error (*WAPE*), coefficient of variance (*CV*), and coefficient of determination (*r*). Among all the proposals, a better fit was found for the ANN models, the correlation coefficient of which did not fall below 0.9. As shown by the sensitivity analysis of inputs (Table 4), time and occupancy level played an important role in the model, despite the seemingly low importance of these parameters. Total hourly precipitation had the least impact on the overall result. The selected neural networks are characterized by the *WAPE* coefficient ranging from 19.93 to 22.77% and the *MAE* coefficient of 10.27 to 11.80 kWh/h. The best of the proposed models, i.e., MLP 2, achieved the value of the correlation coefficient *r* at the level of 0.93. For comparison, the mean absolute error for the proposed SVM models varies from 13.13 to 14.46 kWh/h, and the weighted absolute percentage error from 25.45 to 28.04%. The most precise SVM model—SVM 4—is characterized by a correlation coefficient of 0.88. The proposed prediction methods are widely used for energy prediction due to the possibility of their application in nonlinear dependencies of many variables. As the results show, the use of neural networks in energy prediction enables the achievement of better model fit coefficients. Both methods used were based on the nine variables mentioned above. The greatest differences between the proposed models are visible in the extreme values. The SVM model performs much worse at very low and very high cooling loads. In the event of high values of cooling energy consumption, both the ANN and SVM models lower the predicted values compared to the actual values. This is especially noticeable with the SVM model, which directly affects the lower fit factors.

#### **5. Conclusions**

Reducing energy consumption is an issue that is becoming more and more popular nowadays. This is related to both the obvious economic aspects as well as the growing awareness of the society regarding the limitation of human impact on the environment, including the exploitation of non-renewable resources and emissions of pollutants into the environment. There are new solutions on the market to reduce the energy demand of buildings. Systems of proper management of energy demand, especially in combined energy economies, are also becoming more and more popular. The key issue in such solutions is the ability to accurately determine the demand for energy at a given moment. As a consequence, machine learning techniques are also becoming more and more popular, used to predict the energy load of a building.

The article presents two popular methods of energy prediction: neural networks and support vector machines. For each of the methods, several models differing in characteristic parameters and selected optimization methods are presented. The subject of the analysis is a building of historical importance, the modernization of which is limited due to the minimization of interference with the cubature and the overall appearance of the facility. In addition to replacing the heat source and internal installations, one of the effective methods of reducing operating costs may be the proper management of energy installations, which requires a building energy consumption forecasting system. The models use parameters that may affect consumption, both characterizing external conditions, time of use, and the number of guests. Based on the collected data, five models of neural networks and four SVM models were presented. Each of them was compared according to the proposed forecasting accuracy coefficients. The analysis showed that, despite the initially insignificant influence of some parameters on the obtained results, they played an important role in the predictive model. The best fit coefficients were obtained for models using artificial neural networks to predict the heating load. The proposed model made it possible to estimate the demand for cooling energy with a matching coefficient of 0.93. The quality of the model can be improved by extending the analysis time to several summer seasons. Hotel buildings are a specific object due to the users who have different preferences as to the conditions inside the rooms. Individual control allows them to maintain thermal comfort by changing the temperature, and thus their behavior directly affects energy consumption for cooling purposes.

**Author Contributions:** Conceptualization, M.B. and K.Z.; methodology, M.B.; software, K.Z.; validation, M.B.; formal analysis, M.B.; investigation, K.Z.; resources, K.Z.; data curation, K.Z.; writing—Original draft preparation, K.Z.; writing—Review and editing, M.B.; visualization, K.Z.; supervision, M.B.; project administration, M.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Regional Development Fund, Intelligent Development Program, Grant No. POIR.01.01.01-00-0720/16.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

## **Assessment of the Impact of Occupants' Behavior and Climate Change on Heating and Cooling Energy Needs of Buildings**

#### **Gianmarco Fajilla 1,2,\* , Marilena De Simone <sup>1</sup> , Luisa F. Cabeza <sup>3</sup> and Luís Bragança 2,\***


Received: 9 November 2020; Accepted: 3 December 2020; Published: 7 December 2020

**Abstract:** Energy performance of buildings is a worldwide increasing investigated field, due to ever more stringent energy standards aimed at reducing the buildings' impact on the environment. The purpose of this paper is to assess the impact that occupant behavior and climate change have on the heating and cooling needs of residential buildings. With this aim, data of a questionnaire survey delivered in Southern Italy were used to obtain daily use profiles of natural ventilation, heating, and cooling, both in winter and in summer. Three climatic scenarios were investigated: The current scenario (2020), and two future scenarios (2050 and 2080). The CCWorldWeatherGen tool was used to create the weather files of future climate scenarios, and DesignBuilder was applied to conduct dynamic energy simulations. Firstly, the results obtained for 2020 demonstrated how the occupants' preferences related to the use of natural ventilation, heating, and cooling systems (daily schedules and temperature setpoints) impact on energy needs. Heating energy needs appeared more affected by the heating schedules, while cooling energy needs were mostly influenced by both natural ventilation and usage schedules. Secondly, due to the temperature rise, substantial decrements of the energy needs for heating and increments of cooling energy needs were observed in all the future scenarios where in addition, the impact of occupant behavior appeared amplified.

**Keywords:** occupant behavior; climate changes; energy needs; ventilation; residential buildings; DesignBuilder

#### **1. Introduction**

In most developed countries, buildings are the major energy consumers, and they may not be able to reach the new energy standards [1,2]. In the EU, most of the buildings have more than twenty years and present low energy performance [1]: The percentage of well-designed buildings is less than 2%, with almost 60% of heating systems inefficient and almost 40% of the windows being single glazed [3]. As recognized by the Energy Performance of Buildings Directive (EPBD) [4], buildings are responsible for 40% of the total energy consumption and 36% of global annual greenhouse gas emissions [3,5–7]; these consumptions could drastically increase double or even triple by 2050 if not faced in the right way [8]. As a consequence, governments worldwide have implemented energy requirements in their building regulations to reduce levels of energy consumed by buildings and to promote more energy-efficient envelopes and systems [9].

#### **2. Literature Review**

Nowadays, most researchers agree that occupant behavior plays an increasingly important role in building energy performance [10,11]. Despite the efforts made in improving the envelope of buildings and the efficiency of the systems, reducing energy consumption can be achieved considering also the impact that occupants' behavior (OB) has on buildings consumptions [12–18]. Furthermore, OB is often neglected or too simplified in energy design and assessment, causing large discrepancies between calculated and measured energy performances [12,19,20]. For example, a recent study conducted by Carlucci et al. [19] claimed that occupant behavior related to thermostat control (thermostat setpoints and operation schedules) is often too simplified in the building performance standards and calculation procedures, causing significant uncertainty in the predictions of building energy demand. Moreover, Mora et al. [20] simulated the energy consumption of a residential building considering three occupancy scenarios: Regulations, Current-use, and Statistical. Compared to the Current-use schedules, the Regulation schedules provided a significant underestimation of the heating energy needs, while the statistical schedules led to an overestimation. Different authors [13,14,18] highlighted that OB has an important responsibility in determining the energy consumption of buildings, pointing out that this impact is more significant in the new buildings where the envelope and the systems are optimized. Furthermore, Rouleau et al., in their work [15] claimed that the impact of OB has to be recognized to obtain a reduction in energy consumption. Because OB impacts in many ways on energy consumption (e.g., through heating and cooling systems or the interaction with windows and blinds), they deem that it should be not surprising if there is a huge gap between actual and prevised consumption. Furthermore, Zhang et al. [16] analyzed the role of occupant behavior in building energy performance, concluding that the energy-saving potential of occupant behavior in residential buildings is in the range of 10–25%. Similar results were obtained by [17] that quantify to 20% the achievable energy saving by modifying occupants' behavior using recommendations and feedback. Consequently, occupant behavior in buildings is becoming an increasingly topic so much so that different projects, performed within the framework of the International Energy Agency—Energy in Buildings and Communities Program (IEA-EBC), such as IEA EBC Annex 66 [21] and IEA EBC Annex 79 [22], focused on understanding and studying this issue.

The impact of OB on energy consumption of buildings is also recognized by the Intergovernmental Panel on Climate Change (IPCC) that in the IPCC AR5 [23] reported that factors of 3 to 10 differences can be found worldwide in residential energy use for similar dwellings, due to different usage of natural ventilation and thermal control of the indoor environment.

The reduction of buildings' energy consumption is a growing and global problem, mainly due to the looming threat of climate change. Goal 13 of the 2030 Agenda for Sustainable Development [24] calls for urgent action to tackle climate change and its impacts. Indeed, due to climate change and more frequent extreme events, buildings will have to deal with new climatic conditions for which they were not designed [25]. Thus, an increasing body of literature is now emerging on this topic. A recent work [26] assessed the scientific literature on the energy efficiency of buildings and the climate impact through a comparative analysis of Web of Science and Scopus. It was found that while most of the works focused on technologies for heating, ventilation, air-conditioning, and phase change materials, there is still a knowledge gap in the areas of behavioral changes, circular economy, and some of the renewable energy sources (e.g., geothermal, biomass, wind). The authors in [6] analyzed the impact of climate change on the energy performance of a zero energy building in Valladolid (Spain). Three future weather scenarios (2020, 2050, and 2080) were investigated, and the results showed a drop in the space heating demand and an increase in space cooling. Due to these consumptions' variations, they estimated an increase equal to 25% of the burning biomass to provide more energy to the absorption cooling system. Berardi and Jafarpur [27] assessed the impact of climate change on building heating and cooling energy demand of 16 building prototypes located in Toronto (Canada). Authors estimated for 2070 an average decrease of 18–33% and an average increase of 15–126% for the heating and cooling energy use, respectively. Ciancio et al. [28] simulated the energy performances

of a building in three cities (Aberdeen, Palermo, and Prague) considering three climatic scenarios (2020, 2050, and 2080). In general, decreasing trends for heating energy needs and increasing trends for cooling energy needs were obtained. The highest variations were observed for 2080: A reduction of the heating energy needs from −36% to −80% and an increase of cooling energy needs from +142% to 2316%. In another study, Ciancio et al. [29] analyzed the energy needs of a hypothetic building by varying its location in 19 cities with different climate conditions. The simulations performed for 2020, 2050, and 2080 showed, once again, a general decrease in heating energy needs and an increase in cooling energy needs. The authors highlighted that the effects of climate change will be more predominant in the Mediterranean basin than in other European areas. Same results were also found from other studies, such as [25], that argued that Southern Europe will be more vulnerable to climate change than Northern Europe. Furthermore, the authors in [30] studied the climate change-driven increase of energy demand in residential buildings in the area of Qatar, founding an increase equal to around 30%. They stressed how such an increase would cause higher CO2 emissions, more consumptions of water and fossil fuel, as well as an increase in the impact on the already strained local marine ecosystem. They also suggested renovating the building stocks and substitute fossil fuels with renewable energies (e.g., PV plants, wind farms, and tidal plants) as approaches to reduce the environmental impacts of climate change. Cabeza and Chàfer [8] published a systematic review of the technological options and strategies to achieve zero energy buildings contributing to climate change mitigation. Findings showed that buildings, if properly designed, can help to mitigate the impact of climate change—decreasing both the embodied energy in the materials, used during the construction phase, and the energy demand and use in the operation phase. Moreover, regarding new buildings, authors in [31] proposed an innovative method for designing buildings with robust energy performance under climate change for supporting architects and engineers in the design phase. To the extent of our knowledge, the effect of environmental (climate change) and behavioral variables (such as usage profiles and thermal comfort preferences) on the energy performance of buildings was investigated separately in the literature till now. What is missing are studies that consider both the influencing variables and provide predictions combining the double impact. Table 1 synthesizes the literature review related to this area highlighting: Subject of the study, outcomes and limitations, and considered impacts (occupant behavior/ climate change).



#### **Table 1.** *Cont*.

\* OB = Occupant behavior, CC = Climate Change.

#### *Aim of the Study*

As emerged from the literature review, the impact of occupant behavior, and the effect of climate change on the energy performance of buildings was largely recognized. Their impacts were investigated, highlighting the importance of future scientific contributions to these topics and encouraging more comprehensive studies considering that behavioral variables and climate change were still analyzed separately. Consequently, this paper aims to fill this gap by proposing a study that combines the double effect of these variables on the energy performance of buildings. By considering the information and indications of the available literature, the aim of this study was addressed to assess the impact of both occupants' behavior and climate change on the heating and cooling energy performance of a typical residential unit located in Southern Europe. Here, the energy performance was referred to as the heating and cooling energy needs defined as the heat to be delivered to, or extracted from, a conditioned space to maintain the intended temperature conditions during a given period of time. Energy needs constitute the base of calculation of the primary energy demand that is determined by the energy supply system and the user types of fuel.

In particular, the authors wanted to answer the following research questions:


The answers to these research questions can provide useful indications for scientists and policymakers to assess how human factors and environmental conditions can impact the energy consumptions of buildings, and consequently give due weight to them in future regulations and design criteria.

#### **3. Methodology**

The general schema and the consecutive steps of the investigation are illustrated in Figure 1.

**Figure 1.** Schema of the adopted methodology.

The research can be summarized in four steps:


Furthermore, this section introduces more in detail Step 1 to Step 3: The survey to collect information on the occupants' behavior to be used in energy simulations, the energy model of the residential unit investigated in the study, and the tool adopted to obtain the weather files of future climate scenarios.

#### *3.1. Questionnaire Survey*

Data of a questionnaire survey delivered in Southern Italy were used to obtain use profiles to be provided as input in energy simulations. During two survey campaigns conducted from 2017 [9] to 2019, 237 surveys were collected, and among them, 193 were accepted as valid for these analyses. The questionnaire presents a total of 64 questions grouped into three main categories, as shown in Figure 2.


**Figure 2.** Questionnaire contents.

Consistently with the aim of this paper, the attention was dedicated to the questions regarding the cooling and heating operation habits and the window opening preferences. The responses collected for the buildings located in Rende, characterized by Mediterranean climate conditions and defined as "Csa" according to the Köppen climate classification [32], were considered.

For the selected buildings, the schedules were first subjected to a cleaning process to verify their reliability. After that, the profiles were clustered based on the timing and length of the usage, and typical hourly profiles were obtained for heating, cooling, and natural ventilation.

#### *3.2. Case Study*

Among the collected sample, an apartment built in 2008 located on the second floor of a six-story building, with a gross floor area of 80 m<sup>2</sup> was chosen as a case study. The building structure is made of reinforced concrete, and the external walls consist of double hollow brick layers with an internal air gap partially filled with expanded polystyrene, resulting in a U-value of 0.6 W/m2·K. The windows are double glazing and a frame with thermal break. The heating system, used both for heating and DHW production, is an autonomous wall-mounted gas boiler. A zone thermostat regulates the operating of the heating system, and the heat emitters are aluminum radiators. The cooling system consists of air conditioners installed in the living room and in the bedrooms. METEONORM weather data [33] were used for the dynamic energy simulations conducted by DesignBuilder [34]. The model of the residential unit is shown in Figure 3.

**Figure 3.** The model of the residential unit: (**a**) DesignBuilder model of the building; (**b**) plan of the apartment.

The reliability of the model was verified by the authors in previous work [20] following the ASHRAE Guideline 14-2002 [35]. The predicted results obtained from the simulation of the actual use and the measured data extracted from the energy bills were compared on a monthly scale through the Normalized Mean Bias Error(NMBE) and the Coefficient of Variation of the Root Mean Square Error (CVRMSE). Values lower than the limit values were obtained for both the parameters.

Downstairs there is an unconditioned thermal zone; while upstairs, there is an adiabatic block, due to the presence of another heated dwelling. Horizontal and vertical overhangs were shaped through standard component block considered by the software in shading calculation. Three thermal zones (living area, bedrooms, and bathrooms) were considered, and the characteristic parameters were changed in terms of management of the heating and cooling system, as both activation period and setpoint temperature, and ventilation hourly profiles. The internal heat loads were determined following the indications of the Standard UNI/TS 11300-1 [36] that uses the relation:

$$
\phi\_{\rm int} = 7.987 \, A\_f - 0.0353 \, A\_f^2 \tag{1}
$$

where *Af* is the usable floor area of the house [m2]. The calculated value amounts to 5.56 W/m<sup>2</sup> and groups all contributions of occupancy, miscellaneous equipment, catering process, and lighting. The dynamic simulations were performed by combining different hourly ventilation profiles with heating and cooling operation schedules and setpoints temperature. In the reference case, energy

simulations were conducted for the heating (from 1 October to 30 April) and cooling (from 1 May to 30 September) season by considering the current climate data and a setpoint temperature of 20 ◦C and 26 ◦C, respectively. Further energy assessments were obtained by varying the climatic scenarios (2050 and 2080) and the internal setpoints temperature (18 ◦C and 22 ◦C in the heating season, 26 ◦C and 24 ◦C in the cooling season).

#### *3.3. Climate Scenarios*

In this study, the climate change world weather file generator (CCWorldWeatherGen) [37] was used to create the weather files of future climate scenarios. Several studies used this tool to obtain future weather files [6,25,27–30], and the authors in [38] presented a critical analysis of it. Specifically, CCWorldWeatherGen is a Microsoft Excel-based tool commonly used that, employing the morphing procedure [39], provides weather files for future scenarios using outputs from the UK Hadley Centre Coupled Model (version 3, HadCM3) [40].

The future scenarios selected for this study were 2050 and 2080. The three adopted climate weather files were first analyzed in terms of variations in the external air temperature values. Figure 4a shows the monthly average air temperatures of the current climate, while the Δ*T* between current and future monthly average air temperatures are reported in Figure 4b for 2050 and in Figure 4c for 2080.

**Figure 4.** Monthly average air temperature: (**a**) in 2020, and monthly average air temperature increment Δ*T* (**b**) in 2050 and (**c**) in 2080.

Compared to the current climate, an increase in the monthly average air temperatures for each month of both 2050 and 2080 is projected. In particular, increments from 1.2 ◦C observed in April to 2.8 ◦C in August and from 2.4 ◦C to 5 ◦C, in the same months, were expected for 2050 and 2080, respectively.

#### **4. Results and Discussion**

This section presents the results obtained from the survey and the energy simulations conducted for the heating and cooling season. The results are organized as follow:

• ventilation, heating, and cooling profiles obtained from the survey;


#### *4.1. Ventilation, Heating, and Cooling Profiles*

Tables 2–4 show the typical hourly profiles obtained for heating, cooling, and natural ventilation. Moreover, respondents declared to generally use the heating system from October to April with a typical setpoint temperature of 20 ◦C, and the cooling system from May to September with a setpoint temperature of 26 ◦C. Further setpoints temperature ranging from 18 ◦C to 22 ◦C in winter, and from 24 ◦C to 28 ◦C in summer, were encountered.

**Table 2.** Daily heating schedules (On = 1, Off = 0).

**Table 3.** Daily cooling schedules (On = 1, Off = 0).

The heating schedules varied in terms of both the duration and time of operation. The heating system could operate for 24 h (profile h1), for three hours during the evening (from 19:00 to 22:00) in profile h2, and during the morning (from 07:00 to 09:00) and in the evening (from 19:00 to 22:00) in profile h3.

**Table 4.** Daily Natural ventilation schedules (Open = 1, Close = 0).

As shown in Table 3, the cooling system was installed only in the living and bedroom zones and used with diverse daily schedules. In profile c1, it was used in the hottest hours of the day (from 12:00 to 18:00) in the living zone, and for two time ranges in the bedrooms (from 08:00 to 11:00, and from 14:00 to 17:00). The schedules of the cooling system were more similar between the zones with profile c2: In the afternoon (from 14:00 to 17:00) in the two zones, and in the late evening (from 22:00 to 01:00) only in the bedrooms. Profile c3 was different from the others because the cooling operation was only activated during the late afternoon: From 19:00 to 22:00 in the living zone, and from 22:00 to 07:00 in the bedrooms.

Usually, occupants welcome natural ventilation more than mechanical ventilation, where they can only passively accept the system operation [41]. On the other hand, natural ventilation impacts negatively on the energy needs of a building when the external air temperature is lower than the internal air temperature in winter, or higher in summer, producing greater values of heat losses. On the other hand, benefits from window openings can be obtained in summer when the external air is used for natural cooling during the late afternoon or at night.

Looking at the graphs, shown in Table 4, it can be seen the variations of the occupants' preferences related to ventilation through the seasons. Profile v1 was typical of families who preferred to use continuous hours of ventilation during the day from the morning to the afternoon. The daily schedules were equal among the rooms, but different in duration between the seasons: From 07:00 to 15:00 and from 07:00 to 19:00 in winter and summer, respectively. Profile v2 showed the use of the natural ventilation limited to the morning hours in winter (from 08:00 to 13:00) and concentrated in the coolest hours in the summer. Finally, profile v3 presented an intermittent, but prolonged use throughout the day in winter, and continuous use in the coolest hours in the summer (from 19:00 to 11:00). Similar habits could be seen in the bathrooms area in both v2 and v3 profiles where people used to leave the windows open for the entire day. Natural ventilation profiles, as well as heating and cooling profiles, are linked to occupancy. Generally, it is noted that in homes with greater hours of daily occupancy, there is a more frequent occupant-window interaction and prolonged use of the heating and cooling systems (e.g., heating schedule h1 with continuous activation).

The heating and cooling schedules were combined with the natural ventilation profiles, and nine profiles, both for winter and summer seasons, were applied to perform the energy simulations of a residential unit.

#### *4.2. Monthly Hours of Operation of the Heating and Cooling System in the Climate Scenarios*

Due to the increase of the monthly average air temperature, it is also interesting to analyze how the hours of operation of the heating and cooling systems vary from the current climate to the future scenarios. In this study, "monthly hours of operation" was the sum of the hours in which the heating/cooling system provides the energy necessary to reach and maintain the indoor temperature at the setpoint value.

In particular, Figure 5 shows the monthly hours of operation of the heating system in the current climate (Figure 5a) and the differences (Δ*h*) with respect to 2050 (Figure 5b) and 2080 (Figure 5c), by setting the internal air temperature value equal to 20 ◦C. The energy simulations were performed by considering all the heating schedules (h1, h2, h3) coupled with the ventilation profiles (v1, v2, v3).

**Figure 5.** Monthly hours of operation of the heating system: (**a**) in the current climate, and differences Δ*h* in the year (**b**) 2050 and (**c**) 2080.

The study shows a decreasing trend in the operation hours of the heating system for each month of the future climate scenarios. The major differences arose when the three profiles of the natural ventilation were combined with the heating schedule h1 characterized by continuous activation. In general, December was the month where more variations from 2020 to the future scenarios were observed.

The results for the cooling season, in terms of monthly hours of operation, were obtained with a setpoint temperature of 26 ◦C and are shown in Figure 6.

As a consequence of the external temperature rise, it is possible to observe an increasing trend of the monthly hours of operation of the cooling system in the future climate conditions. May, June, and September registered the main increases with the schedules c1 and c2. This growth was more visible with profile c1 because the cooling system could operate for more hours and mainly in the hottest hours. Considering the schedule c3, the operation of the cooling system was from June to September in 2020, and also needed in May during the future climate scenarios. It mainly happened when the cooling schedule c3 was coupled with the natural ventilation profile v1 because the ventilation occurred in the hottest hours of the day, producing an increase of the internal air temperature, and consequently, a prolonged cooling system operation.

**Figure 6.** Monthly hours of operation of the cooling system: (**a**) in the current climate, and differences Δ*h* in the years (**b**) 2050 and (**c**) 2080.

#### *4.3. Impact of Occupant Behavior on Energy Needs*

Figure 7 shows the heating and cooling energy needs in the current climate with a heating setpoint temperature of 20 ◦C and a cooling setpoint temperature equal to 26 ◦C.

**Figure 7.** Energy needs in the current climate for: (**a**) the heating season; (**b**) the cooling season.

In winter, the energy requirement (Figure 7a) was more influenced by the heating schedules than the natural ventilation type. In particular, values of the order of 2000 kWh, 1000 kWh, and 700 kWh were registered for the heating schedule h1, h3, and h2, respectively. These differences in energy needs were due to the diverse duration of the heating system operation.

On the other hand, the cooling energy need seems to be affected by both the operation type and natural ventilation schedules. A decreasing trend of the energy requirement from the cooling schedule c1 to c3 and from the natural ventilation schedule v1 to v3 was observed. In more detail, the cooling energy need ranged from 714.8 kWh to 619.7 kWh, from 616.4 kWh to 534.7 kWh, and from 606.4 kWh to 511.7 kWh for c1, c2, and c3, respectively. These results can be explained by analyzing the cooling and ventilation profiles. In fact, the cooling system could operate for more hours and in the hottest hours of the day with the schedule c1.

Also, the natural ventilation with profile v1 mainly occurred in the hours in which the external air temperature can be higher than the internal one leading to greater cooling energy needs. In contrast, the schedules v2 and v3 produced a positive effect the energy balance.

In the current climate, h2v2 and c3v3 were the less heating and cooling energy-demanding profiles, while h1v1 and c1v1 were those with the most heating and cooling energy requirement.

#### *4.4. Impact of Climate Changes on the Energy Needs*

The use profiles were also used to assess their impact on future climate scenarios characterized by temperature rise. Figure 8 illustrates the relative differences of the energy needs in 2050 and 2080 compared to 2020.

**Figure 8.** Relative differences between the energy needs calculated in the current climate and in future climate scenarios for: (**a**) the heating season; (**b**) the cooling season.

For all future scenarios, energy needs reductions were observed in the heating season, and energy needs increments in the cooling season. In winter, the impact of climate change was more predominant than the impact of occupant behavior (Figure 8b). In fact, significant variations were found from one year to the next and not in different heating and ventilation profiles. The differences varied from −24% to −26%, and from −47% to −52% in 2050 and 2080, respectively.

In summer, visible variations were observable varying the use profiles and passing from a climatic scenario to another (Figure 8b). In fact, energy requirements increased from +48% to +54%, from +46% to +53%, and from +60% to +73% with the cooling schedule c1, c2, and c3 in 2050, respectively. Moreover, for 2080, cooling need increased from +94% to +107%, from +87% to 100%, and from +121% to +146% with the schedule c1, c2, and c3, respectively.

#### *4.5. Impact of the Heating and Cooling Setpoint Temperatures on Energy Needs*

Occupants can impact the energy performance of buildings also by varying the setpoint temperature of the heating and cooling system.

Figures 9–11 present, for the different climate scenarios, the variations of the heating and cooling energy needs when the setpoint temperatures were modified of ±2 ◦C.

As expected, the decrease in the heating setpoint temperature by 2 ◦C led to a reduction in energy requirements, and the increase in temperature consequently produced an increase in energy need (see Figure 9a). Opposite trends in thermal behavior were observed by varying the cooling setpoint temperature (see Figure 9b).

**Figure 9.** Relative differences of the (**a**) heating and (**b**) cooling energy needs caused by a variation of the setpoint temperature of ±2 ◦C in 2020.

**Figure 10.** Relative differences of the (**a**) heating and (**b**) cooling energy needs caused by a variation of the setpoint temperature of ±2 ◦C in 2050.

**Figure 11.** Relative differences of the (**a**) heating and (**b**) cooling energy needs caused by a variation of the setpoint temperature of ±2 ◦C in 2080.

More in detail, regarding the heating season in the current climate, the energy need decreased from −48% to −54% when the internal air temperature was set equal to 18 ◦C and increased from +62% to +77% when 22 ◦C was the selected setpoint. The maximum variation was found for profile h2v2, for which the energy need was 595.2 kWh at 20 ◦C, and 271 kWh and 1052 kWh at 18 ◦C and 22 ◦C, respectively.

In summer, the energy need increased from +65% to +83%, when the setpoint temperature was 28 ◦C and decreased from –48% to –58% when it was equal to 24 ◦C, with a maximum variation for profile c3v3 with both 24 ◦C and 28 ◦C. In particular, the greatest variations were found for the c3v3 profile that registered an energy requirement of 511.7 kWh at 26 ◦C, and of 935.5 kWh and 215 kWh at 24 ◦C and 28 ◦C, respectively.

The same information as above, but referring to 2050, is shown in Figure 10. For both the heating (Figure 10a) and cooling needs (Figure 10b), the general trends were similar to those noticed in 2020, what changes were the magnitude of the variations.

Specifically, in 2050, the heating energy need encountered higher fluctuations by varying the setpoint temperature (decrement from −53% to −61% and increment from +72% to +89%). The maximum variation, also in 2050, was observed in both cases for profile h2v2. Instead, in summer, the variations due to the occupants' preferences had a minor impact: The energy need increased from +46% to +57% and decreased from −40% to −48%. The results for 2080 are shown in Figure 11.

In 2080, the reduction and increase of the heating setpoint temperature led to remarkable changes in energy requirements (from −59% to −65% for 18 ◦C, and from +90% to +114% for 22 ◦C). In the cooling season, the variations of the setpoint temperature determined more limited modifications in terms of energy needs that increased from +36% to +45% and decreased from −35% to −43%. As happened in 2020, also in 2050 and 2080, the maximum variations were observed for profile h2v2 in winter and c3v3 in summer.

#### *4.6. Discussion*

Energy simulations were first performed with setpoint temperature equal to 20 ◦C in winter and 26 ◦C in summer. In 2020, the heating energy needs were more influenced by heating schedules than ventilation profiles, and values of the order of 2000 kWh, 1000 kWh, and 700 kWh were registered for the continuous and the two intermittent operations, respectively. In summer, the cooling energy needs were affected by both cooling and ventilation operations. They ranged from 511.7kWh to 606.4 kWh, from 534.7 kWh to 616.4 kWh, and from 619.7 kWh to 714.8 kWh in the three operation modes.

In future scenarios, the temperature rise determined the decrement of the heating energy needs and the augmentation of the cooling energy needs, in agreement with the results of the previous studies. Specifically, during the heating season, energy needs reductions from −24% to −26% in 2050, and from −47% to −52% in 2080 were obtained. In summer, energy requirements increased from +48% to +54%, from +46% to +53%, and from +60% to +73% by changing the cooling schedule in 2050. Moreover, the increments obtained in 2080 were around double then those registered in 2050.

In addition to natural ventilation habits and systems operation mode, the occupants' can have different preferences in thermal comfort conditions, thus, variations of the setpoint temperature of ±2 ◦C were considered.

In particular, in 2020, the heating energy needs decreased from −48% to −54% and increased from +62% to +77% when the setpoint temperature was set equal to 18 ◦C and 22 ◦C, respectively. On the other hand, cooling energy needs increased from +65% to +83% and decreased from −48% to −58% with setpoint temperature equal to 28 ◦C and 24 ◦C, respectively.

From 2020 to 2080, the variations of energy needs were smaller for the heating and greater for the cooling. In any case, occupants' behavior in controlling and personalizing the indoor thermal conditions had a consistent impact in each climatic scenario.

To the extent of our knowledge, this study was the first that jointly assessed both the impact of occupant behavior and climate change on the energy performance of buildings. The results of this study can be considered indicative of what could be predicted in other Mediterranean countries.

A limitation of this study consists in the fact that energy evaluations were carried out in one location and for a type of building. Thus, the results are contextual and suggest further investigations to address the implication of both occupant behavior and climate change on the heating and cooling energy needs in diverse building typologies and climatic conditions.

Furthermore, this initial study provided informative results for scientists and policymakers as both human factors and environmental conditions can consistently affect the energy consumptions of buildings. Moreover, if the temperature rise determines the reduction of the energy needs in winter and the increment in summer, different preferences and behavior of occupants can lead to better managing of the systems' operation following energy-saving intentions in every season.

Therefore, adequate attention is needed for the aforementioned aspects in future regulations and design criteria.

#### **5. Conclusions**

Dynamic simulations were conducted to assess how the heating and cooling energy needs of a residential unit were affected by occupants' behavior and climate change. In particular, the impact of occupants' behavior was investigated by applying nine usage profiles of heating, cooling, and natural ventilation in winter and summer. Moreover, the influence of occupant behavior was taken into account by varying the indoor setpoint temperature. Regarding climate changes, three scenarios were considered—2020, 2050, and 2080.

The heating energy needs in 2020 were more influenced by heating schedules than ventilation profiles, while the cooling energy needs were consistently affected by both cooling and ventilation operations.

As expected, reducing the energy needs in winter and a rise in summer were noticed in future scenarios. In addition, due to the temperature increase, the variations of energy needs in 2080 were doubled than those obtained in 2050. More relevant results were highlighted concerning the impact of the setpoint temperature. In fact, the variations of energy needs registered from 2020 to 2080 were higher for the cooling than those for the heating, indicating that standards and codes should place more attention to future prescriptions about this control parameter.

In general, this study quantified how occupant preferences related to heating, cooling, and natural ventilation affect the energy performance of buildings. It was also demonstrated that due to climate change, buildings could be subjected to more critical climate conditions, which will lead them to have higher energy needs and to emit more CO2. In future scenarios, the impacts of occupant behavior will be amplified, and especially the preferences related to the cooling system will have a consistent impact in Mediterranean countries.

**Author Contributions:** Conceptualization, G.F. and M.D.S.; methodology, G.F. and M.D.S.; software, G.F. and M.D.S.; validation, M.D.S.; formal analysis, G.F. and M.D.S.; investigation, G.F. and M.D.S.; data curation, G.F. and M.D.S.; writing—original draft preparation, G.F.; writing—review and editing, M.D.S, L.F.C., and L.B.; visualization, G.F. and M.D.S.; project administration, M.D.S.; funding acquisition, M.D.S. and L.F.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Calabria Region Government with the Gianmarco Fajilla's Ph.D. scholarship (POR Calabria FSE/FESR 2014–2020) grant number H21G18000170006. A part of this publication has received funding from Secretaria Nacional de Ciencia y Tecnologia (SENACYT) under the project code FID18-056. This work was partially funded by the Ministerio de Ciencia, Innovación y Universidades de España (RTI2018-093849-B-C31 - MCIU/AEI/FEDER, UE). This work was partially funded by the Ministerio de Ciencia, Innovación y Universidades - Agencia Estatal de Investigación (AEI, RED2018-102431-T). This work is partially supported by ICREA under the ICREA Academia program.

**Acknowledgments:** The authors would like to thank the Catalan Government for the quality accreditation given to their research group (GREiA 2017 SGR 1537). GREiA is certified agent TECNIO in the category of technology developers from the Government of Catalonia.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **A Method for Establishing a Hygrothermally Controlled Test Room for Measuring the Water Vapor Resistivity Characteristics of Construction Materials**

**Toba Samuel Olaoye 1,\* , Mark Dewsbury <sup>1</sup> and Hartwig Kunzel <sup>2</sup>**


**Abstract:** Hygrothermal assessment is essential to the production of healthy and energy efficient buildings. This has given rise to the demand for the development of a hygrothermal laboratory, as input data to hygrothermal modeling tools can only be sourced and validated through appropriate empirical measurements in a laboratory. These data are then used to quantify a building's dynamic characteristic moisture transport vis-a-vis a much more comprehensive energy performance analysis through simulation. This paper discusses the methods used to establish Australia's first hygrothermal laboratory for testing the water vapor resistivity properties of construction materials. The approach included establishing a climatically controlled hygrothermal test room with an automatic integrated system which controls heating, cooling, humidifying, and de-humidifying as required. The data acquisition for this hygrothermal test room operates with the installation of environmental sensors connected to specific and responsive programming codes. The room was successfully controlled to deliver a relative humidity of 50% with ±1%RH deviation and at 23 ◦C temperature with ±<sup>1</sup> ◦<sup>C</sup> fluctuation during the testing of the water vapor diffusion properties of a pliable membrane common in Australian residential construction. To validate the potential of this testing facility, an independent measurement was also conducted at the Fraunhofer Institute of Building Physics laboratory (IBP) Holzkirchen, Germany for the diffusion properties of the same pliable membrane. The interlaboratory testing results were subjected to statistical analysis of variance, this indicates that there is no significant difference between the result obtained in both laboratories. In conclusion, this paper demonstrates that a low-cost hygrothermally controlled test room can successfully replace the more expensive climatic chamber.

**Keywords:** water vapor resistivity; hygrothermal modeling; condensation; mold; hygrothermal properties; energy efficiency; moisture transport; inter-laboratory testing

#### **1. Introduction**

Over the last three decades, the increased expectations for energy efficient buildings combined with greater thermal comfort has established significant differences between the interior and exterior environmental water vapor pressure. This has created the need to manage water vapor diffusion and moisture, and has led to an increased demand for appropriate hygrothermal assessment [1]. Hygrothermal analysis is capable of calculating the dynamic transport of moisture, heat, and air in a building envelope. In most developed nations, this has become an essential part of the production of durable, healthy, comfortable, and energy-efficient buildings [2,3]. The presence of uncontrolled moisture above a critical limits can result in various degrees of deterioration which can include corrosion, rusting, freezing, and swelling of many materials used in the building [2,4]. The most concerning aspect of uncontrolled moisture in a building is the opportunity for mold to grow within

**Citation:** Olaoye, T.S.; Dewsbury, M.; Kunzel, H. A Method for Establishing a Hygrothermally Controlled Test Room for Measuring the Water Vapor Resistivity Characteristics of Construction Materials. *Energies* **2021**, *14*, 4. https://doi.org/10.3390/ en14010004

Received: 10 November 2020 Accepted: 19 December 2020 Published: 22 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

interior spaces. This can have serious implications for the health of the occupants [5,6]. In addition, recent research has shown that high levels of moisture can impact the energy performance of a building and the quality of the indoor air [7–10].

In Australia, moisture problems have become apparent in many new buildings. Up to 50% of National Construction Code Class 1 and Class 2 buildings constructed in the last 15 years have a visible internal formation of condensation [11]. The complexity involved in understanding water vapor transport through appropriate hygrothermal calculation is posing significant challenges to the design and construction professionals in Australia especially when considering moisture management and energy efficiency in buildings [12–14].

While hygrothermal assessment, the key scientific approach to managing condensation and mold in buildings, has been deployed to address these challenges in many other developed nations, it is an emerging field in the Australia [13]. This may be because there were no building regulations requiring insulation in building envelopes until 2003, and the first regulations regarding risk of condensation management only came into effect in 2019. The long-term impact of moisture accumulation on building durability and human health has now become a critical aspect of the Australian regulatory agenda for new buildings.

Across other developed nations, hygrothermal analysis has evolved from manual calculation methods to computer simulations [15–17]. In the last two decades, this has moved from a limited focus on condensation risk analysis to a greater understanding of moisture accumulation, energy efficiency, and the drying capacity envelopes. Over the same period of time, the simulation method has advanced from steady state to transient simulation [18–20].

Several elements need to be considered in choosing an appropriate approach to hygrothermal modeling. In addition to precision and accuracy, the flexibility to allow selection from a variety of climatic zones and the quality of the climatic data are important aspects [21]. Other things to consider include the simulation runtime, the size of the material data library, and how the vapor diffusion and moisture absorption data have been sourced and validated. For instance, WUFI Pro [15], which appears to be the most popularly used hygrothermal software in Europe and North America, has been considered to be reliable because of its ability to deliver a realistic transient calculation and also because all the construction materials in its data library have been well validated [15,22].

The most appropriate method to source and validate construction material's vapor diffusion properties is to conduct measurements in the laboratory. For many nations, the laboratory measurement of water vapor diffusion characteristics of individual construction materials is evolving, and robust databases are being created. The internationally accepted method to represent vapor diffusion is material vapor resistivity. Due to Australia's slower adoption of highly insulated envelopes and vapor resistivity material data has not been required. It is inappropriate to adopt internationally available data directly for use in Australia without appropriate empirical evaluation of their applicability to materials used in Australia's envelope systems and the physical properties of Australian manufactured construction materials. As of 2019, the Australian National Construction Code requires hygrothermal calculations [23,24] in order for the design of new buildings to be approved. Early adopters are using non-Australian data from international material databases for hygrothermal modeling; however, these data may not provide a true representation of Australian construction materials. Without empirical information regarding the vapor diffusion properties of Australian construction materials, there is the potential that inappropriate decisions will be made.

Four types of laboratory-based test methods are internationally recognized for the quantification of the water vapor diffusion properties of materials. These include the electron-analytical, sweating guarded hot plate, dynamic moisture permeation cell test, and the gravimetric methods [4,25–31]. The testing process requires the establishment of two environments with different vapor pressures on each side of the material. Increasingly, the most preferred method for establishing the water vapor diffusion properties

of most construction materials is the gravimetric method [26,32–36]. This involves the measurement of the mass of moisture that has resulted from water vapor diffusion into or out of a test dish assembly, often referred to as the wet-cup or dry-cup test method, respectively [25,32,37]. Depending on whether it is a wet-cup or dry-cup test, salt solutions, distilled water or a desiccant are used to establish a predetermined relative humidity within the test dish. The material is cut and attached to the test dish and then placed in a temperature and humidity-controlled cabinet or room. The humidity outside the cup, in the room, or cabinet, is controlled so that the desired relative humidity condition outside is achieved [37,38]. The conditions created within the cabinet or test room are designed to replicate the hygrothermal conditions the material may expect to experience as a component of the built fabric. The focus of this paper centers on the establishment of an appropriately hygrothermally controlled test room required for gravimetric vapor diffusion testing.

The general principle for the gravimetric method (shown in Figure 1) is to create two environments with different vapor pressures, by establishing different relative humidities inside and outside the cup, while the temperature remains constant. During the test period, the dish is weighed at regular intervals until the mass does not change, indicating the vapor pressure of the test dish and the room have reached equilibrium. For wet cup gravimetric testing (shown in Figure 2, the vapor flux is expected to go from the cup which has a higher RH through the material being tested to the environment which has a lower RH. The reverse is the case for dry cup gravimetric testing, shown in Figure 3. The process is discontinued after a minimum of four consecutive weighing which shows no change in mass.

**Figure 1.** Diagram of water vapor diffusion [13].

**Figure 2.** Diagram of wet cup test method [13].

**Figure 3.** Diagram of dry cup test method [13].

While many research papers have reported different procedures for quantifying the water vapor diffusion of construction materials using the gravimetric method in a climatic cabinet [34,39,40], no research has reported the development of a hygrothermally controlled test room. However, the demand for more hygrothermally controlled test rooms will increase over the coming years both in Australia and internationally. This is because the demand for energy efficient buildings has increased in many jurisdictions as building codes have moved towards the requirement of near-zero energy consumption in buildings. Hence, the need to establish more hygrothermally suitable construction systems will increase and laboratory testing will be required to establish the hygrothermal properties of individual component materials.

The merits of a hygrothermally conditioned test room over the climatic cabinet is the elimination of experimental errors. During the gravimetric weighing, process errors may arise from opening, closing, and transporting test dishes from the cabinet. In a test room, all weighing activities occur within the climatically controlled space. Despite this distinct advantage, little or no research has reported the design, construction, installation of the equipment, and the operations of such a laboratory. This may be because the acquisition and installation of laboratories is not regarded as a research output. In addition, due to commercial reasons, those engineering firms that have built such rooms have never made available the details of the design, construction, and installation of such a facility. This paper describes the methods employed to develop Australia's first hygrothermal laboratory for quantifying the diffusion properties of materials using common appliances, which included a round-robin test conducted between Fraunhofer Institute of Building Physics laboratory (IBP) Holzkirchen Germany, and this hygrothermal testing laboratory at the University of Tasmania (UTAS), Australia.

The approach employed included establishing a climatically controlled hygrothermal test room with an automatic integrated system which allows heating, cooling, humidifying, and de-humidifying as required. The data acquisition for this hygrothermal test room operates with the installation of environmental sensors connected to specific and responsive programming codes. The room reported here, has been used to successfully complete wet and dry cup vapor diffusion material testing for relative humidities RH between 50% with ±1%RH deviation and temperatures between 23 ◦C with ±1 ◦C fluctuation. The test results indicate that a hygrothermally controlled test room can successfully replace the more expensive climatic chamber.

#### **2. Materials and Methods**

To establish a conditioned hygrothermally controlled test room, it was necessary to design and install environmental equipment that controls the interior temperature and relative humidity within the conditioned room. The accurate control of temperature and relative humidity conditions, within the bandwidths prescribed in ISO 12572, is critical to enable gravimetric based testing of building material vapor resistivity properties. For this research, a test building located at the Newnham campus of the University of Tasmania, was reconfigured to enable the conditioned room to be dynamically controlled. The controls included heating, cooling, humidification, and dehumidification. The second stage involved a round-robin testing of the water vapor resistivity properties of a pliable membranes at Fraunhofer Institute of Building Physics laboratory Holzkirchen Germany, and at this hygrothermal testing laboratory. The following sections discuss the design, installation, operation, and the performance of test room, the inter-laboratory testing that was conducted to compare test facilities and results for measuring vapor resistivity properties.

#### *2.1. Design and Description of the Thermal Test Building*

The University of Tasmania has three thermal test buildings at the Newnham campus in Launceston. They include an unenclosed-perimeter platform-floored building, an enclosed-perimeter platform-floored building and a concrete slab-on-ground floored building. Previous research had established that the well-insulated concrete slab-on-ground floored test building demonstrated the most stable interior temperatures without any stratification in both conditioned and unconditioned modes of operation. This building has an internal floor area of 30.03 m2 (5.48 m by 5.48 m), a ceiling height of 2.44 m and total volume of 73.3 m<sup>3</sup> and has no window, as shown in Figures 4 and 5. The building, constructed in

2006, applied Australian best practice wall and ceiling insulation and air-tightness methods. The combination of the ground keyed concrete slab, external walls with R2.5 in-frame wall insulation, R4.2 ceiling insulation, and a well-installed air barrier system ensured a high-quality test building with minimal internal temperature variability.

**Figure 4.** Floor plan of test building.

**Figure 5.** Architectural section of test building.

#### *2.2. Cabling and Installation of Integrated Data Acqusition System*

The control of air temperature and relative humidity are critical to the successful operation of a hygrothermally controlled test room. To enable accurate control of the test room interior a data acquisition system was used. Normally, data acquisition requires one or more transducers (sensors) to sense, process, and send signals from a measuring instrument to the system, the data acquired is then stored or logged into the central processing unit of a computer or external memory for later analysis. The data acquisition system generally includes: the sensors; a device that converts the primary signal from the sensors into a compactible form with the information processing systems; a computer by which the overall system is able to be managed and on which data from sensors are stored. For this research, DataTaker DT500 dataloggers with a channel extension module (CEM) (see Figure 6) were used. Connection between the Datataker and Dell PC was established via a RS232 communication cable (Figure 7). The De Transfer interface software was used for communication between the DT500 data logger and the Dell PC. Two DT 500 DataTaker data-loggers were used, one for temperature sensors and the second for the relative humidity sensors. An array of four wire PT100 sensors were used to measure temperature. An array of two wire Vaisala HMW40U relative humidity sensors were used to measure relative humidity. Due to the number of terminals required for the array of four wire PT100 sensors, they were connected to both the data-logger and the CEM. The second DT500 DataTaker was used to connect the array of relative humidity sensors used for this project. The primary sensor location was on a pole located in the center of the room (see Figure 8). The need for at least three sensors in each location was based on previous research, which queried the reliability of single sensors and when two sensors had varied measured values [41]. The sensors and other apparatus used to control the room are described in Table 1.

**Figure 6.** Data acquisition system (DT 500 datalogger).

**Figure 7.** Desk control.

**Figure 8.** Environmental control equipment.

**Table 1.** Summary of sensors and other equipment.


#### *2.3. Cooling and Heating System*

Automated heating and cooling were essential for the control of this hygrothermally conditioned test room. Figure 9 shows the position of the air-conditioner within the test room. This equipment is a reverse-cycle heat pump and can heat up to 30 ◦C. When heating above 30 ◦C was required for the room, the wall mounted electric heater shown in Figure 10 was turned on. Silicone DC relays (Figure 11) was used as the power switching interface between the data-logger and the appliances.

**Figure 9.** Air-conditioner.

**Figure 10.** Wall-mounted heater.

**Figure 11.** Silicone DC relays.

#### *2.4. Humidity and Pressure Control System*

The capability to control humidity was essential for this hygrothermally controlled room. For this research, this was achieved through the installation of humidity equipment which enabled water vapor to either be added or removed as required. The power switching for the humidity equipment utilized two solid-state relays shown in Figure 12. The first method to add water vapor to the air was to use a fishpond with a water heater. However, after preliminary testing and discussions with other research collaborators, it was established that there would be a significant water vapor lag with this method. This led to an analysis of quick response humidifiers. This resulted in the selection ofa6L Ultrasonic Cool Mist Steam Nebulizer Diffuser Purifier (shown in Figure 13). This humidifier quickly demonstrated a very fast response to add extra water vapor to the room. Similarly, a Breville Smart dry de-humidifier (Figure 14), was installed to remove excessive water vapor from the room. The power supply for the humidifier and dehumidifier was controlled by a solid-state relay, which in turn was controlled by the DT500 data-logger. In practical terms, when the relative humidity in the room was too high the programmed data logger alarm switched the relay, thus providing power to the dehumidifier. When the desired relative humidity value was achieved, the programmed data logger alarm switched the relay off. Conversely, when the relative humidity was too low, the data logger alarm switched the relay to provide power to the humidifier, thus adding water vapor into the room until the required relative humidity setpoint was reached.

**Figure 12.** Solid state relay.

**Figure 13.** 6 litres Ultrasonic Humidifier.

**Figure 14.** Dehumidifier.

Additionally, a household fan was installed to provide circulation of the air in the room to minimize water vapor stratification.

#### *2.5. Calibration of the Environmental Instruments*

Calibration of the temperature and relative humidity sensors was completed to avoid intrinsic error that may have existed in the devices or data logging equipment. In the first instance, all sensors were carefully chosen for their level of accuracy and long-term reliability. A diagnostic procedure was established to ensure that wiring from the data logger to each sensor did not cause errors in measurements. The on-site calibration utilized pre-calibrated NATA certified temperature and relative humidity sensors provided by Industrial Technik. The calibration of the temperature sensors included zero degrees, room temperature and near boiling temperature. This was to ensure that there were no linear or non-linear errors. Any sensor that had erroneous outputs was replaced. The output

from the relative humidity sensors was compared to a certified and pre-calibrated sensor, whilst the relative humidity was increased and decreased

#### *2.6. Monitoring and Controlling Environmental Conditions*

As previously mentioned, the DataTaker DT500 data logger was used for data acquisition. This system relied on programming code for data acquisition from the sensors and to control the switching relays for the heating, cooling, humidifying, and de-humidifying appliances. The acquisition systems collected temperature and relative humidity data from the sensors and simultaneously stored the data in the memory of Datataker for later use. Figure 15 shows a snapshot of an example of the programming code use to operate and collect temperature data from the PT100 sensors. This code was written according to the sensor type. Similarly, the programming code for acquiring the relative humidity data within the hygrothermal room is shown in Figure 16. In this research, temperature and relative humidity data was collected every 10 min. The examples of the programing code also show alarm codes. The coding shows minimum and maximum values for temperature and relative humidity. The alarms required the data logger to continuously monitor the relative humidity and temperature conditions in the test room. The alarm-controlled power supply to the digital switches on the data loggers. In turn, the digital switches controlled the power supply to the silicone and solid-state relays, which controlled the appliances. The combination of continuous measurement and the control of the four appliances, enabled the room temperature and relative humidity to be adequately controlled by the heating, cooling, humidifying, and dehumidifying appliances.

**Figure 15.** Example of temperature programming code.

**Figure 16.** Example of relative humidity programming code.

#### *2.7. Inter-Laboratory Testing of Wet-Cup and Dry-Cup Dishes*

The procedure for the interlaboratory testing involved the selection of a pliable membrane classified as permeable material in clause AS 4200:1 and carrying out a standard test as referred to in ISO 12572. The independent testing of water vapor resistivity properties was completed on a pliable membrane commonly used in Australian external envelope construction systems. The same material was tested under the same climatic condition of <sup>23</sup> ◦C/50%RH at both the hygrothermal laboratory at Fraunhofer IBP Germany, and UTAS, Australia. Table 2 shows the comparison of the important testing parameters that were used.



It was necessary to employ very similar round glass dishes with diameter of 200 mm. While the depth of the dishes at IBP is 80 mm, at UTAS, the dept is 60 mm. For accuracy, three dishes were used for wet-cup and another three were used for dry-cup gravimetric

measurement both in Germany and in Australia. To achieve the desired humidity testing condition within wet-cup dishes, ammonium dihydrogen phosphate solution was placed in the dish, by both laboratories during the testing. This achieved a dish relative humidity of 93% (Figure 17). Similarly, to achieve the desired testing humidity condition within the dry-cup test dishes, silica gel beads were used at both laboratories, as shown in Figure 18. This achieved relative humidity of 3% within the dishes. Both laboratories employed a 20 mm air space between the top surface of the substrates and the bottom surface of the test specimen. The pliable membrane specimens were then glued to the top edge of the dishes. To avoid water vapor leakages between the dishes and test specimens, the edges between the materials were taped and sealed with molten paraffin wax at 100 ◦C. The dishes were then placed on shelving within these test rooms, as shown in Figures 19 and 20.

**Figure 17.** Wet-cup test method.

**Figure 18.** Dry-cup test method.

**Figure 19.** Shelving in the interior of test room at IBP Germany.

**Figure 20.** Shelving in the interior of test room at UTAS, Australia.

Regular weighing measurements of the test dishes were taken every two hours until equilibrium was achieved. The measurements were in milligrams and all weighing data were recorded. The calculations of the water vapor resistivity properties were obtained mathematically (see Tables 3 and 4). Microsoft Excel 365 was used to complete a statistical analysis of variance to establish if there was any significant difference between the result obtained from the laboratory at Fraunhofer IBP and UTAS.

**Table 3.** Water vapor diffusion properties measured at IBP.



#### Specimen Mean thicknessd (m) Area m<sup>2</sup> specimen grammes (g) vapour flux g = G/A in kg/(s\*m2) vapour permeance = g/dp in kg/(s\*m2\*Pa) vapour resistance Z = 1/W in (s\*m2\*Pa)/kg vapour resistance factor μ equivalent air layer thickness Sd (m) TA4 0.000824 0.0275 7.43 3.34 × <sup>10</sup>−<sup>6</sup> 2.76 × <sup>10</sup>−<sup>9</sup> 3.62 × 108 60.99 0.0503 TA5 0.000804 0.0278 7.40 3.55 × <sup>10</sup>−<sup>6</sup> 2.94 × <sup>10</sup>−<sup>9</sup> 3.40 × 108 57.15 0.0459 TA6 0.000805 0.0275 7.17 3.40 × <sup>10</sup>−<sup>6</sup> 2.82 × <sup>10</sup>−<sup>9</sup> 3.55 × 108 60.81 0.0490 Mean 0.000811 0.0276 7.33 3.43 × <sup>10</sup>−<sup>6</sup> 2.83 × <sup>10</sup>−<sup>9</sup> 3.52 × 108 59.65 0.0484 Standard deviation 1.13 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 0.000160728 0.142243922 1.11 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 9.28 <sup>×</sup> <sup>10</sup>−<sup>11</sup> 1.13 <sup>×</sup> <sup>107</sup> 02.17 0.0023

#### **3. Results**

*3.1. Hygrothermal Control of the Test Room*

This section discusses the result from the climatic control of the hygrothermal test room which was used to quantify the water vapor diffusion properties of the permeable pliable membrane, when the test room was maintained at 50% relative humidity and the temperature remained at 23 ◦C (±1 ◦C) for the material testing periods. It was found that the room would take up to 72 h to initially reach and stabilize at the desired temperature and relative humidity.

During the establishment of the test room, sensors which controlled the operation of heating, cooling, humidifying, and dehumidifying appliances were moved until adequate control of the room was established. The final two versions of the sensor locations are shown in Table 1. The principle reason for the change in sensor location between Version 1 and Version 2 was a measured, and significant time lag for room temperature control. The time lag issues were addressed by the Version 2 configuration.

To demonstrate the potential of this hygrothermally controlled room at UTAS, the temperature and relative humidity during the material testing period was retrieved for analysis. Figure 21 shows the temperature profile of test room for the period of six weeks, while Figure 22 shows the relative humidity profile for this same period which required the relative humidity be kept at 50%. The blue box plot (Figure 23) shows the observations from three temperature sensors located 1800 mm above the floor, the orange box plot shows the observations from three temperature sensors located 1200 mm above the floor, the grey box plot shows the observations from three globe temperature (mean radiant)

sensors located 1200 mm above the floor, and the yellow box plot shows the observations from three temperature sensors located 600 mm above the floor. Summarily the box plot observation indicates that aside from occasional outliers, the temperature in the room was maintained between 23.2 ◦C and 22.6 ◦C, with an average of 22.9 ◦C (±1 ◦C). Figure 24 shows the results from the three relative humidity sensors for the corresponding period, and the box plots show that aside from occasional outliers, the relative humidity was maintained between 49.8% and 50.8%, with an average humidity of 50.4% (±1%).

**Figure 21.** Temperature profile of the room aimed at 23 ◦C (+/−0.5 ◦C) for the testing period 2.

**Figure 22.** Relative humidity profile of the room aimed at 50% for the testing period 2.

**Figure 23.** Box and whisker plot of temperature observations during test 2.

**Figure 24.** Box and whisker plot of relative humidity observations during test 2.

#### *3.2. Comparison of the Interlaboratory Results for the Water Vapor Diffusion Properties*

The gravimetric measurement of change in mass over a particular period commenced as soon as the dishes were placed in the test room. Initially, weighing was completed at two hourly intervals. This was to establish if the dish gained or lost weight (depending on the dry-cup or wet-cup substrate). Tables 3 and 4 show the water vapor resistivity properties measured for the permeable pliable membrane commonly used for Australian construction system.

The analysis of variance that was completed shows that there was no significant difference (*p* = 0.38) between the results of the water vapor resistance factor (Table 5) for the wet-cup test obtained in both IBP and UTAS. Similarly, for the dry-cup test, the there was no significant difference (*p* = 0.77) between the results of the test obtained in both laboratories. Table 6 also indicates that there was no significant difference (*p* = 0.34) between the result of the wet-cup test obtained in both IBP and UTAS for the diffusion-equivalent air layer thickness, and there was no significant difference (*p* = 0.89) between the results of the dry-cup test obtained in both laboratories.

*Energies* **2021**, *14*, 4

**Table 5.** Inter-laboratory comparison of the ANOVA result for the resistance factor (μ) of wet-cup test.


*Energies* **2021**, *14*, 4

 **6.**Inter-laboratory comparison of the ANOVA result for the diffusion-equivalent air layer thickness Sd(m) of dry-cup


#### **4. Discussion**

Firstly, the set-up and configuration of the test room followed many practices common for the establishment of environmentally controlled spaces. The points of interest were the challenges in controlling the room temperature and the configuration and operation of the humidifier and de-humidifier. The ability to keep the temperature and relative humidity within specific bandwidths was critical. The temperature was kept within +/−1 ◦C and the relative humidity was kept within +/− 1% RH. Table 1 makes note of Version 1 and Version 2 for the measurement of dry bulb air temperature. The data logger combined with relay switches demonstrated a simple mechanism to control room temperature. However, there was a recognized time lag and regular over-heating of the test room. After several iterations of data logger programming and the co-location of additional sensors around the air-conditioning appliance, localized temperature stratification near the appliance was identified. An additional PT100 temperature sensor was installed close to the airconditioner thermostat to establish the step difference that was occurring. This extra data allowed for a more informed approach to the data-logger alarm bandwidths, which controlled the air-conditioner power supply.

Secondly, the result of the inter-laboratory measurement of the water vapor resistance factor and the diffusion equivalent air layer thickness of a permeable membrane was investigated to validate the performance of the UTAS laboratory. Under the same experimental procedure and parameters, similar results were obtained, while experimental procedural error was minimized. Recent research [42] had indicated that irrespective of the material to be tested or the test procedure, discrepancies in results may normally occur during any inter-laboratory measurement to determine the water vapor diffusion properties of material through gravimetric cup test. The ANOVA test for this research has demonstrated that discrepancies in the result of interlaboratory measurement of pliable membrane is insignificant. This implies that the hygrothermally controlled room at UTAS can be used for the same experimental purposes obtained at IBP.

The results of the water vapor diffusion properties from the interlaboratory testing with the world leading IBP laboratory indicates that the operation of this laboratory is promising, as this method can be employed to set up a low-cost hygrothermal testing facility.

#### **5. Conclusions and Recommendations**

Essentially, the equipment in the test cells, comprised of an all-embracing range of temperature and relative humidity sensors, and an integrated data acquisition system, which enable flexible monitoring and control of heating, cooling, humidifying, and dehumidifying appliance. This combination of equipment enabled the stabilization of temperature and relative humidity which are key parameters for construction material wet-cup and dry-cup water vapor diffusion testing. The integrated system enabled the stabilization of the temperature and the relative humidity through the use of simple data-logger programming code. The current configuration, operation, and performance of the test room temperature and humidity indicated that the precise profiles required for the vapor diffusion measurement were achieved and maintained for test room conditions of 23 ◦C with a 50% RH.

This paper reports the establishment of Australia's first precisely controlled hygrothermal room for measuring the water vapor diffusion properties of building materials via the use of a conditioned test room. As a key component of this research is to provide national guidance and methods for the establishment of vapor diffusion properties of Australian Construction materials, this is a positive outcome. The use of an environmentally controlled test room for measuring water vapor diffusion properties of building materials is considered more appropriate than other published methods. This is because the process of taking test dishes in and out of conditioned cabinets for weighing allows for the possibility of intrinsic errors. In summary, this research has demonstrated that the establishment of a conditioned hygrothermal test room may not be financially onerous for prospective researchers seeking to establish a hygrothermally controlled laboratory, that can be used to quantify water vapor diffusion properties for locally made construction materials.

**Author Contributions:** T.S.O.—Main author carried out the experiments both at UTAS and IBP, involved in the conceptualization; collects data; analyze data; graphs and visualization; provide the original draft manuscript, and revised manuscript. M.D.—Second author provides guidance to experiment, source for funding to procure equipment; project administration; contribute to data analyze; data curation; contribute to graphs and visualization, edit and provided revision to the manuscript. H.K.—Provides guidance for the laboratory operation, supervision and validation at IBP, and revision of manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** These authors acknowledge the Commonwealth Scientific Industrial Research Organization (CSIRO) for co-funding this research project.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Non-Intrusive Measurements to Incorporate the Air Renovations in Dynamic Models Assessing the In-Situ Thermal Performance of Buildings**

**María José Jiménez 1,\* , José Alberto Díaz 1, Antonio Javier Alonso 2, Sergio Castaño <sup>1</sup> and Manuel Pérez <sup>2</sup>**


**Abstract:** This paper reports the analysis of the feasibility to characterise the air leakage and the mechanical ventilation avoiding the intrusiveness of the traditional measurement techniques of the corresponding indicators in buildings. The viability of obtaining the air renovation rate itself from measurements of the concentration of the metabolic CO2, and the possibilities to express this rate as function of other climatic variables, are studied. N2O tracer gas measurements have been taken as reference. A Test Cell and two full size buildings, with and without mechanical ventilation and with different levels of air leakage, are considered as case studies. One-month test campaigns have been used for the reference N2O tracer gas experiments. Longer periods are available for the analysis based on CO2 concentration. When the mechanical ventilation is not active, the results indicate significant correlation between the air renovation rate and the wind speed. The agreement between the N2O reference values and the evolution of the metabolic CO2 is larger for larger initial values of the CO2 concentration. When the mechanical ventilation is active, relevant variations have been observed among the N2O reference values along the test campaigns, without evidencing any correlation with the considered boundary variables.

**Keywords:** building energy; building envelope; performance assessment; air renovation; nonintrusive measurements; on-board monitoring

#### **1. Introduction**

Buildings use about 40% of the total energy produced globally and have a relevant potential in terms of energy savings and reducing the pollutant emissions to the atmosphere [1]. These issues are driving an increasing interest to foster the energy efficiency in buildings leading to the elaboration and incorporation of related regulations, stressing the demand to broaden the knowledge related to the energy performance of the buildings, and motivating many research initiatives in this area. Presently, the majority of the checks of compliance and energy performance labelling of buildings rely on design values and theoretical assessments or simulations. Nevertheless, many researches have demonstrated that the actual performance of a building can be very different from the one theoretically evaluated [2,3]. The readiness of reliable enough test procedures applicable to as built buildings for assessing their thermal performance, would contribute to eliminate the problems related to the performance gap. The need for tools identifying the sources of the performance gaps, and providing feedback to different stakeholders, is included among the research themes considered by the Energy in Buildings and Communities (EBC) Technology Collaboration Programme (TCP) of the International Energy Agency (IEA) [1]. One of the elements having a significant influence on the energy behaviour of the buildings is the building envelope. The identification of the intrinsic thermal properties characterising

**Citation:** Jiménez, M.J.; Díaz, J.A.; Alonso, A.J.; Castaño, S.; Pérez, M. Non-Intrusive Measurements to Incorporate the Air Renovations in Dynamic Models Assessing the in-Situ Thermal Performance of Buildings. *Energies* **2021**, *14*, 37. https://dx.doi.org/ 10.3390/en14010037

Received: 13 November 2020 Accepted: 21 December 2020 Published: 23 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/).

the as built building envelope from on board monitoring system is recently attracting the attention of many research groups in the context of international collaboration initiatives [4]. In this context, those monitoring systems with a limited set of non-intrusive measurement devices, embedded in the building, as those typically used for billing or for controlling the Heating, Ventilating and Air Conditioning (HVAC) systems are considered as on board monitoring systems. The energy performance assessment of the building envelope can be carried out through data analysis techniques that require the measurement (that can be direct or indirect) of all the effects that contribute to the energy balance in the space that is confined by the building envelope being characterised [5]. One of the contributions to this energy balance is the one from air renovations, either by mechanical or natural ventilation, or by infiltrations as consequence of cracks or material porosity [6].

There are several procedures for the experimental assessment of the air renovation rate in rooms. Some of these procedures are based on pressurisation and others are based on tracer gas techniques [7]. These traditionally applied methods that could give precise results are complex, expensive and highly intrusive for the building users and inhabitants. Additionally, these traditional techniques characterise the air renovations by a constant parameter. Some standardised procedures obtain this parameter under a pressure that is raised regarding the pressure of the building in use [8]. These constant values can introduce some degree of uncertainty on the data based dynamic modelling techniques that are applied for the thermal performance assessment of the building envelope from on-board monitoring systems [5,9,10]. Part of this uncertainty can be driven by the use of the air renovation rate as a constant parameter when actually it is a variable. A review paper that has been recently published identifies the dynamic behaviour of the air renovation rate as an issue contributing to the uncertainty in tracer gas-based methods [11]. Other authors have analysed the uncertainties due to wind in building pressurisation tests [12]. They identified errors in the rage 6–12% for wind speed in the range 6–10 ms−<sup>1</sup> for test carried out under a standard pressure of 50 Pa, while the errors raised up to 35% and 60% for wind speeds of 6 ms−<sup>1</sup> and 10 ms<sup>−</sup>1, respectively under a pressure of 10 Pa. When the air renovation rate is obtained according the standardised building pressurisation tests, the transformation of the pressurised value to the non-pressurised one, can introduce also certain degree of uncertainty in the dynamic models that are used for the energy performance assessment of in-use buildings. The presence of some uncertainty and variability in the air renovation rate due to infiltrations as well as mechanical ventilation, can contribute to understand and explain the behaviour of the Heat Loss Coefficient (HLC) experimentally assessed and its uncertainties [13,14].

The work reported in this paper is focused on the experimental assessment of the air renovation rate analysing the reliability of cheaper and more cost effective techniques regarding the traditional techniques based on tracer gas. The feasibility to characterise air leakage and mechanical ventilation avoiding the intrusiveness of the traditional measurement techniques is analysed. The viability to obtain the air renovation rate itself, as well as the possibilities to express it as function of other variables (such as wind speed, atmospheric pressure, etc.), are studied extending some preliminary studies [15]. Tracer gas measurements based on N2O have been used as reference. Experimental relations between the air renovations and the wind speed, the indoor-outdoor air temperature difference, and the atmospheric pressure have been analysed. The reliability of an alternative method based on the evolution of the metabolic CO2 using wall mounted sensors of CO2 concentration is evaluated. A PASLINK Test Cell [16,17] and two full size buildings are considered as case studies. First the Test Cell and a very simple single zone building, without mechanical ventilation, are considered. Afterwards, a room in an office building has been studied with and without mechanical ventilation. One-month test campaigns have been used for the reference study based on tracer gas measurements using N2O, in both buildings and the Test Cell. Longer periods are available for the analysis based on CO2 concentration.

The next sections are organised as follows: Section 2 presents the considered case studies, and briefly describes the experiment set up and the methodology applied for data analysis, Section 3 presents and discusses the results that have been obtained for the different case studies, and finally Section 4 summarises the conclusions regarding the behaviour of the air renovation rate, discusses the effect of this behaviour on the Heat Loss Coefficient (HLC) and suggest further research on this issue.

#### **2. Materials and Methods**

The next subsections included under this section describe the three considered case studies, the experiment set up, the tests carried out, and finally the methodology applied for data analysis.

#### *2.1. Case Studies*

A PASLINK Test Cell and two full size extensively monitored buildings are considered as case studies [16,17]. These buildings and the Test Cell, briefly described in Section 2.1.1, Section 2.1.2, Section 2.1.3 are at the CIEMAT's Plataforma Solar de Almeria (PSA), in Tabernas (37.1◦ N, 2.4◦ W), Almería (Spain). They are in a rural area where the climate is semi-arid, with large day-night temperature variations.

#### 2.1.1. PASLINK Test Cell

The PASLINK Test Cell consists in a test facility with a high-thermal-insulation test room and an auxiliary room (Figure 1a). The test room has a surface of 4.825 × 2.48 m<sup>2</sup> and its high is 2.47 m. The Test Cell is placed in a large open area without any shading. It has an air conditioning system and measurement devices for testing full-scale building components. Its test room envelope is highly insulated by 40 cm of polystyrene and it is equipped with the Pseudo-Adiabatic Shell (PAS) Concept. This system is based on a thermopile that detects if there is heat flux through the envelope of the test room, and cancels it by means of a heating foil. The interior surface of the test room is finished with an aluminium sheet giving it thermal uniformity. The Test Cell is over a rotating device that enables it for testing in any orientation.

**Figure 1.** Buildings considered as case studies: (**a**) PASLINK Test Cell; (**b**) Single-zone building; (**c**) Office building.

The south wall and the roof of the test chamber are interchangeable, which permits any vertical or horizontal building component to be installed for testing. The tests of air renovations considered in this work correspond to a reference experiment. In this case, the Test Cell incorporates a homogeneous and opaque wall in its replaceable façade.

This test was conducted in the framework of a series of tests that included several photovoltaic modules and electrocromic windows replacing a piece of the component taken as reference. The Heat Loss Coefficients of these components are obtained by subtracting the Heat Loss Coefficient obtained with the photovoltaic modules or the electrocromic windows, from the Heat Loss Coefficient obtained from the reference component. The Test Cell is designed to be very airtight. Typical air renovation rates during testing are between 0.02 and 0.05 renovations per hour [18]. The assessment of its air renovation rate is important in order to check the achieved level of air tightness and to assess the deviations from this level due to the climatic variables.

#### 2.1.2. Single-Zone Building

This building is a small workshop with just one room, and its area is 31.83 m<sup>2</sup> (Figure 1b) [17]. It can give experimental support to diverse research activities maintaining it empty or with low occupancy rates. It was built in 2002. It is near another twin building that is placed 2 m from its east wall. Both are built in an open area without any other obstacles around that could shade them.

This building was designed to reduce the energy demand incorporating the following passive strategies: South orientation, shading elements avoiding the solar gains in summer and maximising them in winter, the windows are double-glazed to reduce heat losses, and diagonally aligned (north-south) to facilitate the natural ventilation, thermal mass incorporated in the building envelope, external insulation and high ceilings.

#### 2.1.3. Office Building Prototype

The so called C-DdI ARFRISOL at PSA is a one floor building with most of the regularly occupied offices facing south (Figure 1c). Its net floor area is 1007.40 m2. It was constructed in 2007 in the framework of the PSE-ARFRISOL project [19]. It is a prototype of a new plant, built on one floor longitudinal plan.

A double-wing structure, that is installed on the roof along the main axis of the building, protects it from the solar radiation. This structure integrates two different types of solar collectors. Uncovered collectors which are designed to operate as radiant coolers by night are over the wing facing north. Flat plate collectors that are designed to supply hot water for the heating, cooling and DHW systems are over the south facing wing. Small solar chimneys that provide night ventilation of the offices are constructed on the central part of this structure. The south windows are protected by an overhang that provides shade during the summertime and facilitates passive heating in winter.

This building is in use, but it must be taken into account that the experiments used for this work were carried out when the considered room was positively empty; at lunch time and also once the working day is finalised (identified every day as test 1, and test 2, respectively).

#### *2.2. Experiment Set Up*

A tracer gas device combined with a gas analyser have been used to carry out Decay experiments based on the evolution of N2O concentration in both buildings and the Test Cell.

The Test Cell and the two buildings are extensively monitored. The monitoring system records minutely read measurements of the following variables:


One-month test campaigns for each building were considered for the analysis. These campaigns were conducted under different conditions: Dynamic heating sequence in the Test Cell maintaining a large indoor to outdoor air temperature difference, free running test in the single-zone building, and space heating maintaining the indoor air temperature in a comfort range in the office building.

#### *2.3. Methodology*

2.3.1. Analysis of the Relations between the Air Renovation Rate and Climate Variables


2.3.2. Analysis of Feasibility to Obtain Air Renovation Rate from Wall Mounted CO2 Sensors

Additionally, the reliability of an alternative method based on the evolution of the metabolic CO2 using wall mounted sensors of CO2 concentration is evaluated in a room of the office building. A reference value (CO2infinite) has been used, such that the variable used for the Decay method is the CO2-CO2infinite. This value was obtained as the average of the CO2 concentration in a period when the room is positively non-occupied (from 9 pm to 7 am), starting when the Decay curve has reached its asymptotic value. An error obtained as the percentage of deviation regarding the reference value (based on N2O), has been represented as function of the maximum value of the CO2 concentration at the beginning of the decay method curve.

#### **3. Results and Discussion**

A reference value has been obtained for each of the considered case studies. These reference values have been obtained using a N2O tracer gas applying the Decay method. The measurements carried out for the different case studies, presented in Figure 2, evidence that the air renovation rates are different for the different case studies.

The air renovation rates obtained from these tests are:


The dependence of these infiltration rates on the considered climate variables, and the feasibility to obtain them from the concentration of the metabolic CO2, is discussed in the next subsections.

**Figure 2.** Decay method based on N2O as tracer gas, applied to the three case studies: (**a**) PASLINK Test Cell (08/10/2018– 11/10/2018); (**b**) Single-zone building 24/02/2016; (**c**) Office building without mechanical ventilation (10/02/2017); (**d**) Office building with mechanical ventilation (02/02/2017).

#### *3.1. PASLINK Test Cell. Infiltrations*

As expected, very low infiltration rates have been obtained for all the tests carried out in the PASLINK Test Cell. These results are shown in Figure 3 and Table 1. In this case, the infiltration rate does not show any relevant correlation with the indoor to outdoor air temperature difference (Figure 3a). This correlation also is not relevant with the atmospheric pressure (Figure 3d). However, the air infiltration rate presents some correlation with other considered variables. It shows significant linear dependency on the wind speed (Figure 3e), and the dependency is remarkable on the absolute value of the variation of the wind speed per unit of time (Figure 3f).

#### *3.2. Single-Zone Building. Infiltrations*

The results obtained for the single zone building are summarised in Table 2. This table shows that the air renovation rate (*n*) presents a large variation in the range 0.16– 0.97 renov/hour. Its average is 0.37 renov/hour, and its standard deviation is 0.26 renov/hour. Figure 4a,c,e,g,i) shows that the *n* value has evident correlation with all the considered boundary variables except the atmospheric pressure (Figure 4g). The most relevant correlation detected is regarding the wind speed (Figure 4c). The absolute value of the variation of the wind speed per unit of time is also relevant (Figure 4i).

**Figure 3.** PASLINK Test Cell. Relations between the air renovation rate and the climatic variables. N2O tracer gas measurement taken as reference. (**a**) Indoor to outdoor air temperature difference; (**b**) product of the indoor to outdoor air temperature difference and the wind speed; (**c**) product of the indoor to outdoor air temperature difference and the wind speed raised to two; (**d**) atmospheric pressure; (**e**) wind speed; (**f**) absolute value of the variation of wind sped per unit of time.

**Table 1.** PASLINK Test Cell. Experimentally determined air infiltration rates and climate variables.


<sup>1</sup> The (N2O) indicates that the values included in the column were obtained using the N2O tracer gas.

#### *3.3. Office Building Prototype*

#### 3.3.1. Infiltrations

The results obtained for the studied room are summarised in Figure 4b,d,f,h,j and Table 3. Considering the analysis based on N2O, the air renovation rate (*n*) presents some variation. However, the observed variation is not so large as in the single-zone building. The *n* value is between 0.61 and 0.75 renov/hour. Its average is 0.67 renov/hour, and it standard deviation is 0.05 renov/hour. Figure 4b,d,f,h,j shows that the *n* value has relevant correlation with all the considered boundary variables except the indoor to outdoor air temperature difference and the atmospheric pressure. The most relevant correlation detected is regarding the wind speed (Figure 4d).

It is noticeable the different behaviour observed for the dependence of the *n* value with the indoor-outdoor temperature difference in this heated room regarding the single zone free running building. The *n* value for the heated office does not show relevant dependence with this variable (Figure 4b). This behaviour is also observed in the Test Cell, also heated during the test campaign, that does not show relevant dependence with this variable (Figure 3a). However, a linear tendency is seen for the free running single-zone building (Figure 4a). This different behaviour could be explained by the different ranges of indoor-outdoor air temperature differences in the case studies (Figures 3a and 4a,b).

Acceptable agreement is observed for the values obtained using the metabolic CO2 ref concentration, measured with the wall-mounted sensors, regarding the reference *n* values based on N2O (Table 3 and Figure 5a). The agreement is very poor when the less accurate CO2 sensor is used (Table 3 and Figure 5a). This behaviour is explained by taking into account that the office has just one user, and consequently, the level of CO2 concentration produced by the metabolic activity is very low, which is leading to relevant uncertainties in the estimations of the *n* values if the used sensor does not have enough resolution. These uncertainties show a decreasing tendency when the CO2 concentration increases (Figure 5b). Taking into account this behaviour a better performance of this sensor is foreseen for larger CO2 concentrations that would be present in rooms with more occupants. This issue will be further investigated.


**Table 2.** Single-zone building. Experimentally determined air infiltration rates and climate variables.

<sup>1</sup> The (N2O) indicates that the values included in the column were obtained using the N2O tracer gas.

**Figure 4.** Relations between the air renovations and the climatic variables. Left: single-zone building. Right: Room of the office building. (**a**,**b**) Indoor to outdoor air temperature difference; (**c**,**d**) wind speed; (**e**,**f**) product of the indoor to outdoor air temperature difference and the wind speed; (**g**,**h**) product of the indoor to outdoor air temperature difference and the wind speed raised to two; (**i**,**j**) absolute value of the variation of the wind sped per unit of time.

**Figure 5.** Office number 1, analysis of infiltrations. Percentage of error of the results obtained from the Decay method using the metabolic CO2 concentration and considering as reference the value obtained from the N2O tracer gas. (**a**) Using the CO2 reference sensor; (**b**) using the cheaper CO2 sensor.

#### 3.3.2. Mechanical Ventilation

The results obtained for the studied room are summarised in Tables 4 and 5. Considering the analysis based on N2O, the air renovation rate (*n*) presents a large variation. It is between 0.95 and 3.08 renov/hour. Its average is 1.98 renov/hour which is very close to the design value (2 renov/hour), and it standard deviation is 0.59 renov/hour. However, the *n* value does not show relevant correlation with any of the considered boundary variables. The observed large spread could be caused by the instability of the electricity that powers the mechanical ventilation system that transmits such instability to the ventilation rate. Other effects, such as hysteresis of the mechanical components of the ventilation system could contribute to produce the detected variations. The causes of the detected large spread will be further investigated in future research works.

Large uncertainties are observed for the values obtained using the metabolic CO2 concentration measured with the wall-mounted sensors (Tables 4 and 5 and Figure 6). These uncertainties are remarkably larger than those observed for the same room without mechanical ventilation (Figure 5). This high uncertainty is attributed the low level of metabolic CO2 concentration produced by just one user. This issue also leads to large uncertainties in the air renovation rate obtained for the same room without mechanical ventilation using the less accurate sensor (Figure 5b). However, such uncertainty is worsened in the case of mechanical ventilation taking into account that the time interval available for each calculation of the *n* value is shortened regarding the case of not using mechanical ventilation.

**Figure 6.** Office number 1, tests with mechanical ventilation active. Percentage of error of the results obtained from the Decay method using the metabolic CO2 concentration and considering as reference the value obtained from the N2O tracer gas. (**a**) Using the CO2 reference sensor; (**b**) using the cheaper CO2 sensor.



(N2O),(CO2)and(CO2\_ref)indicatethatthevaluesincludedinthecolumnrefer to the measurementsusingtheN2Otracergas,theCO2ortheCO2\_refdevicesrespectively.

*Energies* **2021**, *14*, 37

**Table 4.** Office number 1, test 1 for each day. Experimentally determined air infiltration rates when the mechanical ventilation is active, climate variables and deviations between the results obtained using the metabolic CO2 concentration and the N2O tracer gas.


*Energies* **2021**, *14*, 37

**Table 5.** Office number 1, test 2 for each day. Experimentally determined air infiltration rates when the mechanical ventilation is active, climate variables and deviations between the results obtained the using the metabolic CO2 concentration and the N2O tracer gas.


#### **4. Conclusions**

This section summarises the conclusions regarding the behaviour of the air renovation rate and discusses the effect of this behaviour on the experimental assessment of the Heat Transfer Coefficient (HLC).

The following conclusions are extracted regarding the air renovation rate from the different tests carried out:


The behaviour observed in the air renovation rate, showing large variability considering infiltrations and also considering mechanical ventilation, contributes to understand the behaviour of the HLC experimentally assessed and its uncertainties. The following text summarises the conclusions extracted from this work and some ideas for further research, regarding the influence of the air renovation rate on the behaviour of the Heat Loss Coefficient (HLC):


**Author Contributions:** Measurements, S.C.; Data curation, A.J.A. and J.A.D.; Data analysis, elaboration of graphs and synthesis of results, A.J.A., J.A.D. and M.P.; writing—review and editing M.J.J. and M.P., Methodology and writing—original draft preparation, M.J.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Spanish National Research Agency (Agencia Estatal de Investigación) through the In-Situ-BEPAMAS project, reference PID2019-105046RB-I00. Additionally, the operation of the test facilities that supported this study was partially funded by the Spanish Ministry of Economy, Industry and Competitiveness through ERDF funds (SolarNOVA-II project Ref. ICTS-2017-03-CIEMAT-04).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **On the Retrofit of Existing Buildings with Aerogel Panels: Energy, Environmental and Economic Issues**

**Paola Marrone 1, Francesco Asdrubali 2,\*, Daniela Venanzi <sup>3</sup> , Federico Orsini 1, Luca Evangelisti 2, Claudia Guattari 2, Roberto De Lieto Vollaro <sup>2</sup> , Lucia Fontana 1, Gianluca Grazieschi <sup>2</sup> , Paolo Matteucci <sup>3</sup> and Marta Roncone <sup>2</sup>**


**Abstract:** Among the super insulating materials, aerogel has interesting properties: very low thermal conductivity and density, resistance to high temperatures and transparency. It is a rather expensive material, but incentives in the field can improve its economic attractiveness. Starting from this, the thermal behavior of a test building entirely insulated with aerogel panels was investigated through an extended experimental campaign. A dynamic simulation model of a case study building was generated to better comprehend the energy savings obtained through aerogel in terms of energy demand over a whole year. The investigation was completed by computing the carbon and energy payback times of various retrofit strategies through a life cycle assessment approach, as well as by a cost-benefit analysis through a probabilistic financial framework. Compared to conventional insulation materials, aerogel is characterized by a higher energy and carbon payback time, but it guarantees better environmental performance in the whole life cycle. From an economic-financial perspective, the aerogel retrofit is the best in the current tax incentive scenario. However, due to its higher lump-sum investment, aerogel's net present value is very sensitive to tax deductions, and it is riskier than the best comparable materials in less favorable tax scenarios.

**Keywords:** aerogel; thermal behavior; dynamic simulation; retrofitting; LCA; economic analysis

#### **1. Introduction**

Climate-changing gases (GHG), mainly produced by anthropogenic activities, are now considered to be the main responsible factor for the global warming; in fact, the global average temperature has increased by about 1 ◦C compared to the pre-industrial era [1]. Consequently, due to global warming and climate change (CC), large and densely populated areas risk becoming inhospitable [2]. To avoid, or at least reduce, the negative effects of climate change, it is necessary to globally modify the development model aiming at reducing GHG emissions [3,4].

This objective can be pursued by promoting renewable resources [5], inspiring economic development to the principles of the circular economy [6–9], producing low-carbon materials [10,11] and reducing energy consumption [12]. In this context, cities and buildings play a fundamental role. They are among the main responsible factors for energy consumption and GHG emissions (over 30% of the total amount), mainly caused by urban and extra-urban transport, buildings' energy needs (electrical appliances, heating and cooling) and the production of construction materials [13].

**Citation:** Marrone, P.; Asdrubali, F.; Venanzi, D.; Orsini, F.; Evangelisti, L.; Guattari, C.; De Lieto Vollaro, R.; Fontana, L.; Grazieschi, G.; Matteucci, P.; et al. On the Retrofit of Existing Buildings with Aerogel Panels: Energy, Environmental and Economic Issues. *Energies* **2021**, *14*, 1276. https://doi.org/10.3390/en14051276

Academic Editor: Paulo Santos

Received: 19 January 2021 Accepted: 22 February 2021 Published: 25 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The European Green Deal set the target of reducing GHG emissions by at least 50–55% below the levels of 1990 by 2030 [14].

Among the possible actions encouraged by this agreement, strategies to reduce buildings' energy consumption can be listed. In particular, policies targeted at improving the thermal performance of buildings' envelopes [15,16] in terms of reducing heat loss and increasing thermal lag can be implemented.

Alongside the traditional insulator materials [17] made of inorganic constituents (for example, rock wool, expanded polystyrene, etc.), or organic ones (cork, wood fibers, etc.), today non-traditional materials, defined as super-insulating, made with innovative production processes and/or materials, are available [18].

The super-insulating materials are characterized by high performance with a thermal conductivity value lower than 0.020 W/mK, compared to traditional materials (rock wool or glass wool) whose thermal conductivity is equal to 0.035–0.040 W/mK. Comparison can also be made with transition materials, such as expanded polyurethane or propylene, characterized by thermal conductivity values ranging between 0.02 and 0.03 W/mK [19,20]. Furthermore, the high thermal performance of non-traditional insulation materials is characterized by a significant reduction in their thicknesses compared to the traditional ones. Different kinds of innovative and high-insulating materials have already been studied by researchers: reflective multilayer insulation [21–23], vacuum insulation panels (VIPs) [24] and gas-filled panels [25].

Among these innovative materials, aerogel appears to be of great interest, ranking among the most interesting innovative products for the near future [26]. Discovered in the early 1930s [27,28], aerogel is a porous synthetic product, in which the gel's liquid component is replaced with a gas. This solution allowed the creation of a highly performing material in terms of thermal insulation, with a thermal conductivity of about 0.013 W/mK. In fact, several studies have highlighted its excellent thermal performance for opaque envelope applications, integrated in panels [29,30] or mortars [31], and for translucent applications, integrated in panels and frames [32,33]. Cuce et al. [34] presented a comprehensive review on aerogel utilization in buildings: the applications range from energy insulation purposes, to sound insulation, fire retardation and air purification. The use of aerogel in retrofits of historical buildings is very competitive since it permits saving inner space, maintaining the external façades unaltered [35]. Karim et al. [36] proposed a superinsulated plaster made with aerogel particles mixed in the matrix. Finally, their optical properties permit the integration of aerogels in different types of glazing systems [37].

If, on one hand, the high performance of aerogel is nowadays well-known owing to several studies, on the other hand, these studies have also focused on the high cost of this material [35]. Nowadays, this aspect is considered as a very strong limit to its widespread application to the construction sector [38].

However, in Italy this limit can now be partially overcome thanks to the introduction in Italian law of a tax credit of 90% (so-called "bonus façades") for the costs incurred for the retrofit of the building façades (see Budget Law 2020). The standard also includes energy retrofit interventions that meet the so-called minimum requirements and the thermal transmittance limit values of the building envelope [39]. Another incentive that is today guaranteed by the Italian legislation is the so called Superbonus 110% (a tax reduction of the 110% of the expenditure sustained for the works aiming at deep energy retrofits of existing buildings [40]). The insulation of building envelopes is one of the driving interventions that are promoted by Superbonus 110%.

Starting from this, it seems important to evaluate the possibility of employing aerogel for the energy retrofit of existing buildings in order to define its effectiveness in terms of both thermal performance and economic feasibility.

This paper has the following structure: Section 2 provides the aim and scope of the research; Section 3 provides some information about the test rooms and the case study, the experimental campaign in the test rooms, data post-processing, simulations in a case

study building and the cost–benefit analysis; Section 4 presents the results; finally, Section 5 draws conclusions.

#### **2. Aim and Scope**

The thermal performance of aerogel is well-known in the literature. It is a super insulating material able to improve the thermal performance of a wall with reduced thicknesses. On the other hand, aerogel is a rather expensive material, and its use needs a comparison between energy savings and installation costs in order to identify costs and benefits.

Thus, the aim of this study is a wide-ranging analysis, examining and comparing aerogel performance to that of other diffused insulation materials employed as an external insulation layer in regions characterized by a mild climate (central Italy). From an economicfinancial point of view, the analysis here conducted applies a complete financial approach, based on a probabilistic method used to measure both the most probable value of the Net Present Value (NPV) of each retrofit alternative and its probability distribution (however, limited to the monetizable costs and benefits). This approach assumes optimistic and pessimistic estimates (defined in a subjective manner) of the uncertain variables and measures the corresponding range of NPV. Therefore, it derives (under some hypotheses) the variance of the NPV that allows obtaining its probability distribution.

The whole analysis was carried out in order to compare and quantify the advantages/disadvantages of employing aerogel instead of other insulation materials, also in the light of the Italian tax credit.

#### **3. Materials and Methods**

This work integrates four evaluation fields to assess the competitiveness of aerogel in comparison with other insulation materials. After an experimental campaign aiming at studying the real thermal performance of an aerogel coating insulation (described in Section 3.1), a simulation was carried out to evaluate the energy savings achievable by building retrofits using aerogel or other insulation materials (introduced in Section 3.2); the evaluation of the related environmental benefits in the life cycle (see Section 3.3) and the estimation of the achievable economic benefits (see Section 3.4) were finally performed.

#### *3.1. The Experimental Campaign*

The experimental measurement campaigns took place in the external area of the CEFME CTP school for construction, located in Pomezia, a small city close to Rome. According to Italian legislation, the climatic zone of Pomezia is D (on a scale from A to F, with A corresponding to the warmest places and F the coldest), with a degree day value equal to 1536. The experimental investigations involved two test rooms characterized by the same geometry, walls and roof stratigraphy, and the same orientation. One of them is not thermally insulated; the other is insulated with aerogel panels. It is worth noting that the two test rooms, despite their close proximity, do not cast shadows on one another. Figure 1a provides an aerial view of the construction site, and Figure 1b shows the geometrical characteristics of the investigated test rooms. Original vertical walls are characterized by brick construction technique, with plastered tuff blocks, reinforced concrete slabs and internal and external cladding with cement plaster. Table 1 lists the stratigraphy of the test rooms' components.

**Figure 1.** Aerial view of the construction site (**a**), geometrical characteristics of the monitored test rooms (**b**).

**Table 1.** Test Rooms' Components Stratigraphy.


Sample images of the analyzed test rooms are reported in Figure 2, where it is possible to observe the external insulation system during installation (aerogel panels characterized by a thickness equal to 0.01 m) and after installation. The external insulation layer was realized with semi-rigid panels [41] (dimensions equal to 1400 × 720 mm), realized by means of a layer of silica aerogel reinforced with PET (Polyethylene terephthalate) fibers (felt), water-repellent and breathable, with mass density equal to 230 kg/m3, thermal conductivity equal to 0.015 W/mK and specific heat capacity equal to 1000 J/kgK. The external finish of the coat was realized with cement fiber panels, which are also mounted with dowels.

**Figure 2.** Selected test rooms in their original state (**a**) and during aerogel panel installation (**b**,**c**).

As already mentioned, one of the test rooms was monitored as a reference structure for measurements, without any thermal insulation. The other was fully insulated with 0.01-m-thick aerogel panels.

In order to assess the thermal behavior of the reference test room and the insulated one, a heat flow meter sensor and internal and external surface and air temperature probes were installed on the walls [42–45] facing north-west. In particular, heat flow meter sensors were installed on the inner side of the walls, and surface temperature probes were installed on the inner and outer sides of the walls. In addition, internal and external air temperatures were monitored. The experimental campaign was carried out during the winter, specifically during January and February 2020. The schematic representation of the experimental setup is shown in Figure 3. Table 2 lists the technical specifications of the measuring instruments.

**Figure 3.** Schematic representation of the experimental setup in the reference test room (**a**) and in the insulated one (**b**).

**Table 2.** Technical Specifications of the Measuring Instruments.


The measurements of the thermal transmittances of the walls were carried out in compliance with the ISO 9869-1 Standard [46]. The acquired data were processed using the progressive averages method, applying the following formula:

$$\mathbf{U} = \frac{\sum\_{\mathbf{j=1}}^{n} \mathbf{q}\_{\mathbf{j}}}{\sum\_{\mathbf{j=1}}^{n} (\mathbf{T}\_{\text{ai}} - \mathbf{T}\_{\text{ae}})\_{\mathbf{j}}} \tag{1}$$

where **q** is the heat flow density, **Tai** and **Tae** are the temperature of the air inside and outside the analyzed test room, respectively.

The phase shift (briefly defined **PS**) of the thermal waves can be determined as the time difference between the recording time of the highest external surface temperature value (**h**\_**Ts maxe** ) compared to that which corresponds to the highest internal surface temperature (**h**\_**Ts maxi** ).

$$\text{PS} = \mathsf{h}\_{\text{T}\mathsf{s}\ \mathsf{max}\_{\mathfrak{v}}} - \mathsf{h}\_{\text{\tiny\mathsf{T}\mathsf{s}\ \mathsf{max}\_{\mathfrak{i}}}} \tag{2}$$

The thermal wave attenuation (briefly defined **DF**) can be calculated as the ratio of the difference between the maximum internal surface temperature (**Tsmaxi** ) and the average one (**Tsavgi** ), and the difference between the maximum external surface temperature (**Tsmaxe** ) and the average one (**Tsavge** ) [47]:

$$\text{DF} = \left[ \frac{\mathbf{T\_{s\_{\text{max}\_i}}} - \mathbf{T\_{s\_{\text{avg}\_i}}}}{\mathbf{T\_{s\_{\text{max}\_q}}} - \mathbf{T\_{s\_{\text{avg}\_q}}}} \right] \tag{3}$$

In order to carry out a complete and reliable measurements campaign, the thermal behaviors of the two test rooms were analyzed taking into account different scenarios in terms of operational times of the heating system (made with electric fan heaters properly shielded to avoid direct disturbing effects to the sensors).

The first analyzed scenario took five days; during this time, the heating systems was always switched off (this first scenario is defined in the following as *Free-Floating*).

In the second scenario (the so-called *On*), the heating systems were switched on for four days continuously, and at the end of this time, the cooling phase of the two structures was evaluated during the 3 following days.

Finally, in the third scenario (the so-called *On-Off*), the thermal behavior of the two structures was studied by switching on the heating systems for nine hours per day (switching on the fan heaters in the morning at 9.00 a.m. and switching them off at 06.00 p.m.).

#### *3.2. Energy Simulation Model*

The data obtained from the experimental campaign were employed to build a dynamic energy simulation model of an ideal building. The test rooms, in fact, are too small and not representative of a real residential building. The ideal building, which was used by the authors for simulations in previous works, has the same envelope thermal performance (thermal transmittance, phase shift, wave attenuation, etc.) as that of the monitored test rooms, but it is more representative of an actual building since it has transparent surfaces and plants that are essential in residential spaces.

An hourly energy simulation was performed using Design Builder software [48], a computational code based on Energy Plus as an internal simulation engine.

Design Builder was used for modeling a building larger than the actual buildings where measurements were carried out. A simple building with a square shape of a 6 m side, similar to a two-storey detached house already used in other studies [49], was considered as a case study (Figure 4). Each wall has a surface area equal to 36 m2, characterized by the stratigraphy listed previously in Table 1. The fifteen windows adopted in the model are double glazed windows (6 mm–6 mm filled with air in the gap and with a solar factor of 0.7), with a thermal transmittance of 3.094 W/m2K and an area of 1.44 m2 for each one; the frame is made of painted wood and is characterized by a thickness of 8 cm. The shadings of the windows are composed by shutters that are simulated as external systems.

**Figure 4.** 3D view of the case study used for simulation.

The following settings were adopted:


As Italian buildings are usually equipped with only heating system, a natural gas boiler was supposed for the heat generation, and the global efficiency of the system was set as equal to 0.83. In the energy model, an occupancy value of 0.02 people/m<sup>2</sup> has been defined.

The energy need of the building was simulated. Later, different insulating materials were tested, taking always into account a thickness of the insulating layer equal to 0.01 m (equal to the thickness of the aerogel panel tested during the in situ campaign). This choice allows the comparison of different insulation materials with equal saving of inner space in the case of internal application; the use of aerogel is, in fact, a competitive solution in the retrofit of historical buildings when the intervention on the external façade is not possible for architectural conservatory constraints [35]. In particular, the simulated insulating materials are: Expanded PolyStyrene (EPS), rock wool, kenaf and aerogel (whose thermophysical properties are shown in Table 3).



<sup>1</sup> Obtained as the mean of other materials' decay rate.

For the materials, a useful life of 45, 20, 25 and 15 years was considered, respectively, for aerogel, EPS, rock wool and kenaf. However, the materials may not be removed from the walls, and they could continue to partially carry out their task for the whole duration of the building. Therefore, for these kinds of interventions, a linear compound decay rate was estimated as equal to 0.21%, 0.20%, 0.25% and 0.17% per year, respectively, for aerogel, EPS, rock wool and kenaf (Table 3). As far as the duration is concerned, a duration of 50 years was considered for the building.

These insulating materials modified the walls' thermal transmittances, as reported in Table 4 (the insulating material is installed on the outer side of the wall, before plaster). According to this, an energy analysis was carried out to quantify the energy savings obtained by means of different insulating materials.

**Table 4.** U-Value of the Walls Considering Different Insulating Materials.


#### *3.3. Environmental Assessment Based on LCA*

Following the quantification of the energy savings obtained after the implementation of different retrofits, a life cycle assessment (LCA) was performed to determine the effectiveness of the intervention when considering the environmental burdens embodied in the building materials installed. The LCA is an interesting methodology that permits the comparison of the energy requirements of the buildings and the related environmental burdens from a more comprehensive perspective that takes into account the whole life cycle stage of the constructions (production, installation, operation, end-of-life). In fact, different authors have already warned about the burden shifting that characterizes every retrofit intervention [50,51]: the reduction of the operational energy requirement and related environmental burdens is followed by an increase in embodied components linked to the installation of new building materials and systems. Two indicators were introduced to describe the environmental performances of the different external insulation coatings supposed: the Energy Payback Time (*EPBT*) and the Carbon Payback Time (*CPBT*). The first one can be defined as the ratio between the variation of the Embodied Energy (*EE*) of the building following the retrofit and the annual Energy Savings (*ESa*) achieved through the retrofit (see Equation (4)). The latter is similarly the ratio between the variation of the Embodied Carbon (*EC*) of the building and the annual emissions avoided (*CSa*) through the retrofit (see Equation (5)).

$$\text{EPBT} = \frac{\Delta EE}{\text{ESa}}\tag{4}$$

$$\text{CPBT} = \frac{\Delta \text{EC}}{\text{CSa}} \tag{5}$$

The LCA analysis was carried out using Ecoinvent (Ecoinvent, Zurich, Switzerland) data, and when this was not possible, Environmental Product Declaration datasheets were consulted [52]. The EE was calculated using the single-issue indicator Cumulative Energy Demand (*CED*), while the Global Warming Potential (*GWP*) (100 years) was employed to determine the EC of the retrofit. As shown in Figure 5, a "cradle to site" approach was employed for the life cycle assessment. Since the application of external insulation coatings in a low-height building does not imply an energy intensive installation process, stage A5 can be considered negligible.


**Figure 5.** Life cycle stages considered in this study (green marked).

On the other hand stages B1–B5 were not included in the calculation of the payback times since, generally, they are much lower than the useful life adopted in this work for the insulation materials (see Table 3) [53].

#### *3.4. The Cost-Benefit Analysis*

From an economic-financial perspective, the international literature estimates the convenience of retrofit intervention in a partly incomplete manner.

Table 5 provides a systematic review of some typical studies.

The international literature on this topic is very ample and an in-depth analysis of it goes beyond the objective of this study. Therefore, only some studies are analyzed, which can be considered representative of different approaches.

Most studies consider only the energy savings resulting from retrofit interventions and some related items (initial investment, maintenance costs, running and replacement costs, etc.) which are measurable in monetary terms. They do not consider the environment benefits of a retrofit, nor, generally speaking, its impacts on the Internal Environmental Quality (IEQ), due to the difficulty and subjectivity of their economic measurement.

*Energies* **2021**, *14*, 1276


**Table 5.** Studies on the Economic Convenience of a Retrofit Intervention. (1) free-risk rate as a discount rate (corrected by inflation or not). (2) PI = Profitability Index: the ratio between the present value of net benefits and initial investment. (3) NPV= Net Present Value, that uses costcapital(risk-freerate+riskpremium)asadiscountrate.(4)LCCLife-CycleCosts.(5)IEQIndoorEnvironmentalQuality.

 =

 =

Some of them choose the intervention which guarantees the quickest recovery of the investment or the shortest payback time [55,56,61,63–65]. This method is quite easy to apply, but it shows two elements of weakness:


When a more complete approach is provided [54,57–59,62], the present value of differential costs/benefits is calculated (in [55,56,61,65] as a further method) by using a free-risk rate for discounting (often corrected by the expected inflation rate), which does not take into account a premium for the risk of the discounted cash flows.

The analysis here conducted applies a financial approach consistent with the modern financial theory. The net present value (NPV) is used (however, only considering the monetary costs and benefits of a retrofit, in line with the international literature), which measures today's monetary value of the intervention, and it discounts the net cash flows by a rate which considers the time value of the money and the risk premium, calculated with reference to the main risk drivers of the investment. The retrofit is convenient if the NPV is non-negative, and it is the more convenient the higher its value.

Furthermore, a probabilistic approach was used to measure the risk of the NPV of each retrofit alternative. Many studies [54,56,57,60–64] explicitly consider the uncertainty, more often with regard to the technical variables than the economic ones. Some studies consider different possible values of technical input variables (rarely of economic variables, as for example the discount rate in [57,61] and the initial investment and gas price in [57]) and estimate the resulting range of outcome measure, others [54,57,60] use very complex methods to deal with the uncertainty (various sensitivity analysis methods and Monte Carlo techniques), but they are quite methodological exercises: in fact, these techniques are very difficult to apply in a real context, since many of the necessary data cannot be realistically provided, and the approach is quite difficult for the decision-maker to understand.

In this paper, optimistic and pessimistic estimates of the uncertain drivers of NPV were assumed and defined in a subjective manner (i.e., on the basis of the analyst/decisionmaker's forecasts), and the corresponding range of NPV was measured. This analysis provides two useful results for a decision-maker:


Finally, as far as the IEQ aspects are concerned, a multi-criteria methodology (MCDA) is being developed that would measure for the different retrofit interventions the main descriptors of the IEQ in relative terms, with respect to the acceptable ranges defined by EU regulations. This approach would use linear optimization models in order to allow the decision-maker to compare the retrofit alternatives with each other and with the current state of a building.

#### **4. Results and Discussion**

#### *4.1. Experimental Campaign Results*

As mentioned before, the monitoring campaign was carried out during the winter, specifically during the months of January and February 2020, and it was focused on the assessment of the thermal behavior of the studied test rooms in three different scenarios: *Free-Floating* conditions (no heating in the two test rooms), the so-called On scenario (heating system always on) and the so-called On-Off scenario (heating system on only during a specific daily time interval). The obtained results during winter can be summarized as follows:

• Free-floating conditions: Data processing in this phase mainly focused on defining the thermal waves' phase-shift and attenuation according to Equations (2) and (3). In particular, the surface temperature values were analyzed, and their trend over time is reported in Figure 6, where the internal and external surface temperatures for the reference test room are called *Tsi\_ref* and *Tse\_ref*, respectively.

On the other hand, the internal and external surface temperatures for the insulated test room are called *Tsi\_aerogel* and *Tse\_aerogel*, respectively.

It is clear that the internal surface temperatures immediately present a stabilized periodic regime. In particular, the internal surface temperature data measured on the thermally insulated test room provide almost constant values along time, mostly lower than those registered on the reference test room.

Both attenuation and phase shift values were calculated with respect to a daily interval, while the final average value (shown in Table 6) was calculated as the average of the daily attenuation and phase shift values.

**Table 6.** Attenuation and Phase-Shift Average Values Obtained under Free-Floating Conditions.


By comparing these data, it is possible to observe that applying a thin layer of aerogel does not cause a significant variation of the thermal inertia of the wall. In fact, the thermally insulated test room is characterized by an average phase shift just 20.6% higher than the reference one. On the contrary, one centimeter of aerogel, due to its high insulating performance, produced a decrease in the average attenuation of about 64.5% compared to that calculated for the reference test room, with a better indoor air temperature steadiness. Thus, it is possible to affirm that the application of a thin layer of aerogel can improve the dynamic behavior of the structure. However, this is not surprising. In a steady state regime, the layer arrangement makes no difference. On the contrary, under dynamic boundary

conditions, the layer arrangement becomes fundamental, and by interchanging the layers the wall properties change. Hence, this aspect needs to be considered for improving the inertial behaviour of a wall.

• On scenario: The second phase was related to the investigation of the thermal behavior of the two test rooms with the heating always on. In this case, the progressive increase in the air temperature of the two different test rooms was focused, as shown in Figure 7 (before the vertical black dotted line).

**Figure 7.** Indoor air temperatures registered in the test rooms during the second scenario.

Taking into account the thermally insulated test room, it is possible to observe a faster internal air temperature rise if compared with the reference test room. At the end of the always-on heating period, a stabilized regime was not achieved in the thermally insulated test room, as the internal temperature gradually increased. The use of the aerogel led to an internal air temperature of about 33 ◦C, compared to about 27 ◦C (average value) obtained in the reference test room, where an almost periodic regime was identified after about two days.

The absence of the external insulating coat made the reference test room more sensitive to the typical variations of the outdoor air temperatures also during the heating system shutdown. Figure 7 (after the vertical black dotted line) shows a more rapid decrease in the values of the internal air temperature, as expected.

• On-Off scenario: The last part of the winter monitoring was aimed at evaluating the thermal behavior of the two test rooms, assuming that the heating system was switched on and off; i.e., switching on the fan heaters in the morning and switching them off in the evening, thus simulating the irregular working of an actual heating system. The acquired data were employed for evaluating the thermal transmittance of the walls facing north-west. Figure 8 shows the thermal transmittances as a result of the data post-processing based on the progressive average method. The thermal insulation of the test room through the thin layer of aerogel allowed obtaining a thermal transmittance reduction equal to −28.3%.

**Figure 8.** Thermal transmittances obtained from the U-value measurements.

#### *4.2. Energy and LCA Results*

Table 7 and Figure 9 show the results obtained through the software Design Builder. In the reference non-insulated building an energy demand for heating equal to 11,621.5 kWh/year of natural gas was obtained. The energy analysis shows how the installation of different insulating materials allows the reduction in the energy need of the building. Quite similar percentage reductions were obtained when the structure was insulated with EPS, rock wool and kenaf panels. Similar energy savings of about 6% are not surprising, because of similar thermal conductivity values among the different insulating materials. Aerogel, due to its reduced thermal conductivity, allowed achieving a heating energy need of 10,313.7 kwh/year. A percentage difference in terms of heating energy requirement equal to −11.3% was obtained—almost double that obtained using the other insulating materials.

**Table 7.** Data on the Energy Need and Energy Saving of Insulated Buildings Compared with the Reference.


**Figure 9.** Energy needs of the reference building and the insulated one.

The results on the EPBT and CBPT are reported in Table 8. The retrofit with aerogel is characterized by a higher EPBT and CPBT in comparison to traditional insulation materials (e.g., rock wool or EPS). As previously supposed in Section 3.3, the values obtained are lower than the supposed useful life of the insulation materials installed. This means that the burden shifting on embodied components is only temporary and that every intervention is characterized by a positive environmental effect on its life cycle. Figures 10 and 11 show the total CED and the cumulative GWP versus time considering as positive values the energy saved and the emissions avoided: the coating with traditional insulation materials has a lower payback time in comparison with the scenario considering aerogel as insulation material, but the latter results guarantee, after about fourteen years from the installation, a higher energy saving and carbon emission reduction potential.


**Table 8.** Energy and carbon payback times of the various retrofit solutions.

**Figure 10.** Cumulative energy demand for every retrofit intervention (positive values stand for energy saved).

**Figure 11.** Cumulative GWP for every retrofit intervention (positive values stand for emissions avoided).

#### *4.3. Economic Analysis Results*

In order to calculate the NPV, the differential costs and benefits of the four analyzed materials (compared to the current state) were estimated.

In order to estimate benefits, the annual savings related to the methane gas consumption and the tax incentives of "Bonus Facciate" [68] were considered.

The price of methane gas was obtained by the average of prices applied by different suppliers in 2020, equal to 0.0985 euro per kWh. From the historical price series of methane gas over the last 10 years (Eurostat data [69]), an annual growth rate of methane gas of 1.77% was obtained, applied during the whole duration of the retrofit (as a trend estimate).

For the duration and decay rates of the insulating materials, the values reported in Table 3 were adopted.

To estimate the tax incentives, based on the current regulation, a tax deduction of 90% of the total expenditure was assumed over 10 years.

To estimate the cost of capital, the Capital Asset Pricing Model approach has been used, with the following parameters:


$$\text{Beta} = \frac{\text{correlation}\_{\text{gas price, GDP}} \times \text{volatility}\_{\text{gas price}}}{\text{volatility}\_{\text{GDP}}} \tag{6}$$

$$\text{Beta} = \frac{0.77 \times 0.1876}{0.2236} \cong 0.65\tag{7}$$

A cost of capital of 4.42% was obtained as Equations (8) and (9) show:

cost of capital = risk free rate + beta × market risk premium (8)

$$1\ \text{cost of capital} = 1.18 \div 0.65 \times 5 = 4.42\% \tag{9}$$

which was used as a discount rate of the energy savings. The tax incentives were discounted by the risk-free interest rate since they are relatively certain. Table 9 summarizes the costs, benefits and NPV of the four materials.


**Table 9.** Calculation of the Net Present Value (NPV) of materials (data in euros).

<sup>1</sup> Estimated for a surface of 121 m2.

It is possible to conclude that in this scenario aerogel gives the best economic benefits, with a positive expected NPV of EUR 1417.55.

As far as the intervention risk is concerned, a sensitivity analysis was implemented for retrofit interventions of aerogel and rock wool (on the basis of the results of the previous analysis, the latter is the best among the alternatives to aerogel). Optimistic and pessimistic estimates of the main uncertain drivers of the NPV have been forecasted: duration, methane gas price, cost of capital and tax incentive. The assumptions were the following:



**Table 10.** Sensitivity Analysis of the NPV.

<sup>1</sup> Data in euros.

The last two columns measure the coefficients of NPV sensitivity; i.e., how much each driver variability influences the NPV variability.

Figure 12 shows the cumulative probability distribution of NPV of each retrofit, where NPV variance is measured following Equation (10) (by simplifying, the uncertain drivers are assumed to be independent of each other and linearly related to NPV):

$$
\sigma\_{\rm NPV}^2 = \sum k\_i^2 \times S\_i^2 \tag{10}
$$

where *Si* = NPV range between the optimistic estimate *Ui* and the pessimistic one *Li* of uncertain driver *i* (see columns 8–9 in Table 9) and, where *σ<sup>i</sup>* is its volatility (in this case, *ki* is equal to 0.3).

**Figure 12.** NPV cumulative probability distributions of retrofit with aerogel and rock wool.

The main results are the following:


Due to the crucial effect of tax incentives on NPV of both retrofits here assessed, we further analyzed how NPV changes, depending on tax deduction (the most probable values of the other drivers were considered). Figure 13 shows that the aerogel NPV: (i) becomes the more advantageous, in comparison to rock wool, the higher the tax deduction; (ii) is positive for tax deduction larger than 80% (70% for rock wool, instead); and (iii) beats the competing material when tax deduction is larger than 87%. This analysis is important, since it shows that aerogel material is more convenient than competitors only in the fiscal framework here considered (or in a more favorable one, as for example in the case of the Superbonus 110%), while in other scenarios it is not, due to its higher lump-sum investment (even though it provides double energy savings than the alternative materials).

**Figure 13.** NPV as a function of tax incentives.

#### **5. Conclusions**

A small test room, totally insulated with aerogel panels, was investigated by an experimental point of view. The thermal behavior of the aerogel insulated test room was compared with a non-insulated identical test room. Heat transfers across walls were assessed by installing heat-flow meters and air and surface temperature sensors. Experimental data verified the well-known aerogel capability to improve the thermal performance of test room envelopes, even if reduced thicknesses of thermal insulation were applied. In particular, the thermally insulated test room showed an average phase shift 20.6% higher than the reference one. On the other hand, the small layer of aerogel allowed obtaining an average attenuation decrease of about 64.5% when compared to the reference test room. Moreover, 1 cm of aerogel allowed to obtain a thermal transmittance reduction of −28.3%.

Successively, a dynamic simulation model was generated to better comprehend the energy savings obtained through aerogel across a whole year, in terms of energy demand. By comparing aerogel with other commonly used insulation materials, a heating energy demand reduction of −11.3% was found.

Subsequently, the investigation was completed by computing the environmental and energy payback times of this retrofit strategy as well as by a cost-benefit analysis through a probabilistic financial framework. In sum, it is possible to conclude that in the current tax incentive scenario the aerogel retrofit gives the best positive expected NPV, and from the whole LCA perspective it also guarantees both the highest energy saving and emissions avoidance. However, due to its higher lump-sum investment, aerogel's NPV is very sensitive to tax deductions and it is riskier than the best comparable material (roof wool) in less favorable tax scenarios: for example, if a 65% tax deduction is assumed (given the probability distributions of the uncertain variables considered here), aerogel gives a non-negative NPV only in 12% of the cases (roof wool in 38% of the cases, instead).

Therefore, the proposed interdisciplinary study aimed to investigate the environmental, energy and economic impacts associated with the application of aerogel compared to other insulating materials. Furthermore, the proposed methodological approach could also be replicated in other countries characterized by different climatic conditions.

Future developments will address a comparison of the different energy, environmental and economic benefits of aerogel under different climatic conditions, also applying thin aerogel panels to different wall stratigraphy. In addition, the analysis could be performed also considering the energy, environmental and economic benefits of aerogel by analyzing walls thermally insulated with materials of different thicknesses but characterized by the same U-value.

**Author Contributions:** Conceptualization, P.M. (Paola Marrone), F.A. and D.V.; methodology, P.M. (Paola Marrone), F.A., D.V., L.F. and R.D.L.V.; thermal measurements, L.F., L.E., C.G., M.R. and F.O., energy simulations, G.G. and M.R.; economic-financial analysis: P.M. (Paolo Matteucci) and D.V.; data curation, L.E., C.G. and M.R.; writing—original draft preparation, L.E., C.G., M.R., F.O., P.M. (Paolo Matteucci) and D.V.; writing—review and editing, F.A., M.R., G.G., L.E., P.M. (Paolo Matteucci) and D.V.; supervision, P.M. (Paola Marrone), F.A., D.V., L.F. and R.D.L.V.; project coordination, P.M. (Paola Marrone). All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data available on request due to restrictions. The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy reasons.

**Acknowledgments:** The authors would like to thank Eng. Alfredo Simonetti and all the staff at CEFME CTP for the valuable collaboration.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Energies* Editorial Office E-mail: energies@mdpi.com www.mdpi.com/journal/energies

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18