Integrated Energy and Environmental Modeling to Design Cost-Effective Building Solutions at a Regional Level

Abstract

This study introduces a computationally efficient urban building energy model (UBEM) to assess decarbonization strategies for the residential sector at the regional level. The model considers a range of inputs, including building characteristics, climate data, technology penetration, and occupant behavior. The model provides an economic analysis associating emission reduction potential with economic returns through an abatement cost curve, which is critical to designing cost-effective solutions. The model was validated at its full scale in Portugal, using actual consumption data from all municipalities. Key findings showed that lighting upgrades (100% LEDs) are the most cost-effective measure, offering the lowest abatement cost (−521 EUR/tonCO₂eq) and a low discounted payback period of 2 years, while heat pumps for water heating provide the highest emission reduction potential, with an annual reduction of 863 tonnes of CO₂eq annually, equivalent to a 20% reduction in national emissions. Additionally, behavioral measures achieved an annual reduction of 147 tonnes of CO₂eq. The analysis further reveals that, while some measures might have a negative abatement cost at the national level, their economic viability varies locally, with certain municipalities incurring positive abatement costs, highlighting how local context affects the economic viability of decarbonization strategies.

1. Introduction

In recent years, the effects of global warming have become increasingly worrying for humanity as climate change continues to intensify. It has led to the creation of plans and agreements by governments worldwide to address the issue, such as the Paris Agreement, which aims to limit global warming by limiting global temperature increase to well below 2 degrees Celsius [1]. Achieving these climate targets requires coordinated efforts across multiple sectors, each facing its own set of challenges. While sectors like transportation struggle with the complexity of energy consumption and carbon reductions, the building sector has been recognized as having the potential to make a greater contribution to energy and emission reductions [2].

Within the European Union (EU), buildings account for approximately 40% of the total energy consumption and 36% of greenhouse gas (GHG) emissions [3,4]. As a result, the residential sector holds significant potential for energy savings and emission reductions. Addressing this potential is essential to reach the EU’s ambitious climate targets. Policies promoting energy efficiency in buildings, combined with the transition to renewable energy, are seen as crucial steps towards reducing emissions and improving energy performance.

In this context, urban building energy models (UBEMs) are essential tools for analyzing residential energy consumption and emissions, supporting the development of effective policies at scales ranging from individual buildings to neighborhoods or cities [5]. UBEMs enable researchers and policymakers to simulate alternative scenarios for evaluating decarbonization measures, such as energy efficiency upgrades, retrofitting, and renewable integration [6,7].

Several UBEM tools have become widely adopted for urban applications [5]. For example, CityBes, developed by Hong et al. [8], is a platform intended to simulate the energy performance of building stocks across cities, supporting the evaluation of retrofit strategies and energy-saving interventions. In a recent study, Li et al. [9] expanded CityBes capabilities to include district heating and cooling systems, enabling city planners to model district-level energy configurations and assess their impact on energy demand and emissions. Additionally, Fonseca and Schlueter [10] developed CEA, a UBEM designed to simulate energy consumption and emissions at the urban scale, with a case study in Zurich. Their model integrates spatial and temporal data, providing detailed insights into urban districts’ potential for reducing emissions through a range of energy efficiency measures and interventions. Other widely used UBEM platforms include UMI, SimStadm, and CitySIM [5].

While these UBEM platforms offer valuable insights at the neighborhood, district, and city levels, extending them to regional or national applications remains a significant challenge due to several key factors. Johari et al. [11] highlight significant data requirements and computational complexity for UBEM even at the city scale. These challenges suggest that scaling such models to a national level would be substantially more difficult. Additionally, Ali et al. [12] point out that models based on detailed simulations require substantial computational resources, thereby limiting their scalability to larger regions or national levels. Sola et al. [13] similarly note that, while detailed UBEM simulation engines predict the operational energy use of individual buildings very well, it is common for difficulties to appear when scaling up to modeling at larger scales. These difficulties, along with the increased computational requirements, make it challenging for traditional UBEMs, designed primarily for single cities or districts, to maintain efficiency at larger scales, often leading to significantly longer run times.

For instance, Chen et al. [14] simulated 940 buildings in San Francisco using CityBES, which required hours to days, depending on the level of detail and available computational infrastructure. For larger simulations, like the 177,023 buildings across the entire city of San Francisco, the computational demands became even more pronounced.

While current UBEMs perform well at the city or district level, they often lack the computational efficiency necessary for national-scale deployment without the support of high-performance infrastructure. A study by Beltrán-Velamazán et al. [15] further underscores this limitation, highlighting that national-scale UBEMs, though valuable for estimating energy consumption across a country’s building stock, demand supercomputing capabilities to handle simulations efficiently.

This paper addresses this gap by presenting a computationally efficient model capable of simulating multiple regions within a country, accounting for each region’s specific context, and even extending to national-scale simulations. The model was designed to be computationally efficient, enabling simulations at a national scale in less than one minute while still maintaining the necessary detail for the analysis of cost-effective building solutions. This allows the model to bridge the gap between highly localized urban models and broader national models, providing insights into energy consumption and emissions across larger geographical areas without the typical computational constraints of more detailed models.

In addition to its computational efficiency, the model introduces an economic analysis through the use of abatement cost curves. This feature links environmental performance with economic feasibility, providing policymakers with critical insights into the cost-effectiveness of various building solutions designed to reduce CO₂eq emissions.

To demonstrate its applicability, the model was applied to Portugal with desegregated data at the municipal level, covering all 308 municipalities. Portugal was selected as a case study to highlight the model’s adaptability in accounting for diverse regional factors, such as local climate variations, building characteristics, and penetration technologies. The simulation of decarbonization measures reveals valuable information about the energy savings, CO₂eq emission reductions, and cost-effectiveness of various strategies, helping to design the most effective approaches across different contexts. These findings are crucial for guiding resource allocation and informing targeted policy decisions.

2. Materials and Methods

The methodology for developing a regional residential energy model is outlined in this chapter. It includes the analytics used to determine energy consumption in each end use of a dwelling, which are derived from established research.

The model takes into account a variety of inputs, such as the building attributes, climate, and occupant behavioral trends. This hybrid approach is expected to provide an accurate estimate of residential energy consumption by combining both physical and behavioral factors. The model was implemented in Python (version 3.10).

Figure 1 showcases the diagram of the regional residential energy model, providing a visual representation of its various components and their interconnections.

The model estimates the useful energy consumption of residential buildings using inputs such as building constructive solutions, climate data, and occupant behavior. Useful energy refers to the amount of energy required to meet a specific demand (e.g., heating a space or providing hot water), while final energy is the amount of energy consumed via the system to fulfill that demand, considering the efficiency of the equipment.

When evaluating energy consumption for cooking and household appliances, it is important to understand that these calculations are based on final energy consumption, not useful energy. To calculate the useful energy for these two end uses would require detailed information on factors like the types of food being prepared and the materials and efficiency of each appliance. These details are necessary because different foods and cooking methods consume varying amounts of energy, and appliances made from different materials operate at different efficiencies. Collecting these comprehensive data on a national scale is unfeasible, so final energy consumption was adopted for these two end uses as a more practical approach.

Space heating and water heating are often the biggest contributors to overall dwelling energy consumption [16]; according to the IEA, space and water heating accounts for 46% of the energy consumed in buildings [17]. In Portugal, water heating consumption accounts for 20% of the total residential energy consumption, while space heating accounts for 23% [18].

The thermal dynamics of a building involve heat-transfer processes that determine how energy is gained and lost. Heat gains include solar radiation, which heats building surfaces like windows, walls, and roofs, and internal gains from occupants, appliances, lighting, and cooking. The solar heat gains can reduce space heating needs but may cause thermal discomfort during warmer periods.

Heat losses occur through the building envelope (e.g., walls, windows, and roofs), driven by the temperature gradient between the interior and exterior. In cold conditions, heat is lost through conduction, convection, and air infiltration, while in warm conditions, heat can flow inward, increasing cooling demand. Understanding these thermal dynamics is essential to accurately estimating the space heating and cooling demands of the residential building stock.

In the following sections, the model formulation is presented in detail, outlining the mathematical equations and methodologies used to estimate the final energy demand for various end uses in residential buildings. The model covers space heating, space cooling, water heating, lighting, household appliances, and cooking. Each subsection delves into the specific factors and parameters influencing energy consumption for these end uses, such as thermal losses, solar gains, equipment efficiency, and occupant behavior.

2.1. Space Heating

The equations required to compute space heating final energy were adapted from the Regulation on the Energy Performance of Residential Buildings (REH) [19] and other studies [20,21].

Equation (1) was used to calculate space heating’s final energy demand.

$$Q_{SH}={\displaystyle \frac{\left(Q_{t,sh}+Q_{v,sh}-Q_{g,sh}\right)}{{\eta }_{T,sh}}}$$

(1)

where $Q_{t,sh}$ (kWh) represents the losses through transmission, $Q_{v,sh}$ (kWh) is the quantity of heat lost as a result of the ventilation system of a building, $Q_{g,sh}$ (kWh) represents the heat gains, and ${\eta }_{T,sh}$ is the space heating technology efficiency.

The heat losses through transmission are calculated through Equation (2).

$$Q_{t,sh}=0.024·HDD·H_t$$

(2)

where HDD (°C· Day) represents the heating degree days, and $H_t$ (W/°C) is the global heat coefficient.

The heating degree days are a measure of the severity of the outdoor temperature in a given location during a specified time period. The use of heating degree days from the REH ensures that the energy model represents the typical heating demands of buildings in Portugal.

The global heat coefficient is calculated through Equation (3).

$$H_t=U_{wall}·A_{wall}+U_{window}·A_{window}+U_{floor}·A_{floor}+U_{roof}·A_{roof}$$

(3)

where $U_{wall}$, $U_{window}$, $U_{floor}$, and $U_{roof}$ (W/m²·°C) are the overall heat transfer coefficient of the walls, windows, floors, and roofs of a dwelling, and $A_{wall}$, $A_{window}$, $A_{floor}$, and $A_{roof}$ (m²) are the respective areas of each of the elements.

The heat losses through ventilation are calculated with Equation (4).

$$Q_{v,sh}=0.024·HDD·{\displaystyle \frac{c_p·\rho }{3600}}·ACH·Vol$$

(4)

where ACH is the air changes per hour (h⁻¹), $C_p$ (1.005 kJ/kg· K ) is the specific heat of air, $\rho$ ($1.2{\displaystyle \frac{kg}{m^3}}$) is the density of air, and $Vol$ (m³) is the volume of the building under analysis.

The heat gains are calculated using Equation (5).

$$Q_{g,sh}={\eta }_{gu}·\left(Q_{in}+Q_{sol,sh}\right)$$

(5)

where $Q_{in}$ (kWh) are the internal heat gains, $Q_{sol,sh}$ (kWh) is the solar irradiation heat gains, and ${\eta }_{gu}$ is the gain utilization factor.

The gain utilization factor is a ratio of the thermal gains and losses of a building in the heating season, and it is calculated through the following Equation (6).

$${\eta }_{gu}=\left\{\begin{array}{cc}if\gamma \ne 1\wedge \gamma >0& {\displaystyle \frac{1-{\gamma }^a}{1-{\gamma }^{a+1}}}\\ if\gamma =1& {\displaystyle \frac{a}{a+1}}\\ if\gamma <0& {\displaystyle \frac{1}{\gamma }}\end{array}\right.$$

(6)

$$\gamma ={\displaystyle \frac{Q_g}{Q_t+Q_v}}$$

(7)

where “a” can take the values of 1.8, 2.6, or 4.2, according to buildings’ weak, average, and strong thermal inertia [19].

Thermal inertia, which represents a building’s capacity to store and gradually release heat, is influenced by the U-value of the building envelope. Lower U-values, typically found in well-insulated buildings, enhance thermal inertia by reducing the heat transfer through the envelope [22]. A study by Giama et al. [23] showed that older buildings tend to have lower U-values due to outdated construction materials and methods, such as less insulation and single-pane windows, resulting in poorer thermal performance and lower thermal inertia.

Internal heat gains can significantly affect a building’s energy balance and indoor thermal environment, and they are calculated through Equation (8).

$$Q_{in}=0.72·q_{int}·M_{sh}·\overline{A_{floor}}$$

(8)

where $q_{int}$ (W/m²) are the internal gains, $M_{SH}$ represents the duration of the heating season, measured in months, and $\overline{A_{floor}}$ (m²)) is the average floor area.

The internal heat gains are assumed as 4 W/m² [24]; this value is used in energy performance calculations, and it represents the average heat generated via human occupation, cooking, and electrical and lighting equipment.

$Q_{sol,sh}$ (kWh) is the solar heat gains, and it is calculated with Equations (9)–(11).

$$Q_{sol,sh}=G_{south}·\sum _w\left[X_i·F_h·F_o·F_f·A_w·F_g·g_i\right]·M_{sh}$$

(9)

$$F_s=F_h·F_o·F_f$$

(10)

$$A_s=A_w·F_g·F_i$$

(11)

where $G_{south}$ (kWh/m²· month) is referred to as the average monthly solar radiation through the heating season on a vertical south-facing surface, $X_i$ is the window orientation coefficient, $F_h$ is the horizon window obstruction factor, $F_o$ and $F_f$ are window obstruction factors from horizontal and vertical elements, respectively. $A_w$ (m²) is the window area, $F_g$ is the glassed fraction of the window, and finally, $g_i$ is the solar factor, a coefficient that accounts for properties of the glass and shading elements, while $F_s$ is the total obstruction factor, and $A_s$ (m²) is the effective area.

The majority of the coefficients utilized were obtained from [19]. The coefficients that required calculations will be discussed in the following subsections.

2.2. Space Cooling

In the case study through which the model was tested, space cooling consumption accounted for a very small (<5%) [18] portion of the total energy consumption. However, since the intention was to create a replicable methodology, it was still important to include space cooling consumption in the analysis. In other contexts where the model may be applied, this end use may not be negligible, and modeling it ensures the model’s adaptability to different contexts.

The space cooling final energy demand of a building is provided via Equation (12).

$$Q_{SC}={\displaystyle \frac{(1-{\eta }_{gu})·Q_{g,sc}}{{\eta }_{T,sc}}}$$

(12)

where ${\eta }_{T,sc}$ depicts the efficiency of the cooling technology, ${\eta }_{gu}$ is the gain utilization factor for the cooling season, and $Q_{g,sc}$ (kWh) are the solar gains.

Thus, the following Equation (13) is employed to calculate the solar gains:

$$Q_{g,sc}=\sum _w\left[G_{solar}·F_s·A_{s,w}\right]+\sum _o\left[G_{solar}·F_s·A_{s,o}\right]$$

(13)

where $G_{solar}$ (kWh/m²) is the solar gains coefficient, $A_{s,o}$ (m²)) is the effective area of the opaque elements, and $A_{s,w}$ (m²)) is the effective area of the glazed elements.

The first parcel of the formula refers to the windows, while the second parcel refers to the opaque elements. The solar gains of the window elements for space cooling are computed in exactly the same manner as for space heating.

The solar gains coefficient is an assessment of how effectively a window or other building material transmits solar radiation into a structure. It represents the percentage of incident solar radiation that passes through a medium and can be used to heat or cool the inside environment. This coefficient is dependent on a variety of parameters, such as the glazing’s material characteristics, the angle at which solar radiation impacts the building, and the existence of shading devices or other building components that might reflect or obstruct solar radiation.

The effective areas are calculated with Equations (14) and (15).

$$A_{s,w}=A_w·F_g·g_i$$

(14)

where $A_w$ (m²)) is the window area, $F_g$ is the glassed fraction of the window, and $g_i$ is the solar factor.

$$A_{s,o}=U.\alpha ·A_o·R_{es}$$

(15)

where $A_o$ (m²) is the surface area, $\alpha$ is the solar radiation absorption coefficient, $R_{es}$ (m²·°C/W) is the external surface thermal resistance of opaque element, and U (W/m²·°C) is the overall heat transfer coefficient.

In order to compute the heat gain utilization factor, it is first necessary to calculate the heat transmissions using transmission, ventilation, and infiltration from Equations (16) and (17), respectively.

$$Q_{t,sc}=0.72.M_{sc}·H_t\left(T_{ref}-T_{ext}\right)$$

(16)

$$Q_{v,sc}=0.72·M_{sc}·ACH·{\displaystyle \frac{C_p·\rho }{3600}}\left(T_{ref}-T_{ext}\right)·Vol$$

(17)

where $M_{sc}$ is the duration of the cooling season, usually expressed in months, $T_{ref}$ (°C) is a reference temperature set for the cooling season, and $T_{ext}$ (°C) is the average outside temperature in the cooling season. The $T_{ref}$ (°C) for space cooling is set to 25 °C, according to the REH [19].

2.3. Water Heating

The energy needed to heat the water used in a residential building is referred to as the water heating energy demand.

The water heating energy demand for the residential building stock is calculated through Equation (18).

$$Q_{WH}={\displaystyle \frac{C_{p,w}·{\rho }_w·\Delta T·d·n_r·V_w·f_w}{3600·{\eta }_{T,wh}}}$$

(18)

where $C_{p,w}$ (4.186 kJ/kg·K) is the specific heat of water, ${\rho }_w$ (1 kg/l) is the density of water, $\Delta T$ (°C) is the water temperature difference desirable, $n_r$ is the number of residents in the dwelling, $V_w$ (l/person.day) is the volume of water required daily for each resident, $f_w$ is a factor for hydraulic efficiency, and finally, ${\eta }_{T,wh}$ represents the water heating technology efficiency.

2.4. Lighting

The lighting requirements for a dwelling can fluctuate, depending on the size of the home, how the rooms are organized, and the individual requirements of the occupants. The same formulation as in [21,25] was used. Equation (19) calculates the annual energy demand for lighting in a dwelling.

$$Q_L=0.001·AFA·\sum _{Tl}\left[{\displaystyle \frac{S_{Tl}}{{\eta }_{Tl}}}\right]·L·T$$

(19)

where $Q_L$ (kWh) is the annual energy demand for lighting for a single dwelling, $AFA$ (m²) is the average floor area of the dwelling, $S_{Tl}$ is the share of the lighting technology, and ${\eta }_{Tl}$ (Lm/W) is the corresponding lighting efficiency.

Finally, L (Lux) is the lighting requirement, and T (hours/years) is the equivalent amount of lighting hours required per year. According to Daioglou et al. [26], the combination of the last two factors can be expressed as the average lighting requirements, measured in lumen hours per square meter.

2.5. Cooking

The quantity of energy required for cooking depends on an array of factors, such as the type of cooking appliance used, the frequency and duration of cooking, and the efficiency of the cooking appliances used.

Most dwellings include more than one piece of equipment for food preparation; thus, the total amount consumed for cooking is the sum of all the appliances’ consumption [25], given by Equation (20).

$$Q_C=\sum P_{c,equip}·S_{c,equip}$$

(20)

where $Q_C$ (kWh) is the annual final energy consumption for cooking, $P_{c,equip}$ (units/dwelling) is the equipment penetration, and $S_{c,equip}$ (kWh/year) is the specific energy consumption of given cooking equipment.

2.6. Abatement Cost

The CO₂eq cost of abatement is the amount spent to reduce GHG emissions, usually expressed in terms of how much it costs per tonne of CO₂eq that is prevented from being released into the atmosphere. It represents the cost of implementing measures or technologies to reduce emissions. Measures with a negative abatement cost on an abatement curve are often referred to as “no-regret” measures since they simultaneously provide economic and environmental benefits. This can be the result of energy efficiency improvements that lead to cost savings that exceed the costs associated with implementing them [27].

Equation (21) was employed to calculate the net abatement cost of each measure under examination.

$$AC={\displaystyle \frac{I-{\sum }_1^n\frac{S_t}{{(1+r)}^n}}{\Delta E}}$$

(21)

where $AC$ (EUR/tonCO₂eq) represents the abatement cost, I (EUR) is the total investment cost to implement the given measure, $S_t$ (EUR) denotes the annual energy cost savings achieved by the measure, and n is the lifetime of the measure, expressed in years; this represents the expected operational duration of the measure before it needs to be replaced or upgraded. In other words, it is the number of years during which the measure will generate energy savings and emission reductions, r is the discount rate applied to adjust future savings to their present value, and $\Delta E$ (tonCO₂eq) refers to the total emission reductions achieved through the measure over its lifetime.

The discount rate, r, is a key factor in evaluating the economic feasibility of measures, as it determines how future costs and benefits are valued in present terms [28]. Over time, the discount rate applied in energy efficiency and decarbonization projects has generally decreased, both at the European level and globally [29]. This reflects a growing recognition of the importance of long-term environmental sustainability and societal benefits. Historically, higher discount rates reflected short-term financial returns, often devaluing projects with significant future environmental and social gains. However, with an increasing focus on climate change and the transition to low-carbon economies, the rate has been progressively lowered, incentivizing investments in measures with longer payback periods but substantial future benefits.

According to the European Commission’s Guide, Economic Appraisal Vademecum 2021–2027—General Principles and Sector Applications, the use of lower discount rates is recommended for evaluating decarbonization and energy efficiency projects [30]. This approach emphasizes the need to account for the long-term societal and environmental benefits that such projects deliver. Specifically, by applying discount rates between 3% and 5%, the EU ensures that the future benefits, including reduced greenhouse gas emissions and improved air quality, are adequately reflected in cost–benefit analyses [30]. These benefits often take several decades to materialize, particularly in projects with long lifetimes, such as building renovations and renewable energy systems.

2.7. Model Validation

A validation step in an energy model is essential to ensure the model’s accuracy and reliability. The error (%) was calculated for each region with the following Equation (22).

$${\mathrm{Error}}_i(\%)={\displaystyle \frac{({\widehat{C}}_i-C_i)}{C_i}}\times 100$$

(22)

where i represents each of the regions modeled, ${\widehat{C}}_i$ (kWh) represents the consumption estimated by the model for the i-th region, and $C_i$ (kWh) is the actual consumption for the i-th region.

3. Case Study

In the case study section, the focus is on characterizing the specific context to which the model was applied, namely at the municipality level in Portugal. This section provides an overview of the building stock, climate data, and occupant behaviors, offering a detailed understanding of the conditions under which the model operates. However, while this case study focused on Portugal, the methodology is designed to be replicable for other contexts, provided that the necessary input data are available. This flexibility demonstrates the model’s adaptability and its potential application to diverse geographical areas for energy and emissions assessments.

3.1. Demographic and Geographic Characteristics of Portugal

According to INE [31], as of 2021, the population of Portugal is distributed as 9,855,909 in the Portuguese mainland, 236,413 in the Azores autonomous region, and 250,744 in the Madeira autonomous region.

Therefore, the population of Mainland Portugal is by far the largest, accounting for 95.30% of the total population of the country. The remaining 4.7% is divided between the two autonomous regions of Portugal, with the Madeira autonomous region having a slightly larger population than the Azores autonomous region.

With a total area of approximately 92,212 km², the mainland region occupies approximately 89,015 km², while the two autonomous regions—Azores and Madeira—span 2322 and 801 km², respectively.

The mean annual temperature for mainland Portugal is approximately 17 °C, some variations may occur, depending on the location and season. Particularly in the Southern regions, such as Algarve, where mean temperatures can range between 18 °C and 20 °C throughout the year. The cooler climates like the ones observed in northern areas like Porto and the Douro Valley experience an average temperature of 13 °C to 17 °C [32].

3.2. Portugal’s Residential Building Stock

A comprehensive characterization of the residential building stock in Portugal is critical for a successful energy model because the energy performance of buildings is largely determined by their physical and thermal properties. Factors such as building age, construction materials, insulation quality, window types, and heating or cooling systems directly impact how buildings gain and lose heat, and consequently, their energy consumption for space heating and cooling.

An accurate characterization of the building stock is the first step to developing an accurate energy model. By reflecting the diversity of building types and conditions across different regions, this characterization allows the model to perform precise simulations and provide reliable predictions of energy consumption, potential energy savings, and emission reductions.

In Portugal, there are currently 4,142,581 dwellings that are primary residences, according to the 2021 Census [31]. The term primary residence refers to a dwelling or housing unit where a resident usually lives or stays for the majority of the year.

Figure 2 represents the number of dwellings of primary residence per year of construction; the data were extracted from [31].

From Figure 2, it is clear that the majority of primary dwellings in Portugal were constructed between 1961 and 1980. This period is described as the “construction boom”, which occurred as a result of the country’s economic and demographic growth. There was a high demand for housing, resulting in the construction of a large number of buildings that are now part of the existing building stock. Additionally, from Figure 2, it is possible to conclude that approximately 90% of primary dwellings were constructed before 2006.

The maintenance and repair needs of a building can also have substantial ramifications for energy consumption, resulting in an increase in energy usage. For example, if the building envelope (e.g., roof, walls, windows, and doors) is not properly maintained, it can lead to heat losses. This will cause heating and cooling systems to work harder to maintain comfortable indoor temperatures, resulting in an increase in energy usage.

Figure 3 represents the number of buildings that require maintenance or repairs per year of construction; the data were extracted from [31].

From Figure 3, it is possible to conclude that the buildings built between 1961 and 1980 are the ones that require more maintenance. According to Almeida et al. [33], the 20th century saw marked growth in the implementation of reinforced concrete construction in Portugal, which was principally seen between 1960 and 1980. This increase reflected an effort to augment the longevity and resistance to earthquakes of such structures. However, it was later found that the reinforcement bars used in these constructions had inadequate protections against moisture and saltwater exposure, resulting in corrosion and swelling that caused damage to their structure.

3.3. Building Regulations and Legislation in Portugal

The REH is a set of regulations and technical specifications that establishes the energy performance requirements for residential buildings in Portugal [19]. The REH was first published in 2006 and has subsequently undergone multiple updates to reflect advancements in renewable energy and building energy efficiency technologies.

The REH establishes minimum energy performance standards for a number of building envelope components, including walls, roofs, floors, and windows, as well as for heating, cooling, ventilation, and lighting systems. It is suitable for both new and old residential buildings. It also creates guidelines for determining a building’s energy performance and issuing energy performance certifications, both of which are necessary for all newly constructed buildings, as well as for any existing buildings that are sold or rented [19].

The Portuguese Regulation for Energy Performance of Buildings (RCCTE) [24] RCCTE specifies the standards for the thermal performance of buildings. This guidelines were initially established in 1990, and since then, multiple improvements have been made [24].

Weather plays a vital role in the energy performance of buildings, as it directly influences heating, cooling, and overall energy demands. In the Portuguese legislation, the country is classified into six climate zones: three for winter (I1, I2, and I3) and three for summer (V1, V2, and V3) [24]. Index 1 zones have a milder climate, whereas index 3 zones have a harsher climate [34].

The classification of winter climate zones is not based solely on HDD values; it is also influenced by the elevation of each municipality. While HDD captures the heating requirements over time, elevation plays a critical role in determining local temperature variations. Higher elevations tend to experience lower temperatures, which can significantly increase heating needs, even in regions where the HDD values alone might not reflect such extreme conditions.

To address this, the winter zone classification uses a weighted approach, combining both HDD values and the elevation of municipalities. This ensures a more accurate understanding of the heating demands in different areas. For instance, municipalities at higher elevations may fall into harsher winter zones due to colder temperatures that result from elevation, even if their HDD values suggest milder conditions. This classification acknowledges the dual influence of sustained cold temperatures and elevation, where milder zones (I1) are typically characterized by both lower HDD values and lower elevations, and harsher zones (I3) reflect both higher HDD values and greater elevations. The combination of these factors provides a more comprehensive reflection of heating demands across different regions, accounting for both long-term temperature patterns and the local impact of elevation.

The summer zones are determined on the basis of the average outside temperature for the cooling season. This classification helps in understanding the cooling requirements of buildings across different regions. In milder summer zones (V1), lower average temperatures reduce the need for cooling, while in harsher summer zones (V3), higher average temperatures indicate a greater need for cooling to maintain indoor thermal comfort.

3.4. Model Inputs

In this section, some of the required inputs for the energy model used in this study are presented and clarified. The simplifications and assumption made in the model are also presented throughout this section. This includes the simplifications and approximations made in the modeling process, as well as the uncertainties in the input data. Acknowledging these limitations is important for ensuring a transparent and thorough analysis of the results, helping to contextualize the model’s findings within the scope of its design and the data available.

3.4.1. Residential Building Stock

In developing the energy model for this study, certain considerations were made regarding the physical and thermal characteristics of the residential stock. Specifically, it was assumed that all dwellings have a standard wall height of 2.7 m, which, according to Appolloni and D’Alessandro [35], is the minimum wall requirement in Portugal, and the ratio of a dwelling’s length to its side is 1.

3.4.2. Average Floor Area

Based on census data, it is possible to determine the average floor area of the Portuguese residential stock in each municipality. The overall number of residences in each municipality, as well as the number of residential buildings classified according to their area, are both taken into consideration. The average floor area (m²)) for every municipality is determined using Equation (23).

$${\overline{A}}_{floor}={\displaystyle \frac{{\sum }_{i=1}^n{\overline{A}}_i·N_i}{N_{dwellings}}}$$

(23)

where ${\overline{A}}_i$ (m²)) represents the average floor area for each interval, $N_i$ represents the number of dwellings in each corresponding interval, and $N_{dwellings}$ is the total number of dwellings.

3.4.3. Number of Floors

The number of floors in residential buildings is also a crucial parameter for assessing the energy consumption in the Portuguese residential building stock. To this end, Equation (24) was developed to estimate the number of floors in each municipality.

$$N_{floors}=\sum _1^{x}x·N_{RB,xf}$$

(24)

where x represents the number of floors in a given building, and $N_{RB,xf}$ is the number of buildings with x floors.

Data from the Portuguese 2021 census on the number of residential buildings categorized by the number of floors were used. These data indicate the number of buildings in each municipality, ranging from those with just one floor to those with seven or more floors.

Finally, Equation (25) calculates the average number of dwellings per floor in each municipality.

$${\overline{N}}_{d/floor}={\displaystyle \frac{N_{dwellings}}{N_{floors}}}$$

(25)

where ${\overline{N}}_{d/floor}$ is the average number of dwellings per floor.

3.4.4. Overall Heat Transfer Coefficient

The overall heat transfer coefficient (U-value) is a key factor in evaluating the energy performance of residential buildings. It measures how quickly heat is lost through the building’s envelope, taking into account the thermal resistance of building components like walls, floors, roofs, and windows.

One significant factor that can influence the U-value is the period of construction of buildings. This is because building materials and construction practices have improved over time, largely due to advancements in regulations and laws. As a result, newer buildings have better building envelope performance and are better equipped to prevent heat losses. Conversely, older buildings have less thermal insulation or single-pane windows that allow heat to escape more easily.

Figure 4 presents the U-values considered for each envelope element based on the period of construction; the data were extracted from [36].

In this case study, 11 archetypes were used to represent the building stock by construction period, simplifying the modeling process but assuming uniform characteristics within each period. This introduces uncertainty, as it may not fully capture the diversity of construction techniques and materials across buildings.

3.4.5. Glazing Area Percentage

The glazing area percentage (GAP) refers to the proportion of a building’s exterior envelope that is made up of windows or glass surfaces. GAP is an important parameter to consider because it directly impacts a building’s thermal performance, affecting both heat loss and solar gains.

In older buildings, there is a tendency to have less glazing area due to architectural styles and design preferences, which often prioritize solid walls over windows. These solid walls are also more prevalent in historical buildings. As a consequence, older buildings GAP is smaller than that of newer buildings [37].

The following figure, Figure 5, presents the GAP values based on the period of construction considered in this study; the data were extracted from [37].

The variance in GAP across different periods of construction is not as pronounced as in some other building parameters. However, the trend of an increasing GAP over time is still noticeable, indicating a gradual shift toward incorporating more glazing in building designs, especially in post-2000 constructions.

3.4.6. Air Changes per Hour

The (ACH) is a measurement of how many times the air in a room is fully replaced with outdoor air in an hour. It is an essential metric for assessing indoor air quality and energy efficiency in buildings. The following figure, Figure 6, presents the evolution of the ACH based on the period of construction; the data were extracted from [21].

3.4.7. Window Orientation Coefficient

The window orientation coefficient ($X_i$) is a parameter that, unlike some of the other factors presented in the model formulation section, was not directly obtained from the Portuguese legislation. Therefore, it is important to explain how this coefficient was calculated in order to ensure clarity and accuracy in the model’s methodology.

To compute this parameter, it is first necessary to know the distribution of windows orientation for each climatic zone, which was obtained from [38] and is presented in

Table 1. Window orientation per climate zone. Adapted from [38].

Orientation	I1V1	I1V2	I1V3	I2V1	I2V2	I2V3	I3V1	I3V2	I3V3
W	0.14	0.12	0.21	0.15	0.26	0.12	0.08	0.06	0.19
SW	0.08	0.18	0.08	0.12	0.03	0.20	0.09	0.07	0.05
S	0.06	0.11	0.08	0.15	0.10	0.06	0.17	0.25	0.24
SE	0.22	0.12	0.09	0.07	0.03	0.12	0.16	0.18	0.00
E	0.14	0.18	0.21	0.12	0.25	0.12	0.17	0.05	0.28
NE	0.08	0.17	0.08	0.11	0.05	0.20	0.00	0.17	0.00
N	0.07	0.12	0.17	0.15	0.12	0.07	0.16	0.11	0.24
NW	0.21	0.00	0.08	0.13	0.16	0.11	0.17	0.11	0.11

.

Secondly, the orientation factor for different exposures is required. This factor takes into account the orientation of the facade of a building with respect to the cardinal points, and it determines the amount of solar radiation received by the building. The values for this parameter were sourced from the RCCTE [24].

Finally, the window orientation coefficient by climate zone is calculated using Equation (26).

$$X_i=\sum _{j=1}^n{D}_{ij}·F_j$$

(26)

where $X_i$ is the window orientation coefficient for climate zone i, $D_{ij}$ is the distribution of windows’ orientation for climate zone i and orientation j, $F_j$ is the orientation factor for orientation j, and n is the total number of orientations considered.

3.4.8. Solar Gains

Solar gains ($G_{solar}$) are the thermal energy a building receives from solar radiation, which affects its internal temperature. They can reduce heating needs in colder months but increase cooling demand in warmer periods, significantly influencing a building’s overall energy performance. However, they are not directly obtainable from the legislation, requiring calculations based on solar irradiation data and building orientation to accurately estimate their contribution to the building’s energy balance.

The equation used to calculate solar gains ($G_{solar}$) is as follows:

$$G_{solar}=\sum _{i=1}^n{I}_i·D_i$$

(27)

where $G_{solar}$ (kWh/m²), the solar gains, represent the overall thermal energy received in the building from solar energy, and $I_i$ (kWh/m²) indicates the amount of solar energy incident on a surface for each direction (e.g., north and south) during the specified cooling period (June to September).

3.4.9. Emissions and Costs

In order to accurately calculate the CO₂eq emissions of the different sectors, it was necessary to select the CO₂eq emission factors associated with each type of energy vector used.

Table 2. CO₂eq emissions factors and energy costs.

Emission Factors (kgCO2eq/kWh)	Electricity	Natural Gas	LPG	Biomass	Heating Oil	Solar
Energy Conversion 1	0.220 [39]	0.041 [40]	0.028 [40]	-	0.047 [40]	-
Use Phase 2	-	0.203 [18]	0.227 [18]	0 [18]	0.267 [18]	-
Total	0.220	0.244	0.255	0	0.314	0
Cost (EUR/kWh)	0.218 [41]	0.144 [42]	0.213 [42]	0.120 [42]	0.140 [43]	0.03

¹ Emissions resulting from the various stages of the energy supply chain, including extraction, separation, transport, distribution, and refining if it is applicable. ² Emissions resulting from the combustion process itself, considering the chemical reactions involved and the release of pollutants into the atmosphere.

represents the emissions factors utilized in this work, as well as the price of each energy vector.

Biomass was considered to have no direct emissions, as it is often classified as carbon-neutral due to the assumption that the CO₂ released during combustion is offset by the carbon absorbed during the growth of biomass [44].

The CO₂eq emissions factors also take into account the emissions resulting from extracting, transporting, and using the energies considered in the model. These additional non-combustion factors are essential to ensure that a complete picture of energy-related CO₂eq emissions is taken into account when assessing its true environmental impact.

To obtain the price of the kWh produced via the solar thermal panels, Equation (28) was utilized.

$${\mathrm{LCOE}}_{\mathrm{solar}}={\displaystyle \frac{C_{\mathrm{solar}}}{E_{\mathrm{total}}}}$$

(28)

where $C_{\mathrm{solar}}$ (EUR) is the capital cost of the solar thermal panel, and $E_{\mathrm{total}}$ (kWh) is the total energy output produced via the panel over its operational lifetime.

3.5. Equipment and Technology Penetration

In this section, the average penetration of various equipment and technologies across different end uses in residential dwellings in Portugal is presented. The specific penetration for each individual municipality is not shown, as such information would be too extensive, but the national average is provided to suggest an overall idea of the distribution.

3.5.1. Space Heating

The 2021 Census categorizes space heating technologies into four distinct categories: central heating, non-central heating (fixed and movable equipment), fireplaces, and no heating [31]. In the case study, among the total dwellings with access to space heating, movable equipment had the highest penetration at 40.7%, followed by fireplaces at 31.1%, central heating at 20.0%, and fixed equipment at 8.2%.

Additionally between the 2010 and 2021 census, an indicator that provided information on the distribution of space heating technologies by their corresponding energy source was eliminated. As a result, the 2021 census only included data on the distribution of space heating technologies. To obtain the corresponding distribution of space heating technologies by energy source for 2021, a projection was made using the data from the 2010 census [18,45].

3.5.2. Space Cooling

The space cooling technologies considered in the model include air conditioners, fans, and heat pumps. These technologies are commonly used in maintaining comfortable indoor temperatures during hot weather conditions. While heat pumps are a more energy-efficient option, air conditioners and fans are widely available, with fans being the most inexpensive option. It is important to note that, while fans can provide some level of comfort to individuals, they do not have the capability of decreasing the ambient temperature to the desired set point temperature.

The availability of space cooling technologies in residential buildings in Portugal was determined by analyzing the data gathered from the 2021 census, as well as the Energy Consumption Survey in the Domestic Sector 2020 [18,31].

Among the dwellings equipped with cooling technologies, which make up approximately 32.7% of the total, air conditioners are present in around 16.7% of the dwellings with cooling technologies in Portugal. Fans are more widely used, being found in about 58.8% of these dwellings, while heat pumps are available in approximately 45.4% of the dwellings with cooling systems.

3.5.3. Water Heating

The technologies considered for water heating in the model were gas water heaters, electric water heaters, boilers, and solar thermal systems. Each technology was assigned a distribution value between 0 and 1, with the sum of all values in a municipality equal to 1. This approach is based on the assumption that each dwelling has only one piece of water heating equipment, unlike the space cooling and space heating categories, where multiple systems can coexist in a single dwelling or a dwelling may not include any such system at all. This distinction is made because water heating is typically served through a dedicated system, whereas cooling and heating needs may vary, with some dwellings using a combination of devices or none at all, depending on the layout and usage patterns.

In the case study, gas water heaters’ penetration rate was of 57.0%, electric water heaters’ was 14.0%, gas boilers’ was 16.0%, and solar thermal was present in 13% of the dwellings; the data were obtained from [18,46]. According to the data, gas water heaters remain the most widely used technology for water heating in Portugal, while solar thermal systems are the least commonly used.

3.5.4. Lighting

In assessing the lighting technologies present in Portuguese dwellings, several factors were considered. Notably, four different types of light bulbs were taken into account, including incandescent, halogen, fluorescent, and LED bulbs [18]. Notably, in dwellings that had these bulbs present, there were often multiple options available.

The Energy Consumption Survey in the Domestic Sector 2020 [18] indicates that LED lighting is the most widely used technology in the case study, being present in 80.10% of dwellings, with an average of 12.60 bulbs per household. Fluorescent lighting follows, with a presence in 56.60% of dwellings, although with a significantly lower average of 3.20 bulbs. Halogen bulbs are installed in 23.50% of dwellings, with an average of 5.80 bulbs per household, while incandescent bulbs are still found in 35.10% of dwellings, with an average of 5.70 bulbs.

3.5.5. Cooking

In this study, four major cooking technologies were considered: cookers with ovens, cookers without ovens, independent ovens, and cooking robots. Among the technologies assessed, cookers with ovens had the highest penetration at 52.00%, followed closely by cookers without ovens and independent ovens, both at 51.60%. Cooking robots, with a penetration rate of 13.90%, were the least common technology among the options considered in the case study. This distribution highlights the prominence of traditional cooking technologies, while newer technologies, such as cooking robots, are still less widespread. These results were extracted from the Energy Consumption Survey in the Domestic Sector 2020 [18].

3.5.6. Household Appliances

In this study, a diverse number of household appliances were considered, namely fridges, freezers, dishwashers, washing machines, and clothes dryers, among others. These are the appliances most common in a residential setting, and they have a greater impact on the total energy consumption of a dwelling.

It is important to also take into account the energy efficiency of electronic appliances, as it can vary significantly, impacting energy consumption. As the 2020 distribution of efficiency for electronic devices was not available, it was necessary to use the 2010 distribution presented in [45] in this study. It is important to note, however, that electronic devices are not frequently replaced, and thus, the 2010 distribution may still be relevant for many dwellings. Nevertheless, updated data must be made available to improve the accuracy of energy consumption models. Figure 7 presents the energy efficiency share for the case study; the data were extracted from [18,45].

The analysis of household appliances in Portugal revealed that high-efficiency appliances make up a mere 2% of the total. The most common efficiency rating is D, accounting for 43% of appliances, followed by E class with 19%, Class F with 14%, and lower efficiency ratings representing 12% of appliances. This distribution does not account for dryers since they possess their own separate energy efficiency label.

It is also important to note that some of the data used in this study, particularly regarding appliance energy efficiency, originate from the 2010 distribution, prior to the introduction of the new energy efficiency classifications. To ensure consistency with current standards, the data for appliances were updated to match the latest energy classification brochure. As a result, the first four appliances in Figure 7 (fridges, freezers, dishwashers, and washing machines) do not show an “A” rating, reflecting the revised classifications.

The model additionally included electronics related to entertainment, such as television sets and boxes. It was important to include these devices in the model due to their high penetration rate (99.2%) and high usage across dwellings nationally.

Unlike certain end uses for which detailed data on equipment penetration are available at the municipal level, the information for cooking technologies, lighting, and household appliances is only accessible at the NUTS I level.

This presents a limitation because different regions may vary in their cooking habits, access to technologies, and income levels, leading to differences in the actual use of these technologies. However, due to the lack of disaggregated data, assuming a uniform penetration rate for all municipalities was a necessary simplification. This approach allowed for the inclusion of these technologies in the model but may have obscured local differences, potentially leading to less precise estimates at the municipal scale.

4. Results

This section presents the results and findings of the regional residential energy model, providing insights into the energy consumption patterns and dynamics within the residential sector. Firstly, the validation process conducted in this work will be presented, ensuring the accuracy and reliability of the regional residential energy model. Subsequently, an analysis of residential energy consumption, its associated emissions, and costs will be presented and discussed.

4.1. Validation

A comparison and a validation of the regional residential energy model’s results were performed for electricity and natural gas.

Electricity and natural gas consumption data were used to ensure the model’s accuracy and reliability using reference data from the DGEG website [42] for the year 2021. Validating the model using 2021 data is particularly important because the census data from 2021 were used extensively in the model, making it representative of the country’s state during that year.

Figure 8 and Figure 9 display the normal distributions of errors associated with the residential electricity and natural gas consumption estimated via the model, respectively.

Both models demonstrate strong performance, with a slight underestimation of electricity consumption (mean error of −0.36%) and a slight overestimation of natural gas consumption (mean error of 0.70%).

The standard deviation of errors for the electricity model is 17.75%, while the natural gas model involves a slightly higher standard deviation of 17.85%. This indicates that, although the models perform effectively on average, there is variability in their predictions. The electricity model’s error typically ranges between −18.11% and +17.39%, and the natural gas model’s error spans from approximately −17.15% to +18.55%. These ranges represent the spread of errors in the model’s predictions, meaning that, while the average error is close to zero, individual predictions can vary considerably from the actual values. The variability is likely due to the complex and diverse factors that influence energy consumption across municipalities, such as differing household behaviors, appliance usage patterns, and the unpredictability of certain local factors that the model may not fully capture.

Although these models are not without limitations, they offer a robust representation of overall consumption trends. The relatively low mean errors suggest that the models are largely unbiased at a national scale, although some deviations may occur for individual municipalities. Given the complexities involved, the performance of both models is well within acceptable ranges for this type of analysis, as supported by other UBEM studies where acceptable error ranges for energy consumption predictions typically fall between 1% and 19%, with this range reflecting the absolute error calculated from comparisons of predicted energy use against real-world utility data [47].

Additionally, Chen et al. [14] noted that deviations at smaller scales, such as individual municipalities, are acceptable as long as UBEMs remain effective in capturing broader consumption trends. Their study emphasized that these deviations are considered a necessary trade-off, given the need to balance model complexity with computational efficiency.

In this case study, some data, such as the equipment penetration of cooking appliances, were only available at the multi-municipal level, and they were assumed to be homogeneous across all the municipalities at this level. Should more granular data become accessible in the future, they would enhance the precision of localized results and further strengthen the model ability to support decarbonization strategies tailored to municipal contexts.

4.2. As-Is Scenario

Characterizing the current state of the case study is important for understanding its energy consumption and emissions patterns. This baseline not only offers insight into how energy is being used and where emissions are originating from but also serves as a reference point for evaluating the impact of future interventions.

Figure 10 and Figure 11 depict a characterization of energy consumption patterns across the various end uses in the residential dwellings within the case study, offering insights into both the proportional distribution of energy consumption and the energy use per dwelling.

The figures illustrate that, in this specific case study, space heating and cooking represent the largest portions of energy consumption. The high standard deviation observed in space heating (Figure 11) is largely a consequence of each municipality’s specific weather conditions. More specifically, municipalities in the northern regions of the country exhibit the highest energy consumption for space heating, which is closely associated with higher HDD values.

Space heating consumption is also influenced by other factors, such as building age, envelope thermal insulation quality, and heating system efficiency, and this combination of factors, with HDD playing a significant role, directly contributes to the larger standard deviation observed in space heating consumption throughout the case study.

For cooking, household appliances, and lighting, the low standard deviations can be explained by the lack of detailed municipal-level data for these end uses. In the absence of such data, assumptions had to be made at the national level. For instance, lighting requirements, including hours of usage and the types of appliances used, were standardized across all municipalities. These generalized inputs contribute to less variability in the model’s outputs, as they assume similar consumption behaviors across regions. This results in relatively low standard deviation values, as there is little differentiation between municipalities in terms of these energy uses.

In the case of water heating, while detailed information on equipment penetration is available at the municipal level, the actual usage patterns—such as the amount of water consumed and the temperature settings—were assumed to be constant across all households. This decision, while necessary due to data limitations, contributes to the medium level of standard deviation observed.

For space cooling, despite the relatively low average energy consumption, there is a higher standard deviation compared to other end uses. This can be attributed to the climatic variability across the case study area. Municipalities in the interior of the country, which experience higher outside temperatures during the cooling season, tend to have greater energy consumption for space cooling, while coastal or cooler regions have lower space cooling needs. This geographic disparity introduces greater variability in the energy consumption for space cooling, even though the overall demand remains low compared to other end uses.

Finally, understanding the total emissions per end use in the case study is crucial for identifying which energy-consuming activities contribute the most to overall GHG emissions. This breakdown allows for a more targeted approach when developing decarbonization strategies, as it highlights the specific areas where emission reductions can be most impactful.

Figure 12 presents the energy source distribution by end use, and Figure 13 presents the total emissions per end use across the case study, providing a clearer picture of how energy consumption translates into CO₂eq emissions.

Although space heating is one of the most energy-intensive end uses, a significant portion of its consumption comes from biomass, which in this study is considered to produce no emissions. This assumption explains why space heating, despite its high energy consumption, contributes less to the total emissions when compared to other end uses.

Unlike other energy-intensive end uses like household appliances, water heating, and cooking, which exhibit consistently high consumption across all municipalities, space heating exhibits significant regional variation. As previously discussed, this is largely due to differences in HDD, where northern regions with higher HDD values experience much greater demand for space heating compared to their lower HDD counterpart. This regional disparity means that space heating plays a major role in energy consumption and emissions in colder areas while contributing relatively little in warmer regions.

Space cooling and lighting are the least-emitting sectors, which aligns with the fact that these end uses rely solely on electricity and represent a smaller share of overall energy use, as shown in Figure 10 and Figure 12.

4.3. Decarbonization Scenarios

In this section, various decarbonization measures are explored with a focus on reducing emissions across different building types and energy systems. These measures are evaluated to highlight potential energy savings, environmental benefits, and economic considerations.

The next step is particularly important, as it transforms the model from only estimating emissions and energy consumption into a design tool. By providing detailed insights into the cost-effectiveness and environmental impact of different measures, the model enables stakeholders make informed decisions about solutions that balance sustainability and investment returns.

Although the number of potential measures to test is extensive, a set of scenarios has been carefully selected based on their appropriateness for the case study. These measures represent a diverse mix of passive design improvements, active technological upgrades, and behavioral changes tailored to the specific needs and context of the study area.

Table 3. Measures under study overview.

Type of Measure	Measure ID	Measure Description	Scope
Passive measures	E1	Window renovations	I3 Municipalities
	E2	Thermal insulation renovations	I3 Municipalities
Active measures	W1	Solar thermal	High-irradiation municipalities 1
	W2	Electrification of water heating	Nationally
	W3	Electrification of water heating	High population density *
	H1	Electrification of space heating	Nationally
	H2	Electrification of space heating	Highest penetration of fossil fuels *
	L1	100% LED lighting	Nationally
	A1	More efficient appliances	Nationally
	A2	More efficient appliances	High population density *
	A3	More efficient appliances	High income *
Behavioral measures	B1	Shower water conservation	Nationally
	B2	Decreased TV consumption	Nationally

¹ Performed in the following NUTS III municipalities: Algarve, Alentejo Central, Baixo Alentejo, Alto Alentejo, and Alentejo Litoral. * Performed in the top 50 municipalities of this scope.

provides a comprehensive overview of the measures and design scenarios evaluated in this study.

This study tested three distinct categories of scenarios: passive, active, and behavioral. Passive measures focus on improving the energy efficiency of buildings by enhancing the design and characteristics of the building envelope. The building envelope refers to the physical barrier between the conditioned indoor space and the external environment, including elements like walls, roofs, windows, and floors. These measures are generally non-mechanical and focus on reducing the building’s energy demand by leveraging design principles that naturally maintain thermal comfort. Active measures, in contrast to passive ones, involve the use of mechanical or electrical systems to manage energy demand and improve the energy efficiency of a building. These systems typically include HVAC, and lighting, among others. Active measures often focus on upgrading equipment to more energy-efficient alternatives or integrating renewable energy sources into a building [48].

In addition to passive and active measures, behavioral measures focus on influencing the behavior of building occupants to promote energy-efficient practices. Unlike passive and active measures, which are related to physical upgrades or equipment, behavioral measures aim to reduce energy consumption by changing how people use energy.

Despite biomass being recognized as a renewable energy source, it was not included in the decarbonization measures. This decision was based on concerns that incentivizing its use could promote deforestation, as increased demand for biomass may lead to unsustainable harvesting of forests, ultimately undermining the environmental goals of the study.

For each measure, a base investment cost was established, which reflects the specific characteristics of the case study that are presented in

Table 4. Base investment costs.

Measure ID	Intervention	Investment
E1	EPS 150 (80 mm thickness)	45 EUR/m2 [49]
E2	Window framing and double glass	380 EUR/m2 [49,50]
W1	Solar thermal panel	1500 EUR/unit [49]
W2, W3	Heat pump water heater	2200 EUR/unit [49]
H1, H2	Heat pump space heating	2800 EUR/unit [49]
L1	LED bulbs	4 EUR/W [50]
	Fridge	950 EUR/unit [49]
A1, A2, A3	Washer	300 EUR/unit [49]
	Dryer	450 EUR/unit [49]
B1, B2	Education and awareness programs	3.30 EUR/resident [51]

. While tailored to the conditions of this case study, the costs can be adjusted and adapted for application in other geographical or socioeconomic contexts.

5. Discussion

An abatement cost curve was utilized to support the design of cost-effective solutions by comparing decarbonization measures. The abatement cost curve illustrates the relationship between the cost per tonne of CO₂eq avoided (vertical axis) and the annual emission reductions achieved via each measure (horizontal axis). This enables a clear comparison of different strategies based on their financial cost and their ability to reduce emissions on a yearly basis.

By applying this method, we can identify measures that are both cost-effective and capable of contributing significantly to overall emission reductions.

When interpreting an abatement cost curve, measures can appear on either the negative or positive side of the cost axis, which has different implications for their cost-effectiveness. Measures on the negative axis (below zero on the cost axis) represent negative abatement costs. These measures not only reduce emissions but also save money over time. The savings they generate, typically through lower energy use or increased efficiency, more than offset their initial costs. Measures on the positive axis represent positive abatement costs, meaning they require a net investment for each tonne of CO₂eq reduced.

While the abatement cost curve is a valuable tool for assessing cost-effectiveness, it is important to recognize that it does not capture additional benefits that certain measures may provide. For instance, measures such as building insulation or renewable energy installations can improve thermal comfort, enhance indoor air quality, or offer energy security through local production. Additionally, measures like investments in renewable energy infrastructure might yield future financial gains, especially if energy prices fluctuate or carbon pricing becomes more stringent. Thus, while a measure with a positive abatement cost may seem less attractive from a purely financial perspective, its broader societal and environmental contributions should not be disregarded. These extra benefits can play a crucial role in long-term decarbonization strategies, making such measures essential in achieving comprehensive climate goals.

An abatement cost curve serves as a design tool to assist decision-makers in interpreting cost-effectiveness alongside other objectives, such as maximizing emission reductions or addressing societal benefits. Factors such as policy goals, long-term sustainability, and the specific context of the region or sector should all play a role in shaping decisions. Ultimately, the curve provides valuable insight into the relative costs of different strategies, but the final choice depends on the balance between economic feasibility, environmental impacts, and other strategic priorities that are unique to each decision-making scenario.

Figure 14 presents the abatement cost curve for the measures tested in the case study on a national level.

The passive measures, E1 and E2, which involve the installation of thermal insulation and the upgrading of windows in I3 winter-zone municipalities, both show a positive abatement cost on the abatement cost curve. Moreover, they demonstrate relatively marginal emission reductions when compared to some of the other measures. This can largely be attributed to the nature of passive measures, which typically involve high upfront investment costs for materials, labor, and installation, particularly for building-envelope upgrades.

Additionally, the effectiveness of these passive measures in reducing energy consumption for space heating and cooling can be limited in certain cases. For example, while thermal insulation and upgraded windows reduce heat transfer, their impact on actual energy consumption depends on the existing efficiency of the heating and cooling systems. If the heating and cooling systems in place are already efficient, the marginal energy savings from insulation upgrades can be lower, thus resulting in a higher cost per tonne of CO₂eq reduced. This is especially true in cases where the baseline energy consumption for space heating is not as high, making the reduction in consumption smaller relative to the investment.

It is important to note that space heating is largely dependent on biomass, which accounts for approximately 60% of the energy used. Biomass is often regarded as carbon-neutral, meaning measures aimed at reducing energy consumption for space heating, such as improved insulation or window upgrades, have a limited direct impact on emission reductions in this scenario. However, the remaining 40% of energy used for space heating comes from electricity, natural gas, LPG, and heating oil, which do contribute to emissions. As results passive measures, while they are beneficial in terms of energy efficiency, they have a reduced overall impact on emissions in cases where biomass dominates the heating mix.

However, this does not diminish the importance of such measures in improving thermal comfort and ensuring energy efficiency, even in situations where the direct emissions impact may be less significant. To assess whether these benefits are truly justified or worthwhile, an additional and more detailed analysis would be necessary.

All the active measures in this study demonstrated a negative abatement cost, indicating that they result in net economic gains over their lifetimes.

Among the active measures, W2, which involves the installation of heat pumps to replace fossil fuel-based water heating systems on a national level, is the most impactful in terms of direct emission reductions. Heat pumps are an energy-efficient technology that significantly reduces dependence on fossil fuels, offering substantial reductions in CO₂eq emissions; however, W2 presents a higher abatement cost compared to other measures in this study, primarily due to the high upfront costs associated with purchasing and installing heat pump systems. Despite these costs, the measure still offers a positive return on investment over time, making it financially viable in the long term.

Although W2 offers substantial emission reduction potential, implementing it on a national scale presents significant challenges. Replacing all fossil fuel-based water heating systems with heat pumps across the country would require a large financial investment and logistical coordination, making it unfeasible in the short term. However, this measure provides a valuable perspective on the long-term environmental gains that could be achieved.

Measure L2, which involves converting the entire residential sector to LED light fixtures, stands out as the measure with the lowest abatement cost and offers the best economic returns, along with considerable emission reductions. In the case study, a large portion of residential lighting still relies on non-LED lights, leaving substantial room for improvement. The low abatement cost associated with this measure can be attributed to the affordability of LED technology and its high efficiency, making it an attractive option for decarbonization efforts in this region.

The higher efficiency of LED lights, coupled with their relatively low cost, underscores their potential as a key technology in the future decarbonization plans for this case study. Although replacing all lighting fixtures on a national scale at once may not be feasible, a gradual approach could be taken. This could include phasing out non-LED fixtures from the market and incentivizing the adoption of LEDs. Given the already low price and long lifespan of LEDs, this transition would likely be easy to implement as older lighting systems naturally phase out, ensuring a smooth and cost-effective shift toward energy-efficient lighting.

Finally, it is important to closely examine the behavioral measures. In these measures, the investment required was based on the cost per resident, which represents the education needed to shift behaviors. These costs include expenses for education campaigns aimed at informing the population. Both B1 (water conservation) and B2 (reduced TV watch time) achieved a negative abatement cost, indicating that they result in net economic gains over time; however, the abatement cost for B1 is lower than that for B2. This difference arises because water conservation leads to more significant energy savings by reducing the amount of hot water needed, which lowers the energy demand for heating. In contrast, reducing TV watch time has a smaller impact since the energy used by TVs is much lower than the energy required for water heating.

These measures demonstrate that behavioral changes can have a substantial impact on the emissions of the residential sector. Encouraging shifts in daily activities, such as conserving water or reducing unnecessary energy use, can effectively lower household emissions. However, while promoting behavioral changes has clear benefits, achieving a nationwide shift in behavior is challenging, and such changes require persistent and well-designed education campaigns to ensure long-term adoption.

While the proposed decarbonization measures offer significant insights, certain practical aspects, such as physical space requirements for the installation of technologies, were not considered in this analysis. These limitations, although important for real-world implementation, fall outside the scope of this study.

In addition to analyzing the cost-effectiveness of the measures, the discounted payback period was also evaluated for each intervention. The discounted payback period refers to the amount of time it takes for the cumulative discounted savings from a measure to equal its initial investment. This period accounts for the time value of money, meaning that future savings are discounted to reflect their present value. By doing so, it provides a more accurate representation of how long it takes for the benefits of a measure to outweigh its costs.

Table 5. Discounted payback period.

Measure ID	Discounted Payback Period
E1	41 years
E2	47 years
W1	10 years
W2	15 years
W3	15 years
H1	11 years
H2	10 years
L1	2 years
A1	16 years
A2	15 years
A3	13 years
B1	1 year
B2	1 year

presents the discounted payback period for the decarbonization measures under study.

The analysis in Table 5 reveals that the highest discounted payback periods were found in the passive measures, particularly in Measure E2. Notably, the discounted payback period for E2 is significantly longer than the expected lifetime of windows (≈25 years), indicating that the financial returns from energy savings will not cover the initial investment within the useful lifespan of the measure. As discussed earlier, passive measures are generally less effective at reducing energy consumption, which means that their high upfront costs are harder to recover through savings.

Most of the active measures, with the exception of L1 (LED lighting), demonstrated medium to high discounted payback periods. Despite this, these payback periods are still within the expected lifespan of the equipment. With proper maintenance, systems such as solar thermal panels and heat pumps can last significantly longer than their discounted payback periods, ensuring that the initial investment is recouped before the equipment reaches the end of its useful life. This makes these measures both financially viable and effective in the long run, as they continue to provide energy savings and emission reductions even after the payback period has passed.

In contrast, L1 (LED lighting) achieved a short discounted payback period, primarily due to the low cost of LED technology. LEDs are relatively inexpensive to install and operate, which allows for faster financial returns. Moreover, LED lights are known for their long lifespans, often lasting up to 50,000 h of use, which translates to several years of continuous service. This combination of low cost, high efficiency, and long life makes LED lighting an essential component of any decarbonization strategy, offering both immediate financial savings and long-term emission reduction.

Finally, behavioral measures showed the lowest discounted payback periods, largely due to the immediate energy savings that result from changes in daily habits. These measures require minimal upfront investment, as the primary costs are associated with education and awareness campaigns, while the energy savings are realized quickly. Actions such as reducing unnecessary energy consumption, like optimizing appliance use or conserving water, lead to significant reductions in household energy demand, translating into rapid monetary savings.

Because behavioral shifts can be implemented without costly infrastructure changes, the financial return is seen almost immediately. This makes them highly cost-effective in the short term while also contributing to emission reductions. These measures underscore the importance of engaging residents in decarbonization efforts, showing that even small adjustments in energy use habits can have a notable impact on overall energy consumption and emissions. Behavioral measures, therefore, play a vital role in achieving immediate gains while laying the foundation for broader decarbonization strategies.

The final analysis aimed to assess whether the abatement cost of a specific measure remained consistent across different municipalities or regions.

At the municipal level, the measure evaluated was W2, which has the highest abatement potential, and it involved the replacement of gas water heaters with heat pumps. The municipal-level abatement cost and emissions savings for measure W2 are presented in Figure 15.

As shown in the previous figure, Figure 14, the national-level abatement cost for measure W2 was found to be slightly negative. Now, at the municipal level, Figure 15 reveals that the abatement cost is not uniform across municipalities, showing significant variability. In some cases, the negative abatement cost observed at the national level shifts to a positive abatement cost locally, indicating that the measure does not perform as efficiently in all municipalities.

A potential factor influencing the variability in abatement costs could be the number of residents per household. Larger households typically use more hot water, increasing energy consumption; as a result, upgrading to more efficient systems, such as heat pumps, can lead to substantial cost savings and emission reductions. On the other hand, smaller households may have lower energy demands, making the transition less impactful and, therefore, less cost-effective.

For the regions NUTS II, Norte and Azores abatement cost curves were created to assess the decarbonization measures that are applicable to each region from the set analyzed, and these curves are presented in Figure 16 and Figure 17.

As depicted in Figure 16 and Figure 17, the abatement cost curves for NUTS II, Norte and Azores, illustrate notable differences in the performance of the decarbonization measures across these two regions. The most striking contrast is with measure H1, which demonstrates a negative abatement cost in Norte, indicating that the investment is recovered within the equipment’s lifetime due to the region’s harsher winter climate. In contrast, the same measure in the Azores has a positive abatement cost, suggesting that the milder climate makes this measure less economically viable. Furthermore, the abatement potential of H1 also differs substantially between the two regions, highlighting the influence of local conditions on the effectiveness of decarbonization measures.

Given this variability, it is important to explore how different municipal-level factors drive abatement costs and measure performance. The variability in the penetration rates of technologies, building characteristics, and other local conditions shows that a uniform approach may not be effective across all municipalities; therefore, future work should focus on conducting sensitivity analyses to understand how different parameters and variables affect the abatement cost of decarbonization measures.

Sensitivity analyses are valuable because they give decision-makers a clearer idea of how a measure’s abatement cost might change under different conditions. By understanding local factors like building stock, population density, and the use of fossil fuel-based equipment’s, municipalities can prioritize which measures to focus on without running extensive simulations for each one. Identifying the most influential variables helps authorities prioritize measures, simplifying decision-making and optimizing resource use. This approach allows municipalities to adopt more effective, scalable strategies and make smart, cost-efficient decisions that fit their specific local needs.

6. Conclusions

A computationally efficient regional residential energy model was developed and validated with a full-scale demonstration in Portugal. The model offers insights into both economic and environmental impacts, helping design decarbonization strategies by analyzing abatement costs, emission reduction potential, and discounted payback periods.

While the model offers distinct advantages, such as computational efficiency, adaptability, and the inclusion of a cost-effectiveness analysis, it also faces certain limitations. These limitations include the necessity of simplifications, assumptions, and the use of archetypes. Despite these limitations, the model demonstrated good performance at the national level, achieving a mean electricity and natural gas consumption estimation below 1%.

The novelty of this model lies in enabling a nation-wide design of cost-effective building solutions with regional disaggregation, as a consequence of its computational efficiency. This is advantageous when compared to traditional UBEMs, since they lack the computational efficiency necessary for national-scale deployment without the support of high-performance infrastructure.

The results highlight that active measures like LED lighting and heat pumps offer quicker payback periods and larger emission reductions compared to passive measures, such as insulation upgrades, which require higher upfront investments and result in lower energy savings. Specifically, LED lighting achieves a discounted payback period of 2 years, while heat pumps range from 10 to 15 years, depending on the measure. In contrast, thermal insulation has a much longer payback period of 41 years. Additionally, the role of behavioral changes demonstrates that even small adjustments in daily energy use can significantly impact energy savings and emission reductions, achieving annual savings of 147 tonnes of CO₂eq and 653 MWh.

The insights provided via this model can contribute meaningfully to a coordinated national decarbonization strategy. By identifying the most effective interventions across diverse municipalities, the model can guide policymakers in designing efficient, regionally tailored approaches to emissions reduction.