Energy and daylighting trade-offs in residential window design: multi-objective optimization for hot-arid regions Authors: Mahdi Ramezani, Somayyeh Taheri, Mohammad Sharajabian Gorgabi, Arash nourbakhsh sadabad Published: 2025 ABSTRACT The relationship between the geometrical parameters of windows and the energy and daylight performance of buildings is inherently complex and requires systematic investigation. This study addresses this challenge by optimizing the geometry of south-facing windows in residential apartments in Yazd, a city with a hot-arid climate. The novelty of this work lies in its holistic integration of a comprehensive set of window parameters, including width, height, sill height, horizontal and vertical position, and subdivision patterns, across eight typical residential room configurations derived from field surveys. Unlike previous studies that focused on isolated parameters or simplified spaces, this research systematically explores the combined influence of window geometry and room dimensions through a generative design approach. The methodology involves parametric modeling in Grasshopper using Python scripting, simulation of energy and daylight performance with Ladybug Tools, and multi-objective optimization through the NSGA-II. Three objectives were considered: minimizing cooling demand, minimizing heating demand, and maximizing average Useful Daylight Illuminance (UDI). Statistical analyses, including correlation and regression, were applied to identify the most influential parameters, and simulation results were validated against actual energy consumption data from residential units in Yazd. The results indicate that window to wall ratio (WWR), window area, and window height have the greatest influence on energy and daylight performance. The optimal configuration is a vertically elongated south-facing window with a WWR of about 20%, a height greater than 2 m, a sill height below 0.5 m, and horizontal centering on the façade. Compared to baseline cases, this configuration achieves a 3.9–5.2% reduction in total energy consumption, a 19.7–23.2% reduction in cooling demand, a 2.3–2.4% reduction in heating demand, and an 8–12% improvement in average UDI, while maintaining acceptable glare levels. These findings confirm that optimized window geometry can simultaneously enhance daylight quality and reduce energy demand, offering evidence-based guidelines for sustainable. Introduction Windows have a significant impact on the energy performance and daylighting of residential apartment facades in the city of Yazd. Considering the global energy crisis, the optimization of the geometrical parameters of windows appears to be essential. In general, approximately 40% of global energy consumption is associated with maintaining thermal comfort in the building sector1. Studies indicate that energy consumption in this sector is increasing at an accelerating rate, and it is projected that by 2040, this figure will rise by more than 50%2. Despite the critical importance of energy in the building sector, architectural design and construction practices in Iran are commonly carried out without systematic rules or evaluations, relying solely on the architect’s personal experience. Meanwhile, the physical attributes of buildings are considered among the most significant factors influencing building energy performance, and the correlation between building morphology and energy efficiency has been extensively studied3. Within this context, the geometrical parameters of windows in the building facade are identified as one of the most influential physical components impacting energy performance4. In recent years, numerous studies have focused on the optimization of window parameters based on energy and daylighting performance1,2,5,6. Nevertheless, the importance of this issue demands that it be systematically incorporated into energy performance standards and regulations in various countries. In this regard, there is a significant absence of regulations concerning window specifications in the national building codes of Iran. Accordingly, the aim of this research is to optimize the geometrical parameters of windows on all four geographic orientations of a residential apartment facade in the city of Yazd, based on the building’s energy and daylighting performance. The simultaneous consideration of energy and daylighting stems from the inherent trade-off between these two performance objectives6,7. Given the complexity of window design and the evaluation of energy and daylighting performance, the use of parametric computational systems appears essential. So, no prior work has combined specific set of window geometrical parameters with a multi-objective optimization across a diverse range of room dimensions in this climatic context. In this study, a parametric algorithm is employed within a Generative Design framework to enhance building performance8,9. This approach allows for the prediction and optimization of window geometrical parameters including window width, window height, sill height, window position, and horizontal and vertical divisions within the residential facade design. The generative algorithm is implemented in the Grasshopper environment using Python programming language. Subsequently, the building’s energy and daylight performance are simulated using Ladybug Tools. Finally, the multi-objective optimization process is carried out via the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) using the Wallacei plugin. In general, a considerable number of studies have addressed the impact of windows on building energy performance, and the volume of such research continues to grow. Most studies investigate a wide range of window related parameters. Among them, some focus specifically on the type of windows and glazing. For instance, the study by Vukadinović et al.10 examined the impact of different façade wall constructions, glazing types, and window-to-wall ratios (WWR) on the energy performance of a detached residential building, revealing the significant influence of window characteristics on cooling and heating demands. Similarly, Zhang et al.11 analyzed various types of windows for a building located in Belper, UK, and found that the number of glazing layers, U-Value and SHGC have a substantial impact on energy consumption. The findings suggest that increasing window size is beneficial only when the solar heat gain outweighs thermal losses. Xing et al.12 also studied window elements in the façade of an office building to propose strategies for reducing energy losses and environmental impacts. The results highlight the considerable effects of WWR, window orientation, and glazing type on energy use. Some studies focus specifically on smart technologies. Cannavale et al.4 investigated electrochromic window systems in residential buildings for optimizing energy efficiency and daylighting. Mohammad and Ghosh13, evaluated aerogel technologies as an effective strategy for improving the energy performance of windows. Their study included eleven types of granular and monolithic aerogel windows in the UK climate, demonstrating that variations in WWR, SHGC and U-Value can lead to a 33% increase in window thermal transmittance and a 15.5% reduction in heating demand. Araújo et al.14 examined the influence of different switching temperature ranges and visible transmittance values of thermochromic coatings on window glazing in office spaces across various climates. Krarti15 analyzed control strategies for smart glazing windows in office environments and confirmed the impact of window size and orientation on energy performance. Other studies have addressed general window characteristics. Cherier et al.16 optimized WWR, window orientation, and glazing type based on building energy performance and solar gain, confirming the significant impact of these variables. Mehdizadeh-Rad et al.1 studied the effect of different window types on cooling load in a university building, finding that glazing type can affect cooling demand by 5–12%. Troup et al.17, using survey data from the CBECS, investigated the influence of WWR in real-world office buildings and found it to be a strong predictor of cooling energy use, and to some extent, lighting and ventilation. Xu et al.18 categorized the components and thermophysical properties of 3,597 windows in China, confirming the influence of window characteristics based on orientation and climatic region. Some studies have also explored window configuration parameters. Elghamry and Hassan19 analyzed the effects of window parameters including shape, design, dimensions, WWR, location, and orientation on thermal comfort, energy consumption, cost, and CO₂ emissions under hot semi-arid climate conditions. Their results show that window configuration significantly affects energy use, though it has little impact on relative humidity; among the parameters studied, WWR and orientation were the most influential. Azmy and Ashmawy20 studied the effect of different window positions and orientations on energy consumption in a typical office room located in Cairo’s hot climate. Their findings indicate that WWR affects energy performance by more than 25%, and that roof openings are optimal for daylighting, as they illuminate over 60% of the room area. Alwetaishi and Benjeddou21 focused on the influence of WWR on energy loads across different orientations and climate zones, showing that the optimal WWR for energy and daylighting performance should vary between 20% and 30%, depending on orientation, to achieve optimal daylight factor and energy efficiency. In higher-altitude locations, this ratio may be slightly higher. Based on the conducted investigations, none of the aforementioned studies have provided a precise analysis of the geometrical parameters of the window, such as window location, the number of horizontal and vertical cells of the window, and the horizontal and vertical spacing between the window cells. Furthermore, in none of the reviewed studies has the impact of space dimensions on the relationship between the geometrical parameters of the window and building energy performance and daylight performance been addressed. While the current body of research extensively investigates the impact of geometrical window parameters on building performance, there remains a notable gap concerning the integration of traditional architectural knowledge from the Yazd region, particularly regarding vernacular window design. Yazd’s historical architecture embodies culturally and climatically responsive principles that evolved over centuries to optimize daylighting and thermal comfort. A brief consideration of these traditional approaches enriches the contextual understanding of why contemporary window design strategies diverge from vernacular precedents. Table 1Comparative study between subject of the research and the literature Review.ResearchDifference of the Present StudyAraújo et al.14Considering the impact of window geometric specifications on energy performance and daylightingCherier et al.16Integrating a comprehensive set of geometric specifications of window across diverse room configurationsMehdizadeh-Rad et al.1Considering the common dimensions of rooms in contemporary Iranian culture based on field surveys in research methodsTan et al.22Considering the horizontal and general position of the window in the room viewAkram et al.2Considering various horizontal and general window divisionsSorooshnia et al.23Integrating a generative design approach for automatic window generation in room facadesCannavale et al.4Investigating energy and daylight performance of windows in various room configurations Table 1Comparative study between subject of the research and the literature Review.ResearchDifference of the Present StudyAraújo et al.14Considering the impact of window geometric specifications on energy performance and daylightingCherier et al.16Integrating a comprehensive set of geometric specifications of window across diverse room configurationsMehdizadeh-Rad et al.1Considering the common dimensions of rooms in contemporary Iranian culture based on field surveys in research methodsTan et al.22Considering the horizontal and general position of the window in the room viewAkram et al.2Considering various horizontal and general window divisionsSorooshnia et al.23Integrating a generative design approach for automatic window generation in room facadesCannavale et al.4Investigating energy and daylight performance of windows in various room configurations Comparative study between subject of the research and the literature Review. Accordingly, the main research gap addressed in this study is the lack of integration of a comprehensive set of window geometrical parameters and room dimensional variations in relation to energy and daylight performance, as well as the absence of a systematic guideline for designing south-facing windows in residential buildings in the city of Yazd (Table 1). This divergence underscores the necessity for advanced optimization techniques that systematically evaluate geometric parameters within diverse room dimensions and climatic conditions pertinent to Yazd. By bridging this gap, the present study not only addresses modern energy and daylight performance challenges but also situates its contribution within the continuum of local architectural heritage and innovation. The novelty of this research lies in applying a comprehensive generative design framework that integrates a full set of window geometrical parameters with multi-objective evolutionary optimization, across eight common residential room configurations, to derive evidence-based guidelines for hot-arid housing design (Table 1). Accordingly, the objectives of this study are: To model and optimize the key geometrical parameters of windows, including width, height, sill height, horizontal and vertical position, number of subdivisions, and spacing patterns, within the facades of residential buildings in Yazd.To analyze how variations in room geometry (length, width, and height across eight spatial types) influence the optimal configuration of window parameters.To perform multi-objective optimization of window design with respect to cooling demand, heating demand, and average Useful Daylight Illuminance (UDI), thereby quantifying trade-offs between energy and daylight performance.To examine Pareto front solutions in order to identify dominant geometrical patterns that balance energy efficiency and visual comfort.To determine the relative significance of each window parameter on energy consumption, thermal loads, daylight autonomy (DA), glare probability, and UDI through statistical analysis.To validate the simulation and optimization results using real energy consumption data from residential units in Yazd.To propose context-specific design guidelines for window geometry that not only enhance energy and daylight performance but also offer a foundation for performance-based updates to Iran’s national building codes. To model and optimize the key geometrical parameters of windows, including width, height, sill height, horizontal and vertical position, number of subdivisions, and spacing patterns, within the facades of residential buildings in Yazd. To model and optimize the key geometrical parameters of windows, including width, height, sill height, horizontal and vertical position, number of subdivisions, and spacing patterns, within the facades of residential buildings in Yazd. To analyze how variations in room geometry (length, width, and height across eight spatial types) influence the optimal configuration of window parameters. To analyze how variations in room geometry (length, width, and height across eight spatial types) influence the optimal configuration of window parameters. To perform multi-objective optimization of window design with respect to cooling demand, heating demand, and average Useful Daylight Illuminance (UDI), thereby quantifying trade-offs between energy and daylight performance. To perform multi-objective optimization of window design with respect to cooling demand, heating demand, and average Useful Daylight Illuminance (UDI), thereby quantifying trade-offs between energy and daylight performance. To examine Pareto front solutions in order to identify dominant geometrical patterns that balance energy efficiency and visual comfort. To examine Pareto front solutions in order to identify dominant geometrical patterns that balance energy efficiency and visual comfort. To determine the relative significance of each window parameter on energy consumption, thermal loads, daylight autonomy (DA), glare probability, and UDI through statistical analysis. To determine the relative significance of each window parameter on energy consumption, thermal loads, daylight autonomy (DA), glare probability, and UDI through statistical analysis. To validate the simulation and optimization results using real energy consumption data from residential units in Yazd. To validate the simulation and optimization results using real energy consumption data from residential units in Yazd. To propose context-specific design guidelines for window geometry that not only enhance energy and daylight performance but also offer a foundation for performance-based updates to Iran’s national building codes. To propose context-specific design guidelines for window geometry that not only enhance energy and daylight performance but also offer a foundation for performance-based updates to Iran’s national building codes. Methodology This study focuses on the multi-objective optimization of the geometrical parameters of windows in the city of Yazd, Iran, based on building energy performance and daylighting. Accordingly, the research is conducted through the following steps: Step 1: Selecting the geometrical parameters of windows through a content analysis of the relevant literature;Step 2: Coding a generative design algorithm for window configuration based on these geometrical parameters;Step 3: Integrating the building energy and daylight performance simulation algorithm and the NSGA-II into the generative design algorithm;Step 4: Executing the optimization process of the geometrical parameters of windows based on energy and daylight performance objectives;Step 5: Analyzing the optimal solutions obtained from the Pareto front (Fig. 1). Step 1: Selecting the geometrical parameters of windows through a content analysis of the relevant literature; Step 1: Selecting the geometrical parameters of windows through a content analysis of the relevant literature; Step 2: Coding a generative design algorithm for window configuration based on these geometrical parameters; Step 2: Coding a generative design algorithm for window configuration based on these geometrical parameters; Step 3: Integrating the building energy and daylight performance simulation algorithm and the NSGA-II into the generative design algorithm; Step 3: Integrating the building energy and daylight performance simulation algorithm and the NSGA-II into the generative design algorithm; Step 4: Executing the optimization process of the geometrical parameters of windows based on energy and daylight performance objectives; Step 4: Executing the optimization process of the geometrical parameters of windows based on energy and daylight performance objectives; Step 5: Analyzing the optimal solutions obtained from the Pareto front (Fig. 1). Step 5: Analyzing the optimal solutions obtained from the Pareto front (Fig. 1). Based on the aim of the study, the energy and daylight performance simulations of the window are conducted for a single room. The generalization of the results to the entire building in the daylighting performance section has been achieved through the use of the UDI metric. In general, daylight performance and UDI levels are independent of adjacent spaces; therefore, the optimized window parameters obtained for each room dimension can be applied to any residential space. Furthermore, the generalizability of the energy performance results to the whole building has been validated based on the energy performance analysis of ten residential buildings and a comprehensive review of the literature. According to previous studies, the presence of neighboring spaces around a room can improve its energy performance but does not affect the optimal geometrical parameters of the room itself. Additionally, the energy performance validation process in this research confirms the accuracy of the simulation results. Initially, through a qualitative content analysis of the relevant literature, the geometrical parameters of the window including window width, window height, window location, the number of horizontal and vertical subdivisions, and the horizontal and vertical spacing between the window grid cells on the façade of a residential apartment were identified. This selection encompassed a comprehensive set of window geometrical parameters. Fig. 1Research Methodology Framework Proposed in This Study. Research Methodology Framework Proposed in This Study. Subsequently, based on the extracted geometrical parameters, a generative design algorithm was coded to configure the window on the southern façade of a room with defined dimensions. In this context, a review of the literature revealed that the dimensions of the room in which the window is installed have a significant impact on both energy and daylight performance. Given that the primary focus of this study is the optimization of window geometrical parameters, the room dimensions were defined as a control variable. Accordingly, a field survey was conducted to examine the interior spatial dimensions of residential units in the city of Yazd. A total of 85 residential units were analyzed, and eight common spatial configurations of rooms in Yazd’s residential apartments were selected (Fig. 2). Fig. 2Selected spaces and their dimensions for the optimization process. Selected spaces and their dimensions for the optimization process. Next, an algorithm for generating a window within a room with specified length, width, and height was developed, based on window width, window height, window sill height, window location, and the number of horizontal and vertical subdivisions. This algorithm was coded using the Python programming language within the Grasshopper environment and Rhinoceros 8 software. The resulting code consists of 3 input genes representing the room’s geometrical parameters and 8 input genes representing the window’s geometrical parameters, with the output being the window surface area (Fig. 3). Fig. 3Coding of the generative design process for window configuration based on its geometrical parameters. Coding of the generative design process for window configuration based on its geometrical parameters. In this context, the 3 input genes for the room’s geometrical parameters include the east-west dimension, the north-south dimension, and the room height, while the 8 input genes for the window’s geometrical parameters comprise window width, window height, horizontal window position, vertical window position, number of horizontal divisions, number of vertical divisions, horizontal spacing between window cells, and vertical spacing between window cells (Fig. 4). Fig. 4Parameters of the Generative Design for Window Configuration Based on Its Geometrical Parameters. Parameters of the Generative Design for Window Configuration Based on Its Geometrical Parameters. After defining the generative design algorithm for window configuration, the range of variation for the input genes was determined (Table 2). The length and width of the room were based on eight common spatial configurations of rooms in residential units in Yazd city. The room height was fixed at 3 m. The range of variation for window width and height was defined by dividing the east-west dimension and the room height into 20 segments. The number of horizontal and vertical divisions ranged from 1 to 10 segments, and the horizontal and vertical spacing between window cells was defined from 0 to 1 m in 11 increments. Table 2Input genes for the optimization Process.Input GenesMinimum StepMaximum StepNumber of StepsStep SizeWindow WidthREW/20REW20REW/20Window HeightRH/20RH20RH/20Number of Horizontal Divisions of the Window110101Number of Vertical Divisions of the Window110101Horizontal Distance Between Window Cells0.01110.1Vertical Distance Between Window Cells0.01110.1Horizontal Position of the Window-(REW-WW)REW-WW200(REW-WW)/100Vertical Position of the Window-(RH-WH)RH-WH200(RH-WH)/100REW: Dimensions of the Room’s East-West Side; RH: Dimensions of Room’s Height; WW: Window’s Width; WH: Window’s Height. Table 2Input genes for the optimization Process.Input GenesMinimum StepMaximum StepNumber of StepsStep SizeWindow WidthREW/20REW20REW/20Window HeightRH/20RH20RH/20Number of Horizontal Divisions of the Window110101Number of Vertical Divisions of the Window110101Horizontal Distance Between Window Cells0.01110.1Vertical Distance Between Window Cells0.01110.1Horizontal Position of the Window-(REW-WW)REW-WW200(REW-WW)/100Vertical Position of the Window-(RH-WH)RH-WH200(RH-WH)/100REW: Dimensions of the Room’s East-West Side; RH: Dimensions of Room’s Height; WW: Window’s Width; WH: Window’s Height. Input genes for the optimization Process. REW: Dimensions of the Room’s East-West Side; RH: Dimensions of Room’s Height; WW: Window’s Width; WH: Window’s Height. Subsequently, energy performance and daylighting simulations were integrated with the generative design algorithm within Rhinoceros 8, Grasshopper, and the Ladybug Tools plugin, based on the climatic data from the Yazd weather station database (IRN_YA_Yazd.Sadooghi.Intl.AP.408210_TMYx). Furthermore, building materials were selected according to the ASHRAE 90.1 2019 standard recommendations for the Yazd climate zone (2B) during the simulation process. The simulation employed the Midrise Apartment program and the Apartment subgroup based on the ASHRAE 90.1 2019 standard (Table 3). The outputs from the energy performance and daylight simulations included total energy consumption, cooling demand, heating demand, average UDI, average DA of the room, and the average DGP for the room. Table 3Envelope and systems Properties.TypeU ValueSHGCR Value(SI)R Value(IP)Solar TransmittanceVisibleTransmittanceWindowSimple Glazing WindowGeneric Exterior Window0.450.250.219261.2487970.370.40352315TypeR Value (SI)R Value (IP)Solar ReflectanceVisibleTransmittanceWallTypical Insulated Steel Framed Exterior Wall-R122.17139612.3297610.5RoofTypical IEAD RoofHighly Reflective-R264.63829726.3374730.8FloorTypical Insulated Steel Framed Exterior Floor-R274.67730926.5589910.2TypePeopleLoadLightingLoadInfiltrationLoadHot WaterLoadElectric Equipment LoadHVACSystemBuildingTypical MidriseApartment0.028people/m29.4W/m20.000569m3/s-m20.14925829895L/h-m26.7W/m2Ideal Air System Envelope and systems Properties. Simple Glazing Window Generic Exterior Window Visible Transmittance Typical IEAD Roof Highly Reflective-R26 People Load Lighting Load Infiltration Load Hot Water Load HVAC System Typical Midrise Apartment 0.028 people/m2 9.4 W/m2 0.000569 m3/s-m2 0.14925829895 L/h-m2 6.7 W/m2 While the ASHRAE 90.1 2019 standard was referenced for material selection to ensure a recognized baseline of thermal properties, special attention was also given to reflect the typical construction practices in Yazd. Where possible, local material data and construction methods commonly employed in residential buildings were incorporated into the modeling to enhance the accuracy and relevance of the simulations. Utilizing localized construction data helps capture the specific thermal behavior and energy performance characteristics inherent to the region, which may differ from generic standards developed for broader contexts. Future studies could further refine this approach by systematically comparing results based on ASHRAE standards versus locally sourced material properties to quantify the impact on simulation outcomes more explicitly. Following this, for each of the eight selected rooms, an optimization process was defined to optimize the geometrical parameters of the window based on cooling demand, heating demand, and the average UDI in the Yazd climate. The optimization processes were carried out using the NSGA-II in Rhinoceros 8, Grasshopper, and the Wallacei plugin. The input genes of the optimization algorithm included window width, window height, number of horizontal subdivisions (rows), number of vertical subdivisions (columns), horizontal distance between window cells, vertical distance between window cells, horizontal window position, and vertical window position based on the predefined variation ranges (Fig. 5). Fig. 5The explanation of parameter ranges and control variables. The explanation of parameter ranges and control variables. In the final stage of the study, after conducting eight optimization processes, the analysis of the Pareto frontier optimal solutions was performed. For the results of the eight optimization processes, window geometrical parameters including WWR, window height, number of horizontal subdivisions, number of vertical subdivisions, horizontal spacing between window cells, vertical spacing between window cells, horizontal window position, and vertical window position alongside room energy and daylight performance indicators including total energy consumption, cooling demand, heating demand, average UDI, average DA, and average annual DGP were examined(Fig. 5). The analysis of the Pareto frontier data involved descriptive statistics, Pearson correlation analysis, and linear regression tests conducted using SPSS Statistics 27. For the validation of the simulation model, ten residential units (Table 4), which were part of an initial survey of 85 units, were selected. A table containing the background data of the ten residential units used for validation is presented, including the year of construction, number of floors, total floor area, building dimensions, orientation, apartment unit area, floor level of the unit, window-to-wall ratio (WWR) of all façades, number of occupants, and occupancy load. The actual electricity and gas consumption of these units was collected from monthly bills over one year and converted into kilowatt-hours. Subsequently, energy consumption for the same units was calculated using the simulation model. Table 4Contextual data of the ten residential units used for validation.Building CodeYear of ConstructionNumber of Floors on the pilotTotal AreaEast-West SideNorth-South SideArea of ​​Residential UnitOrientation to the NorthFloor of the UnitSouth WWRNorth WWREast WWRWest WWRNumber of ResidentsOccupancy LoadNo. 12019422512.518193020.7000.240.0177No. 22022615114.510.41135810.60.20030.0198No. 3202041931216.08175530.60.450060.031No. 42020596128751220.40.20020.0208No. 5202051881314.46168040.70.250040.0212No. 62023491137765400.40020.0219No. 7201851601610143830.600.3040.025No. 82020487117.9731230.550.450030.0344No. 92018616212.512.96141520.700040.0246No. 102017422312.517.84199030.7000.2550.0224 Table 4Contextual data of the ten residential units used for validation.Building CodeYear of ConstructionNumber of Floors on the pilotTotal AreaEast-West SideNorth-South SideArea of ​​Residential UnitOrientation to the NorthFloor of the UnitSouth WWRNorth WWREast WWRWest WWRNumber of ResidentsOccupancy LoadNo. 12019422512.518193020.7000.240.0177No. 22022615114.510.41135810.60.20030.0198No. 3202041931216.08175530.60.450060.031No. 42020596128751220.40.20020.0208No. 5202051881314.46168040.70.250040.0212No. 62023491137765400.40020.0219No. 7201851601610143830.600.3040.025No. 82020487117.9731230.550.450030.0344No. 92018616212.512.96141520.700040.0246No. 102017422312.517.84199030.7000.2550.0224 Contextual data of the ten residential units used for validation. A comparison of the results showed that the average difference between the actual and simulated values was less than 10%, and the Pearson correlation test yielded a coefficient of 0.86, indicating a strong and direct relationship between the actual and modeled data. The use of a subset of the original sample, while ensuring consistency between the datasets, allowed for a valid comparison within the studied population. These findings confirm the accuracy of the simulation model in reflecting actual energy consumption patterns. For model validation, the two statistical indicators NMBE and CVRMSE were evaluated based on both total energy consumption and energy use per square meter (Table 5). Table 5NMBE and CVRMSE report for model validation.Energy ConsumptionEnergy Consumption per AreaNMBE−4.0038411%−4.7562548%CVRMSE6.6928377%9.2747881% Table 5NMBE and CVRMSE report for model validation.Energy ConsumptionEnergy Consumption per AreaNMBE−4.0038411%−4.7562548%CVRMSE6.6928377%9.2747881% NMBE and CVRMSE report for model validation. The results showed an NMBE below 5% and a CVRMSE below 15%, which are within the acceptable limits defined by ASHRAE standards, thereby confirming the model’s reliability. Future work may involve validation with an independent set of units to strengthen the model’s generalizability in broader contexts. Furthermore, to assess reliability, sensitivity analysis of the generative design simulation algorithm results was conducted using Pearson correlation tests between simulation inputs and outputs. The significance level was close to zero and below 0.05, and the multivariate correlation coefficient with energy performance exceeded 0.7, confirming the reliability of the simulation algorithm. Results Based on the presented content, eight optimization processes were conducted. The optimization results demonstrate a high convergence among the Pareto frontier responses. The Pareto frontier charts are illustrated in Fig. 6. The optimization outcomes indicate a direct influence of the geometrical parameters of the window on energy and daylight performance indices. According to the eight completed optimization processes, variations in the geometrical parameters of the window led to a minimum 2.5% and maximum 3.9% reduction in room energy consumption. These changes reduced the energy use intensity (EUI) up to 18.93 kWh/m², and the total energy consumption up to 506.66 kWh. Fig. 6Pareto Frontier Charts of the Eight Conducted Optimization Processes. Pareto Frontier Charts of the Eight Conducted Optimization Processes. The cooling demand accounted for the largest portion of this reduction, as variations in window geometrical parameters across the eight optimization processes resulted in a minimum 19.7% to maximum 23.2% decrease in cooling demand and a minimum 2.3% to maximum 2.4% decrease in heating demand. Moreover, these variations reduced the cooling demand up to 24.12 kWh/m² and 652.65 kWh and the heating demand up to 5.18 kWh/m² and 147.78 kWh. This finding suggests that changes in the geometrical parameters of the window have a more significant effect on cooling demand. In other words, it can be stated that the geometrical parameters of the window have the greatest influence on cooling demand among the indicators of building energy performance. To enable a more precise examination of energy and daylighting performance, descriptive statistics of the performance objectives were presented based on the Pareto-front solutions obtained from the eight conducted optimization processes (Table 6). Table 6Descriptive statistics results of Pareto frontier outcomes for energy performance and daylight Metrics.RangeMinimumMaximumMeanMeanSth. DeviationStatisticStatisticStatisticStatisticSth. ErrorStatisticStatisticRectangle12m23–4 mEnergy Consumption18.939972370.129224389.069196376.214048101.0108563345.53668816830.655Cooling Demand24.12869094.411055118.539745103.088860571.3002352347.12168167950.718Heating Demand5.188717198.018314203.207031200.614049930.3027864861.6584298822.750DA Average69.4566672.63375072.09041727.400972233.68705610720.194838007407.831UDI Average72.3108346.63583378.94666749.946361104.46837181824.474280401598.990DGP Average3.27364699.99953196.72588598.792578170.1836816511.0060658361.012Rectangle12m24 m–3 mEnergy Consumption18.163765367.821067385.984832374.811249841.0917510766.07861273436.950Cooling Demand23.01542892.171294115.186722101.679545521.3553952867.54652157156.950Heating Demand4.851905198.286731203.138636200.620566770.2833770061.5777763952.489DA Average81.1254163.76666784.89208335.359085904.33211172924.120177302581.783UDI Average75.2725005.17833380.45083353.948373554.44461591624.746574104612.393DGP Average4.09093799.96135495.87041798.452184160.2227175321.2400387371.538Square16m24 m–4 mEnergy Consumption15.598966335.481349351.080315341.568809470.9105182924.98711407624.871Cooling Demand20.05939281.536677101.59606990.019515401.1730561776.42509329141.282Heating Demand4.470028176.963506181.433534179.038156570.2719856691.4897268622.219DA Average72.4516670.76333373.21500035.396083304.05049132322.185454664492.194UDI Average73.2655565.82472279.09027858.202481434.29980305423.550991256554.649DGP Average3.392674100.00000096.60732698.455600730.1946105621.0659259471.136Rectangle35m27 m–5 mEnergy Consumption15.005689271.627776286.633465276.723210370.7699066564.21695242617.783Cooling Demand17.79073058.58945476.38018465.599909970.9430445825.16526790426.680Heating Demand3.384396137.142789140.527185138.612162830.1875435791.0272184901.055DA Average61.8905712.87900064.76957131.589723803.63971469519.935538414397.426UDI Average74.4432865.74800080.19128654.542357174.46851648524.475072776599.029DGP Average2.86771599.97280497.10508998.673091100.1660544930.9095179180.827Square36m26 m–6 mEnergy Consumption12.617375268.618686281.236061273.229672970.6882412103.83196488414.684Cooling Demand15.92755958.08892974.01648864.574054420.8775261394.88585876523.872Heating Demand3.310183134.708436138.018619136.144481030.1977245151.1008835101.212DA Average62.2676541.91802564.18567927.467849393.52308271419.615694384384.775UDI Average76.2028402.65925978.86209952.421449584.54866051025.325869887641.400DGP Average3.030818100.00000096.96918298.684799870.1609034010.8958722210.803Rectangle45m29 m–5 mEnergy Consumption11.868828258.334762270.203590261.973634610.5812202803.33885631011.148Cooling Demand15.11898253.48632868.60531058.687567450.7546906724.33536784418.795Heating Demand3.250154129.087143132.337297130.774929640.1808027241.0386325741.079DA Average63.7083670.00000063.70836725.992235553.55391128120.415665995416.799UDI Average78.2521430.00000078.25214346.332940614.56324770926.213862333687.167DGP Average3.135982100.00000096.86401898.783182550.1620684130.9310121500.867Rectangle45m25–9 mEnergy Consumption11.489212261.392235272.881447265.804177970.7184855893.93530764615.487Cooling Demand14.43405256.89082971.32488162.864087300.9002510854.93087826824.314Heating Demand3.054748129.040524132.095272130.428953200.1897456321.0392796271.080DA Average49.6950001.18612250.88112225.017527172.96506599216.240335282263.748UDI Average68.0671435.54316373.61030645.243989773.83394412220.999376800440.974DGP Average2.928202100.00000097.071798−98.717881770.1731558650.9484137300.899Square49m27 m–7 mEnergy Consumption10.340009249.696965260.036974253.739523160.5788232053.22275121110.386Cooling Demand13.31959051.57562164.89521157.292798350.7460427114.15379001917.254Heating Demand3.015934122.630625125.646559123.935587190.1740483490.9690601950.939DA Average56.65581.271557.927325.4383523.229936117.9835234323.407UDI Average76.68962.150778.840349.9407814.758289326.4930337701.881DGP Average2.752637100.00000097.24736398.792886350.1587989220.8841549800.782 Table 6Descriptive statistics results of Pareto frontier outcomes for energy performance and daylight Metrics.RangeMinimumMaximumMeanMeanSth. DeviationStatisticStatisticStatisticStatisticSth. ErrorStatisticStatisticRectangle12m23–4 mEnergy Consumption18.939972370.129224389.069196376.214048101.0108563345.53668816830.655Cooling Demand24.12869094.411055118.539745103.088860571.3002352347.12168167950.718Heating Demand5.188717198.018314203.207031200.614049930.3027864861.6584298822.750DA Average69.4566672.63375072.09041727.400972233.68705610720.194838007407.831UDI Average72.3108346.63583378.94666749.946361104.46837181824.474280401598.990DGP Average3.27364699.99953196.72588598.792578170.1836816511.0060658361.012Rectangle12m24 m–3 mEnergy Consumption18.163765367.821067385.984832374.811249841.0917510766.07861273436.950Cooling Demand23.01542892.171294115.186722101.679545521.3553952867.54652157156.950Heating Demand4.851905198.286731203.138636200.620566770.2833770061.5777763952.489DA Average81.1254163.76666784.89208335.359085904.33211172924.120177302581.783UDI Average75.2725005.17833380.45083353.948373554.44461591624.746574104612.393DGP Average4.09093799.96135495.87041798.452184160.2227175321.2400387371.538Square16m24 m–4 mEnergy Consumption15.598966335.481349351.080315341.568809470.9105182924.98711407624.871Cooling Demand20.05939281.536677101.59606990.019515401.1730561776.42509329141.282Heating Demand4.470028176.963506181.433534179.038156570.2719856691.4897268622.219DA Average72.4516670.76333373.21500035.396083304.05049132322.185454664492.194UDI Average73.2655565.82472279.09027858.202481434.29980305423.550991256554.649DGP Average3.392674100.00000096.60732698.455600730.1946105621.0659259471.136Rectangle35m27 m–5 mEnergy Consumption15.005689271.627776286.633465276.723210370.7699066564.21695242617.783Cooling Demand17.79073058.58945476.38018465.599909970.9430445825.16526790426.680Heating Demand3.384396137.142789140.527185138.612162830.1875435791.0272184901.055DA Average61.8905712.87900064.76957131.589723803.63971469519.935538414397.426UDI Average74.4432865.74800080.19128654.542357174.46851648524.475072776599.029DGP Average2.86771599.97280497.10508998.673091100.1660544930.9095179180.827Square36m26 m–6 mEnergy Consumption12.617375268.618686281.236061273.229672970.6882412103.83196488414.684Cooling Demand15.92755958.08892974.01648864.574054420.8775261394.88585876523.872Heating Demand3.310183134.708436138.018619136.144481030.1977245151.1008835101.212DA Average62.2676541.91802564.18567927.467849393.52308271419.615694384384.775UDI Average76.2028402.65925978.86209952.421449584.54866051025.325869887641.400DGP Average3.030818100.00000096.96918298.684799870.1609034010.8958722210.803Rectangle45m29 m–5 mEnergy Consumption11.868828258.334762270.203590261.973634610.5812202803.33885631011.148Cooling Demand15.11898253.48632868.60531058.687567450.7546906724.33536784418.795Heating Demand3.250154129.087143132.337297130.774929640.1808027241.0386325741.079DA Average63.7083670.00000063.70836725.992235553.55391128120.415665995416.799UDI Average78.2521430.00000078.25214346.332940614.56324770926.213862333687.167DGP Average3.135982100.00000096.86401898.783182550.1620684130.9310121500.867Rectangle45m25–9 mEnergy Consumption11.489212261.392235272.881447265.804177970.7184855893.93530764615.487Cooling Demand14.43405256.89082971.32488162.864087300.9002510854.93087826824.314Heating Demand3.054748129.040524132.095272130.428953200.1897456321.0392796271.080DA Average49.6950001.18612250.88112225.017527172.96506599216.240335282263.748UDI Average68.0671435.54316373.61030645.243989773.83394412220.999376800440.974DGP Average2.928202100.00000097.071798−98.717881770.1731558650.9484137300.899Square49m27 m–7 mEnergy Consumption10.340009249.696965260.036974253.739523160.5788232053.22275121110.386Cooling Demand13.31959051.57562164.89521157.292798350.7460427114.15379001917.254Heating Demand3.015934122.630625125.646559123.935587190.1740483490.9690601950.939DA Average56.65581.271557.927325.4383523.229936117.9835234323.407UDI Average76.68962.150778.840349.9407814.758289326.4930337701.881DGP Average2.752637100.00000097.24736398.792886350.1587989220.8841549800.782 Descriptive statistics results of Pareto frontier outcomes for energy performance and daylight Metrics. Rectangle 12m2 3–4 m Rectangle 12m2 4 m–3 m Square 16m2 4 m–4 m Rectangle 35m2 7 m–5 m Square 36m2 6 m–6 m Rectangle 45m2 9 m–5 m Rectangle 45m2 5–9 m Square 49m2 7 m–7 m According to the findings, the best-performing solutions among the eight optimization processes, based on each individual performance objective are presented in Fig. 7. The results indicate that solutions with higher WWR generally provide better performance in terms of heating demand and all daylighting metrics. In contrast, solutions with lower WWR are associated with improved cooling performance. Due to the dominant share of cooling demand compared to heating demand in the studied climate, solutions with lower WWR also tend to be more energy-efficient overall. Moreover, the results show that the best combined energy and daylighting performance was achieved by solutions featuring single-cell windows. With an increase in room depth relative to the daylight-exposed façade, the optimal energy and daylight performance of the building is achieved with a higher window-to-wall ratio. Based on the Pareto front solutions, it can be concluded that at a constant window-to-wall ratio, horizontal elongation of the window reduces energy consumption and cooling demand, while vertical elongation of the window decreases heating demand in residential rooms in the city of Yazd. Also, at a fixed WWR, vertically extended single-cell windows or horizontally extended windows with horizontal divisions lead to higher average UDI. Fig. 7Best Pareto frontier responses of the eight optimization processes performed, shown according to each performance objective. Best Pareto frontier responses of the eight optimization processes performed, shown according to each performance objective. These results highlight the trade-offs and performance contradictions that arise from different geometric window configurations. Therefore, selecting the appropriate window geometry is critical for achieving optimal building performance. In this context, descriptive statistics, correlation analyses, and regression tests of the Pareto-optimal solutions were conducted to further explore these relationships. Accordingly, the best responses among the eight optimization processes based on each performance objective are presented in Fig. 8. Fig. 8Best Pareto Frontier Responses of the Eight Optimization Processes Based on Each Performance Objective. Best Pareto Frontier Responses of the Eight Optimization Processes Based on Each Performance Objective. Figure 8 shows a strong positive correlation between the WWR and daylighting metrics such as UDI, DA, and DGP, highlighting the significant impact of increasing window area on natural light availability. However, the correlation between WWR and heating demand is weaker but still statistically significant, reflecting the complex trade-offs involved. This indicates that larger windows improve daylight penetration and visual comfort but may also lead to increased heat loss during colder periods, slightly raising heating requirements. To further investigate in more detail and quantitatively, descriptive statistics of the entire set of Pareto frontier responses from the eight optimization processes were examined. The mean WWR is 0.2, and the median window height is 2.22 m (Table 7); indicating that a significant portion of the Pareto frontier solutions feature windows with a WWR below 0.5 and vertical elongation. Moreover, among all Pareto frontier responses, the median number of horizontal rows in the window cell grid is 1.15, and the median number of vertical columns is 3.81. Based on this, it can be stated that according to the mean of the Pareto frontier solution set across the eight optimization processes, the window cell grid typically consists of one row and four columns. Additionally, the median horizontal distance between window cells is 0.228 m, and the median vertical distance between window cells is 0.206 m. Additionally, the mean vertical position of the window was found to be −0.481 and the median horizontal position − 0.109 (Table 7). Table 7Descriptive statistics results of window geometrical parameters for all Pareto frontier responses from the eight optimization Processes. N MinimumMaximumMeanStd. DeviationVarianceWWR2400.000.940.20540.169400.029Window Height2400.152.982.22110.890830.794Number of Rows2401101.150.8070.651Number of Columns2401103.813.29110.833Horizontal Position of the Window240−0.91.0−0.1090.31600.100Vertical Position of the Window240−1.01.0−0.4810.45160.204Vertical Distance between Cells2400.00.50.2060.20010.040Horizonal Distance between Cells2400.00.50.2280.21620.047 Table 7Descriptive statistics results of window geometrical parameters for all Pareto frontier responses from the eight optimization Processes. N MinimumMaximumMeanStd. DeviationVarianceWWR2400.000.940.20540.169400.029Window Height2400.152.982.22110.890830.794Number of Rows2401101.150.8070.651Number of Columns2401103.813.29110.833Horizontal Position of the Window240−0.91.0−0.1090.31600.100Vertical Position of the Window240−1.01.0−0.4810.45160.204Vertical Distance between Cells2400.00.50.2060.20010.040Horizonal Distance between Cells2400.00.50.2280.21620.047 Descriptive statistics results of window geometrical parameters for all Pareto frontier responses from the eight optimization Processes. Subsequently, for a more detailed and granular analysis, the descriptive statistics of the Pareto frontier results were examined separately for each of the eight optimization processes (corresponding to each room) (Table 8). The range of mean values for the WWR across the eight studied spaces varies from 0.134 for the 5-meter-deep Room 9 to 0.332 for the 9-meter-deep Room 5. The highest ratio corresponds to the room with the greatest depth relative to the southern facade, while the lowest ratio pertains to the room with the largest southern facade area (Table 8). Accordingly, the optimal WWR can be considered dependent on two factors: room depth and room area. The median window height within the optimal solutions ranges from 2.03 m to 2.48 m (Table 8). The lowest median window height belongs to Room 9 with a 5-meter depth, and the highest to Room 6 with a 6-meter depth. The mean window height exceeding 2 m across all eight optimization processes, combined with a mean WWR below 0.35, reflects a tendency of the optimal solutions toward vertical elongation of the window (Table 8). Table 8Descriptive statistics of Pareto frontier results for WWR and window height by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mWWR0.010.940.23193Window Height0.302.980.88073Rectangle; 12m2 − 4 m–3 mWWR0.010.340.10453Window Height0.302.980.93105Square; 16m2 4 m–4 mWWR0.010.390.12314Window Height0.452.981.04855Rectangle; 35m2 − 7 m–5 mWWR0.010.900.17552Window Height0.452.980.62680Square; 36m2; 6 m–6 mWWR0.020.500.14744Window Height0.452.980.73272Rectangle 45m2 − 9 m–5 mWWR0.0010.380.11055Window Height0.152.980.98898Rectangle 45m2 − 5–9 mWWR0.050.710.20927Window Height0.302.980.96085Square; 49m2; 7 m–7 mWWR0.020.500.15250Window Height0.302.980.87077 Table 8Descriptive statistics of Pareto frontier results for WWR and window height by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mWWR0.010.940.23193Window Height0.302.980.88073Rectangle; 12m2 − 4 m–3 mWWR0.010.340.10453Window Height0.302.980.93105Square; 16m2 4 m–4 mWWR0.010.390.12314Window Height0.452.981.04855Rectangle; 35m2 − 7 m–5 mWWR0.010.900.17552Window Height0.452.980.62680Square; 36m2; 6 m–6 mWWR0.020.500.14744Window Height0.452.980.73272Rectangle 45m2 − 9 m–5 mWWR0.0010.380.11055Window Height0.152.980.98898Rectangle 45m2 − 5–9 mWWR0.050.710.20927Window Height0.302.980.96085Square; 49m2; 7 m–7 mWWR0.020.500.15250Window Height0.302.980.87077 Descriptive statistics of Pareto frontier results for WWR and window height by space Dimensions. Relationship of optimal WWR and room depth is largely influenced by the trade-off between daylight provision and thermal performance in deeper spaces. Although it is true that deeper rooms generally require more daylight, increasing window size beyond a certain point leads to a disproportionate increase in cooling loads due to increased solar heat gain, especially in the hot climate of Yazd. Therefore, the optimization process tends to favor a smaller WWR for deeper rooms to reduce excessive cooling demand, while still maintaining adequate daylight levels through adjustment of window location and shape. This trade-off highlights the importance of simultaneously considering daylight benefits and energy costs when determining the optimal window geometry at different room depths. Based on the analysis of the horizontal and vertical positions of the window center relative to the center of the southern facade in the Pareto frontier responses (Table 9), the median vertical position across all response sets is a negative value, indicating that the window center is located lower than the facade center. Table 9Descriptive statistics of Pareto frontier results for general and horizontal window position by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mHorizontal Window Position−0.70.5−0.4870.2270Vertical Window Position−1.0−0.2−0.6500.2556Rectangle; 12m2 − 4 m–3 mHorizontal Window Position−0.31.00.1160.4259Vertical Window Position−1.00.5−0.5680.4929Square; 16m2 4 m–4 mHorizontal Window Position−0.70.1−0.2170.2102Vertical Window Position−1.01.0−0.3230.5600Rectangle; 35m2 − 7 m–5 mHorizontal Window Position−0.90.9−0.0930.4518Vertical Window Position−1.00.0−0.6270.2924Square; 36m2; 6 m–6 mHorizontal Window Position−0.40.4−0.0870.1544Vertical Window Position−1.00.1−0.6160.4375Rectangle 45m2 − 9 m–5 mHorizontal Window Position−0.20.50.0550.1460Vertical Window Position−1.00.1−0.1390.3840Rectangle 45m2 − 5–9 mHorizontal Window Position−0.50.3−0.0730.2273Vertical Window Position−1.00.1−0.5300.3030Square; 49m2; 7 m–7 mHorizontal Window Position−0.30.1−0.1100.0978Vertical Window Position−1.00.7−0.4260.5489 Table 9Descriptive statistics of Pareto frontier results for general and horizontal window position by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mHorizontal Window Position−0.70.5−0.4870.2270Vertical Window Position−1.0−0.2−0.6500.2556Rectangle; 12m2 − 4 m–3 mHorizontal Window Position−0.31.00.1160.4259Vertical Window Position−1.00.5−0.5680.4929Square; 16m2 4 m–4 mHorizontal Window Position−0.70.1−0.2170.2102Vertical Window Position−1.01.0−0.3230.5600Rectangle; 35m2 − 7 m–5 mHorizontal Window Position−0.90.9−0.0930.4518Vertical Window Position−1.00.0−0.6270.2924Square; 36m2; 6 m–6 mHorizontal Window Position−0.40.4−0.0870.1544Vertical Window Position−1.00.1−0.6160.4375Rectangle 45m2 − 9 m–5 mHorizontal Window Position−0.20.50.0550.1460Vertical Window Position−1.00.1−0.1390.3840Rectangle 45m2 − 5–9 mHorizontal Window Position−0.50.3−0.0730.2273Vertical Window Position−1.00.1−0.5300.3030Square; 49m2; 7 m–7 mHorizontal Window Position−0.30.1−0.1100.0978Vertical Window Position−1.00.7−0.4260.5489 Descriptive statistics of Pareto frontier results for general and horizontal window position by space Dimensions. The mean value of the horizontal position of the window in all room models, except for the 3 × 4 m room, is approximately zero, indicating that the window is positioned at the center of the horizontal axis of the façade. In the 3 × 4 m rooms, however, the values of −0.487 and − 0.217 indicate a tendency for the window to shift toward the left side of the façade (Table 9; Fig. 9). Fig. 9Means of Geometric Characteristics of Window for Pareto Frontier Solutions. Means of Geometric Characteristics of Window for Pareto Frontier Solutions. The mean number of cell rows was approximately one in all cases. This indicates that most of the optimal solutions consist of a single row of window cells. Additionally, the median number of window cell columns ranged from 1.8 to 7.32, showing considerable variation. The highest value was observed for the 3 × 4 m room, while the lowest corresponded to the 3 × 4 m room (Table 10). Table 10Descriptive statistics results of Pareto frontier outcomes for the number of rows and columns of window Cells, categorized by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mNumber of Rows121.0330.183Number of Columns171.8002.074Rectangle; 12m2 − 4 m–3 mNumber of Rows121.1930.402Number of Columns197.0322.763Square; 16m2 4 m–4 mNumber of Rows111.0000.000Number of Columns163.6002.027Rectangle; 35m2 − 7 m–5 mNumber of Rows131.1660.461Number of Columns1105.5333.360Square; 36m2; 6 m–6 mNumber of Rows1101.3541.644Number of Columns1103.1613.465Rectangle 45m2 − 9 m–5 mNumber of Rows111.0000.000Number of Columns195.0003.509Rectangle 45m2 − 5–9 mNumber of Rows161.2330.971Number of Columns162.0001.965Square; 49m2; 7 m–7 mNumber of Rows171.2251.087Number of Columns192.1932.676 Table 10Descriptive statistics results of Pareto frontier outcomes for the number of rows and columns of window Cells, categorized by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mNumber of Rows121.0330.183Number of Columns171.8002.074Rectangle; 12m2 − 4 m–3 mNumber of Rows121.1930.402Number of Columns197.0322.763Square; 16m2 4 m–4 mNumber of Rows111.0000.000Number of Columns163.6002.027Rectangle; 35m2 − 7 m–5 mNumber of Rows131.1660.461Number of Columns1105.5333.360Square; 36m2; 6 m–6 mNumber of Rows1101.3541.644Number of Columns1103.1613.465Rectangle 45m2 − 9 m–5 mNumber of Rows111.0000.000Number of Columns195.0003.509Rectangle 45m2 − 5–9 mNumber of Rows161.2330.971Number of Columns162.0001.965Square; 49m2; 7 m–7 mNumber of Rows171.2251.087Number of Columns192.1932.676 Descriptive statistics results of Pareto frontier outcomes for the number of rows and columns of window Cells, categorized by space Dimensions. The mean vertical distance between window cells in most rooms ranges between 0.10 and 0.16 m. However, the median of the optimal responses for 3 × 4 m room and 4 × 4 m room are respectively 0.361 m and 0.413 m, both exceeding 0.16 m (Table 11). Table 11Descriptive statistics results of Pareto frontier outcomes for vertical and horizontal distances between window Cells, categorized by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mVertical Distance between Cells0.00.50.1330.1539Horizonal Distance between Cells0.00.50.2070.1964Rectangle; 12m2 − 4 m–3 mVertical Distance between Cells0.00.50.3610.1726Horizonal Distance between Cells0.00.50.2350.1624Square; 16m2 4 m–4 mVertical Distance between Cells0.20.50.4130.1008Horizonal Distance between Cells0.00.50.2830.2379Rectangle; 35m2 − 7 m–5 mVertical Distance between Cells0.00.40.1430.1478Horizonal Distance between Cells0.00.50.3670.2123Square; 36m2; 6 m–6 mVertical Distance between Cells0.00.50.2030.2183Horizonal Distance between Cells0.00.30.0550.0995Rectangle 45m2 − 9 m–5 mVertical Distance between Cells0.00.50.1650.2026Horizonal Distance between Cells0.00.40.0650.1380Rectangle 45m2 − 5–9 mVertical Distance between Cells0.00.40.1070.1413Horizonal Distance between Cells0.00.50.2300.2152Square; 49m2; 7 m–7 mVertical Distance between Cells0.00.50.1650.2026Horizonal Distance between Cells0.00.40.0650.1380 Table 11Descriptive statistics results of Pareto frontier outcomes for vertical and horizontal distances between window Cells, categorized by space Dimensions.MinimumMaximumMeanStd. DeviationRectangle; 12m2 − 3–4 mVertical Distance between Cells0.00.50.1330.1539Horizonal Distance between Cells0.00.50.2070.1964Rectangle; 12m2 − 4 m–3 mVertical Distance between Cells0.00.50.3610.1726Horizonal Distance between Cells0.00.50.2350.1624Square; 16m2 4 m–4 mVertical Distance between Cells0.20.50.4130.1008Horizonal Distance between Cells0.00.50.2830.2379Rectangle; 35m2 − 7 m–5 mVertical Distance between Cells0.00.40.1430.1478Horizonal Distance between Cells0.00.50.3670.2123Square; 36m2; 6 m–6 mVertical Distance between Cells0.00.50.2030.2183Horizonal Distance between Cells0.00.30.0550.0995Rectangle 45m2 − 9 m–5 mVertical Distance between Cells0.00.50.1650.2026Horizonal Distance between Cells0.00.40.0650.1380Rectangle 45m2 − 5–9 mVertical Distance between Cells0.00.40.1070.1413Horizonal Distance between Cells0.00.50.2300.2152Square; 49m2; 7 m–7 mVertical Distance between Cells0.00.50.1650.2026Horizonal Distance between Cells0.00.40.0650.1380 Descriptive statistics results of Pareto frontier outcomes for vertical and horizontal distances between window Cells, categorized by space Dimensions. Based on the median geometrical parameters of the window in the optimal responses for each room, windows have been defined for eight rooms (Fig. 10). As shown in the figure, the resulting responses share significant commonalities. Fig. 10Means of Geometric Characteristics of Window’s Cells for Pareto Frontier Solutions. Means of Geometric Characteristics of Window’s Cells for Pareto Frontier Solutions. Overall, all responses derived from the median geometrical parameters of the window in the optimal solutions feature vertically elongated windows with short sill, characterized by horizontal segmentation in some cases and the absence of horizontal segmentation in others (Fig. 11). Fig. 11Best Responses Derived from the Median Geometrical Parameters of the Window in the Optimal Solutions for Each Room. Best Responses Derived from the Median Geometrical Parameters of the Window in the Optimal Solutions for Each Room. Subsequently, to examine the relationship between the geometrical parameters of the window and the energy and daylight performance of the space, Pearson correlation analysis was conducted between the window geometrical parameters of the Pareto frontier response set and the corresponding energy and daylight performance metrics (Table 12). According to the results, the room area, east-west dimension, and north-south dimension have the greatest impact on energy performance, including total energy consumption, cooling demand, and heating demand. However, the WWR and window area have the most significant influence on daylight performance, including the average UDI, average DA, and average DGP (Table 12). Table 12Results of pearson correlation test between window geometrical parameters and room energy Performance.Energy ConsumptionCooling DemandHeating DemandAverage DAAverage UDIAverage DGPRoom EW Side DimensionPearson Correlation− 0.821**− 0.809**− 0.810**− 0.088− 0.063− 0.064Sig. (2-tailed)0.0000.0000.0000.1670.3280.321Room NS Side DimensionPearson Correlation− 0.752**− 0.699**− 0.768**− 0.146*− 0.102− 0.070Sig. (2-tailed)0.0000.0000.0000.0220.1090.277Room AreaPearson Correlation− 0.973**− 0.933**− 0.976**− 0.147*− 0.101− 0.085Sig. (2-tailed)0.0000.0000.0000.0210.1130.186Window WidthPearson Correlation− 0.295**− 0.180**− 0.360**0.429**0.400**0.413**Sig. (2-tailed)0.0000.0050.0000.0000.0000.000Window HeightPearson Correlation− 0.0140.088− 0.0770.452**0.548**0.485**Sig. (2-tailed)0.8300.1690.2310.0000.0000.000Window AreaPearson Correlation− 0.322**− 0.149*− 0.422**0.692**0.672**0.693**Sig. (2-tailed)0.0000.0190.0000.0000.0000.000WWRPearson Correlation− 0.0400.143*− 0.153*0.746**0.691**0.751**Sig. (2-tailed)0.5350.0240.0170.0000.0000.000Number of RowsPearson Correlation− 0.056− 0.056− 0.055− 0.023− 0.065− 0.029Sig. (2-tailed)0.3820.3840.3920.7150.3100.653Number of ColumnsPearson Correlation0.0960.0300.134*− 0.090− 0.126*− 0.103Sig. (2-tailed)0.1350.6360.0360.1590.0490.106Vertical Distance_between CellsPearson Correlation0.343**0.383**0.310**0.345**0.250**0.337**Sig. (2-tailed)0.0000.0000.0000.0000.0000.000Horizonal Distance between CellsPearson Correlation0.030− 0.0230.062− 0.121− 0.034− 0.180**Sig. (2-tailed)0.6410.7190.3330.0580.5960.005Horizontal Window PositionPearson Correlation− 0.211**− 0.243**− 0.187**− 0.124− 0.104− 0.113Sig. (2-tailed)0.0010.0000.0030.0520.1030.076Vertical Window PositionPearson Correlation− 0.114− 0.115− 0.1110.0710.0090.085Sig. (2-tailed)0.0750.0720.0830.2700.8840.185N246246246246246246*. Correlation is significant at the 0.05 level (2-tailed).**. Correlation is significant at the 0.01 level (2-tailed). Table 12Results of pearson correlation test between window geometrical parameters and room energy Performance.Energy ConsumptionCooling DemandHeating DemandAverage DAAverage UDIAverage DGPRoom EW Side DimensionPearson Correlation− 0.821**− 0.809**− 0.810**− 0.088− 0.063− 0.064Sig. (2-tailed)0.0000.0000.0000.1670.3280.321Room NS Side DimensionPearson Correlation− 0.752**− 0.699**− 0.768**− 0.146*− 0.102− 0.070Sig. (2-tailed)0.0000.0000.0000.0220.1090.277Room AreaPearson Correlation− 0.973**− 0.933**− 0.976**− 0.147*− 0.101− 0.085Sig. (2-tailed)0.0000.0000.0000.0210.1130.186Window WidthPearson Correlation− 0.295**− 0.180**− 0.360**0.429**0.400**0.413**Sig. (2-tailed)0.0000.0050.0000.0000.0000.000Window HeightPearson Correlation− 0.0140.088− 0.0770.452**0.548**0.485**Sig. (2-tailed)0.8300.1690.2310.0000.0000.000Window AreaPearson Correlation− 0.322**− 0.149*− 0.422**0.692**0.672**0.693**Sig. (2-tailed)0.0000.0190.0000.0000.0000.000WWRPearson Correlation− 0.0400.143*− 0.153*0.746**0.691**0.751**Sig. (2-tailed)0.5350.0240.0170.0000.0000.000Number of RowsPearson Correlation− 0.056− 0.056− 0.055− 0.023− 0.065− 0.029Sig. (2-tailed)0.3820.3840.3920.7150.3100.653Number of ColumnsPearson Correlation0.0960.0300.134*− 0.090− 0.126*− 0.103Sig. (2-tailed)0.1350.6360.0360.1590.0490.106Vertical Distance_between CellsPearson Correlation0.343**0.383**0.310**0.345**0.250**0.337**Sig. (2-tailed)0.0000.0000.0000.0000.0000.000Horizonal Distance between CellsPearson Correlation0.030− 0.0230.062− 0.121− 0.034− 0.180**Sig. (2-tailed)0.6410.7190.3330.0580.5960.005Horizontal Window PositionPearson Correlation− 0.211**− 0.243**− 0.187**− 0.124− 0.104− 0.113Sig. (2-tailed)0.0010.0000.0030.0520.1030.076Vertical Window PositionPearson Correlation− 0.114− 0.115− 0.1110.0710.0090.085Sig. (2-tailed)0.0750.0720.0830.2700.8840.185N246246246246246246*. Correlation is significant at the 0.05 level (2-tailed).**. Correlation is significant at the 0.01 level (2-tailed). Results of pearson correlation test between window geometrical parameters and room energy Performance. *. Correlation is significant at the 0.05 level (2-tailed). **. Correlation is significant at the 0.01 level (2-tailed). Furthermore, to prioritize the geometrical parameters based on energy and daylight performance, linear regression analyses were conducted between each energy and daylight performance factor and the window geometrical parameters of the Pareto frontier response set (Table 13). According to the standardized beta values, for energy consumption, the east-west dimension, north-south dimension, room area, and WWR were found to be the most significant factors, in that order. Additionally, based on the regression findings, the north-south dimension, east-west dimension, WWR, and room area had the greatest impact on cooling demand, respectively. For heating demand, the east-west dimension, north-south dimension, room area, and window area had the most significant effects, in that order (Table 13). Table 13Standardized beta results of linear regression tests between window geometrical parameters and room energy Performance.Energy ConsumptionCooling DemandHeating DemandAverage DAAverage UDIAverage DGPEW Side Dimension− 0.788− 0.744− 0.7980.118− 0.3880.270NS Side Dimension− 0.751− 0.751− 0.734− 0.238− 0.597− 0.038Room Area0.2580.2430.262− 0.2630.392− 0.431Window Width0.0300.0600.0100.1860.2680.177Window Height0.0020.035− 0.0190.1700.3900.221Window Area− 0.111− 0.135− 0.0940.2560.3900.109WWR0.1760.3430.0680.4400.1160.523Number of Rows− 0.0030.002− 0.0070.0280.0200.018Number of Columns− 0.008− 0.014− 0.0050.071− 0.0180.095Vertical Distance between Cells− 0.025− 0.014− 0.0310.055− 0.0350.070Horizonal Distance between Cells0.0270.0100.036− 0.0500.090− 0.113Horizontal Position of the Window0.0180.0090.023− 0.0070.006− 0.014Vertical Position of the Window0.0190.0200.0180.0730.0510.102 Table 13Standardized beta results of linear regression tests between window geometrical parameters and room energy Performance.Energy ConsumptionCooling DemandHeating DemandAverage DAAverage UDIAverage DGPEW Side Dimension− 0.788− 0.744− 0.7980.118− 0.3880.270NS Side Dimension− 0.751− 0.751− 0.734− 0.238− 0.597− 0.038Room Area0.2580.2430.262− 0.2630.392− 0.431Window Width0.0300.0600.0100.1860.2680.177Window Height0.0020.035− 0.0190.1700.3900.221Window Area− 0.111− 0.135− 0.0940.2560.3900.109WWR0.1760.3430.0680.4400.1160.523Number of Rows− 0.0030.002− 0.0070.0280.0200.018Number of Columns− 0.008− 0.014− 0.0050.071− 0.0180.095Vertical Distance between Cells− 0.025− 0.014− 0.0310.055− 0.0350.070Horizonal Distance between Cells0.0270.0100.036− 0.0500.090− 0.113Horizontal Position of the Window0.0180.0090.023− 0.0070.006− 0.014Vertical Position of the Window0.0190.0200.0180.0730.0510.102 Standardized beta results of linear regression tests between window geometrical parameters and room energy Performance. Furthermore, based on the reported standardized beta coefficients, the WWR, room area, window area and the north-south dimension of the room, respectively, had the greatest influence on the average DA of the space. Similarly, the north-south dimension, room area, window height, window area, and east-west dimension had the most significant impact on the average UDI in the room. Discussion This study investigates the impact of window geometrical parameters on the energy and daylight performance of residential spaces in Yazd city. Previous research has extensively explored the relationship between window geometrical parameters and energy and daylight performance (Mebarki et al., 2021; Alwetaishi & Taki, 2020); however, no study has systematically examined window geometrical parameters across a diverse range of room types. Some studies have analyzed the influence of window parameters on energy and daylight performance in a simplified single space (Tan et al., 2020; Gasparella et al., 2011), while others have focused on multi-space buildings with full detail (Mehdizadeh-Rad et al., 2022). Given the significant influence of spatial geometry on energy and daylight outcomes, prescribing optimal window parameters for design necessitates a structured investigation across a variety of spaces a focus of the present research. The findings offer comprehensive insights into optimal window geometrical parameters based on an extensive analysis of typical interior spaces in residential units. The findings of this study offer clear and quantifiable evidence that optimizing the geometry of south-facing windows can substantially enhance both energy efficiency and daylight performance in residential buildings located in hot-arid climates such as Yazd. These results have significant potential to inform actionable policy recommendations and support the development of performance-based guidelines within Iran’s national building codes, which currently lack specificity in this domain. By translating the optimization outcomes such as the recommended 20% WWR, increased verticality, and strategic window placement into prescriptive or performance driven regulations, policymakers can promote more sustainable architectural practices. Furthermore, integrating such evidence-based parameters into national standards could facilitate the widespread adoption of passive design strategies, ultimately contributing to reduced energy consumption and environmental impact across Iran’s residential sector. The results reveal that changes in spatial geometry affect the optimal window geometrical parameters with respect to energy and daylight performance. Specifically, an increase in the luminous facade area corresponds to a decrease in the WWR based on energy and daylight criteria. The highest WWR values were observed in rooms with the greatest depth relative to the southern facade, whereas the lowest WWR values were associated with rooms having the largest southern facade area. Furthermore, the study shows that, in general, vertical elongation of windows improves energy and daylight performance; however, as room area or depth relative to the southern facade increases, window height must be proportionally increased. The optimal window height range corresponds to approximately 6.67% to 6.82% of the room height. Among the findings, WWR emerged as the most influential window geometrical parameter on energy and daylight performance, regardless of variations in room geometry. Other studies have similarly identified WWR as the key factor affecting energy and daylight outcomes but have not considered how spatial geometry changes impact the optimal WWR value [17; 21]. This research determined the optimal southern facade WWR for residential buildings in Yazd to range between 0.134 and 0.332. Correspondingly, in Alwetaishi and Benjeddou reaserch21 recommend WWR limits of 35%, 25%, and 20% for northwest, southeast, and southwest facades respectively in the hot climate of Taif, Saudi Arabia. Additionally, in Alwetaishi and Benjeddou reaserch, Alwetaishi and Benjeddou21 define suitable WWR values for optimal energy and daylight performance in various climates as 20% to 30%, aligning with the present study’s results. Conversely, Tawfeeq and Qaradaghi24 optimized WWR in Sulaymaniyah, Iraq, suggesting desirable values of 65% for southern and eastern facades and 95% and 30% for northern and western facades, respectively. Considering the climatic similarity between Sulaymaniyah and Yazd, these findings contrast with the results of the current research. While the hot-arid climates of Yazd and Sulaymaniyah suggest comparable environmental conditions, the discrepancy in optimal WWR between this study and the findings of Tawfeeq and Qaradaghi24 likely stems from a range of contextual and methodological differences beyond climate alone. Key contributing factors may include variations in building construction assumptions, such as insulation levels, thermal mass, glazing properties, and ventilation strategies, all of which influence the building’s thermal behavior and daylight responsiveness. Additionally, internal loads including occupancy patterns, equipment use, and lighting schedules can significantly affect energy demand profiles, thereby altering the optimization outcomes. Perhaps most critically, the divergence in optimization objectives whether prioritizing cooling load minimization, thermal comfort, DA, or a combination there of can lead to markedly different conclusions regarding ideal window dimensions. These distinctions highlight the importance of contextual specificity and the need for locally calibrated design guidelines when translating simulation-based findings into practice. Moreover, this study examined the optimal window placement on the southern facade in Yazd based on energy and daylight performance. This issue has only been addressed in one other study23, which did not extensively analyze the geometrical window parameters assessed here. According to our findings, positioning the window center below the midpoint of the facade height and at the horizontal midpoint of the window improves room energy and daylight performance. This conclusion holds despite variations in room geometrical parameters. The study also analyzed the number of rows and columns of window cells within the range of 1 to 10 and explored the variation in horizontal and vertical spacing between cells from 0 to 1 m. While some previous studies have considered window segmentation, their analyses were limited to small horizontal divisions22,25. Our results indicate that for optimal energy and daylight performance, the number of cell rows should be fixed at 1, while the number of columns may vary and should exceed 1. Additionally, vertical spacing between cells in most rooms ranged between 0.10 and 0.16 m. For practicing architects in Yazd and similar hot arid contexts, the study’s findings can be distilled into a practical and accessible design guideline that bridges the gap between simulation-based research and real-world application. A simplified “rule of thumb” emerging from the optimization results would be: for south-facing rooms, begin with a WWR of approximately 20%, adopt a vertically elongated window shape with a height exceeding 2 m, and position the sill at a height below 0.5 m, centered horizontally on the façade. These parameters not only enhance daylight availability but also yield measurable reductions in both cooling and heating demands. By integrating these recommendations into early-stage design decisions, architects can make informed choices that align with energy efficiency goals and passive design principles, without the need for complex simulation tools. Finally, the study found that variations in window geometrical parameters led to reductions in energy consumption by 3.9% to 5.2%, cooling demand by 19.7% to 23.2%, and heating demand by 2.3% to 4.2%. Sorooshnia et al.23 reported a 16.12% reduction in building energy consumption in Sydney, Australia, based on window parameter modifications. Azmy and Ashmawy20 demonstrated that in Cairo’s hot climate, a WWR exceeding 25% significantly influences energy performance. In general, based on the literature review, several structural limitations have been considered in the research methodology. The primary limitation in this study is the use of a single-room model. According to the findings of several studies26,27, the external form of a building has a significant influence on its energy and daylight performance. Moreover, as reported in other studies [28; 29], the spatial layout and dimensions of indoor spaces also have a considerable impact on the building’s energy and daylight performance. Therefore, given the wide variability of external forms and internal layouts, energy and daylight simulations were conducted for a single room to enable a comprehensive analysis of the effect of window geometrical parameters (Table 14). Additionally, the room dimensions were defined as a control variable and examined through eight optimization processes. Another limitation of this study is the focus on south-oriented daylighting. Typically, rooms in residential apartments in Yazd have windows only on one façade, and most apartment buildings include a south-facing façade to receive sunlight. Accordingly, optimizing the window parameters for south-facing orientations was of primary importance in this research, while other orientations were not examined due to the comprehensive nature of the study (Table 14). Finally, another methodological limitation is the exclusion of shading systems. Based on previous studies [22; 30], shading devices have a substantial effect on building energy and daylight performance. Therefore, to conduct a comprehensive assessment of the influence of window geometrical parameters, the simulations were performed without considering any shading system (Table 14). Table 14The limitations of the Reserach.Subject of the LimitationReasonsReferencesUse of a Single-Room ModelThe significant impact of the layout and dimensions of interior spaces on energy performance and daylightThe great influence of the building’s exterior form on energy performance and daylightFattahi Tabasi et al.26Mukkavaara & Sandberg27Zhang et al.28Rahbar et al.29Focus on South-Facing WindowsThe presence of windows on only one side of most rooms in residential apartments in Yazd cityThe presence of a southern front in most residential apartments in Yazd city for receiving sunlightCherier et al.16Exclusion of External Shading DevicesThe significant impact of shading devices on energy performance and daylightTan et al.22Zhang et al.30 Table 14The limitations of the Reserach.Subject of the LimitationReasonsReferencesUse of a Single-Room ModelThe significant impact of the layout and dimensions of interior spaces on energy performance and daylightThe great influence of the building’s exterior form on energy performance and daylightFattahi Tabasi et al.26Mukkavaara & Sandberg27Zhang et al.28Rahbar et al.29Focus on South-Facing WindowsThe presence of windows on only one side of most rooms in residential apartments in Yazd cityThe presence of a southern front in most residential apartments in Yazd city for receiving sunlightCherier et al.16Exclusion of External Shading DevicesThe significant impact of shading devices on energy performance and daylightTan et al.22Zhang et al.30 The limitations of the Reserach. The significant impact of the layout and dimensions of interior spaces on energy performance and daylight The great influence of the building’s exterior form on energy performance and daylight Fattahi Tabasi et al.26 Mukkavaara & Sandberg27 Zhang et al.28 Rahbar et al.29 The presence of windows on only one side of most rooms in residential apartments in Yazd city The presence of a southern front in most residential apartments in Yazd city for receiving sunlight Tan et al.22 Zhang et al.30 Future research should expand the investigation to include multi-zone models with internal heat transfer, explore the effects of different orientations and facade compositions, and assess the performance of integrated shading strategies. Additionally, incorporating occupant behavior, dynamic internal loads, and advanced glazing technologies would further enhance the applicability of the findings and support the development of more holistic, context-sensitive building design guidelines. A comparison between the findings of the present study and Chap. 19 of the Iranian National Building Regulations reveals that both sources emphasize the critical role of the building envelope, particularly windows, in optimizing energy performance. According to Chap. 19, the design of transparent elements must aim to reduce heat transfer, control solar radiation, and ensure adequate natural lighting, while the window U-value must be determined in accordance with the climatic conditions. This regulation mandates the use of double glazing, appropriate shading devices, and reasonable window-to-wall ratios to prevent excessive cooling and heating loads. Similarly, the present study demonstrates that reducing the window-to-wall ratio and selecting optimal window dimensions for the hot and dry climate of Yazd lead to a significant decrease in cooling demand and an improvement in annual energy performance. Therefore, the findings of this research are consistent with the principles of Chap. 19, which aims to reduce thermal losses and manage energy loads through proper design of transparent façade elements. Regarding daylighting, Sects. 19 − 5 of the National Building Regulations highlights the importance of utilizing natural light during daytime hours to reduce electricity consumption and provide visual comfort. The results of the present study indicate that the window-to-wall ratio and window area exert the greatest influence on daylighting indicators, including Useful Daylight Illuminance (UDI), Daylight Autonomy (DA), and Daylight Glare Probability (DGP). These findings quantitatively refine and extend the qualitative objectives of Chap. 19, providing climate-specific guidance for Yazd: optimizing the use of natural light while avoiding excessive cooling loads. Consequently, the present research can serve as a foundation for revising or more precisely localizing the design criteria for transparent elements in Chap. 19, especially for hot and dry regions of Iran. However, the National Building Regulations do not specify any optimal geometrical parameters for windows. Aside from general guidance on window-to-wall ratio, the regulations do not address how individual geometrical parameters of windows relate to building energy performance or daylighting performance. Given the substantial impact of transparent façade elements on both energy and daylighting outcomes, this gap highlights the need for fundamental updates in the national regulations. The findings of this study can therefore provide a scientific framework for revising and improving these regulatory standards. Conclusion Based on the findings, it can be concluded that the geometrical parameters of south-facing windows have a significant impact on both the energy consumption and daylighting performance of residential buildings in Yazd. The results indicate that the optimized window geometries led to reductions of up to 2.5% in total energy consumption, 23.2% in cooling demand, and 2.4% in heating demand. Implementing these optimized window configurations in the façades of residential buildings, through the formulation of a comprehensive set of design and construction guidelines, can substantially reduce energy use while improving daylighting performance. This approach can therefore contribute to environmental sustainability and sustainable urban development at a broader scale. The results further reveal that the optimal geometrical parameters of windows in terms of energy and daylighting performance are highly dependent on room depth and floor area. Hence, predictive modeling can be employed to establish their relationships and integrate these as quantitative indices within future building codes and façade design regulations. The study found that the optimal window-to-wall ratio (WWR) for south-facing façades in Yazd’s residential buildings ranges between 13.4% and 33.2%, with an optimal window height corresponding to 67.6–82.6% of the room height. The optimal configuration typically consists of vertically elongated windows with a single row and multiple horizontal divisions. Consequently, windows with a vertical proportion, height greater than 2 m, short sill height, and WWR below 30%, divided into horizontal segments, provide the most efficient performance for Yazd’s hot-arid climate. Notably, this configuration aligns closely with the traditional window typologies of Yazd, such as the “Se-dari” and “Panj-dari” layouts (Fig. 12). Fig. 12The impact of window geometric characteristics on energy performance and daylighting. The impact of window geometric characteristics on energy performance and daylighting. Overall, among all window geometry parameters, the WWR, window area, and window height have the greatest influence on building energy performance, particularly in optimizing cooling and heating demands and UDI.