Open AccessArticle

A Study of the Impacts of Different Opening Arrangements of Double-Skin Façades on the Indoor Temperatures of a Selected Building

Qing Sun

^1,2,

Junwei Song

^1,*,

Ying Yu

¹,

Hongbo Ai

² and

Long Zhao

College of Electromechanical Engineering, Xi’an University of Architecture and Technology, Xi’an 710055, China

Architectural Design and Research Institute, Xi’an University of Architecture and Technology, Xi’an 710055, China

Author to whom correspondence should be addressed.

Buildings 2024, 14(12), 3893; https://doi.org/10.3390/buildings14123893

Submission received: 9 October 2024 / Revised: 25 November 2024 / Accepted: 2 December 2024 / Published: 5 December 2024

(This article belongs to the Section Building Energy, Physics, Environment, and Systems)

Download

Browse Figures

Versions Notes

Abstract

The aim of this study is to evaluate the indoor temperature of a double-skin façades (DSF) high-rise building in Xi’an under different window opening arrangements, and to assess their impact on the operating time of the air-conditioning system. Compared to conventional buildings, double-skin façade (DSF) buildings can reduce energy consumption. While current research trends focus primarily on heat transfer and materials, there is limited exploration of window opening arrangements. To address this gap, VENT engineering software 2018 was used to simulate indoor temperatures at various window opening angles and determine the optimal arrangement. Additionally, the extreme random tree (ET) algorithm was employed to develop a model for indoor temperature prediction. Climate data were sourced from an online database and processed using the Spearman correlation coefficient method. Window opening arrangements were designed using orthogonal tests, and the performance of the DSF was evaluated with computational fluid dynamics (CFD) software (Fluent) 2023R1. An analysis of temperature variation in the double-skin façade (DSF) curtain wall revealed that the ET algorithm predicted indoor temperatures with 93% accuracy at a 50° window opening angle. Optimal window opening arrangement 2 resulted in a 2.7% reduction in the average interior temperature, a 3.6% reduction at a height of 1.2 m, and a decrease in air-conditioning runtime by 1.33 h. The extreme random tree (ET) algorithm was found to be more accurate than other methods in predicting DSF performance. These findings provide insights for optimizing the control and application of double-skin façades and suggest potential synergies with other systems.

Keywords:

DSF; window-opening arrangements; extreme trees classifier; CFD

1. Introduction

Building energy consumption represents a significant share of total energy use: 36.3% in China [1], 40% in the United States [2], and 40–50% of global carbon emissions [3]. Although the application of double-skin façades (DSFs) is increasing, their daily operation is still limited to basic opening and closing. The integration of DSFs with other systems, particularly window-opening arrangements, remains suboptimal. A well-designed window-opening arrangement is essential for maximizing the energy-saving potential of DSFs in buildings. However, research on the optimal window-opening arrangement for DSF systems is limited, and exploring the various control factors of these systems is of considerable importance.

A DSF system consists of two glass panels that create an air gap or buffer zone between the inner and outer layers. The space may be designed to be ventilated, sealed, or movable, serving to regulate air circulation and provide thermal and acoustic insulation. DSF systems are widely employed in high-rise buildings to enhance energy efficiency, optimize indoor environmental quality, facilitate natural ventilation and lighting, and improve architectural aesthetics. DSF systems include ventilated, sealed, movable, and photovoltaic curtain walls, which incorporate photovoltaic materials. DSF systems are categorized based on the method of window operation, as illustrated in Figure 1.

Studies have shown that naturally ventilated buildings with DSF structures can reduce energy consumption by approximately 40% compared to air-conditioned buildings [4]. Zheng et al. investigated a multilayer DSF system in Xiamen, which effectively lowered indoor temperatures by 6.9 °C compared to a conventional single-glazed wall, resulting in a 49.76% energy saving in summer [5]. Duan et al. analyzed factors such as DSF shape using EnergyPlus 21.0 and found that a serrated DSF outperformed a traditional flat DSF, with energy savings of 33.04% for cooling and 28.44% for heating, resulting in an annual energy saving of 31.11% [6]. In terms of material innovation, some scholars have discussed the use of bio-based materials in DSFs. Lahayrech S. et al. [7] used recycled materials and bio-based polyester resins to manufacture tiles for use in DSF systems, where they served as thermal insulation with positive results.

Yoon proposed an operational arrangement that integrates active cooling (AC) and passive cooling (PC) modes, as well as PC and hybrid heating (HH) modes, based on the DSF system. This arrangement was investigated through building simulation modeling, which showed that it can save 28% of cooling energy and 98% of thermal energy compared to a single-layer curtain wall system [8]. Joe et al. proposed an arrangement that integrates a Heating, Ventilation, and Air Conditioning (HVAC) system with DSF temperature control. This system maintains the indoor temperature within a specified range and triggers HVAC cooling when the temperature exceeds a critical value, resulting in a 3.4% reduction in energy consumption [9]. Lin et al. proposed a window-opening arrangement for a naturally ventilated louvered DSF system, where the louvers’ angle in the DSF cavity is adjusted based on window positioning. Their results showed that louvered DSF could reduce the daily cooling load by up to 28.4% compared to conventional DSF systems [10]. Choi et al. set temperature intervals to determine the window-opening arrangement of the DSF. The windows were opened when the temperature exceeded 21 °C and closed when the room temperature exceeded 26 °C to prevent overheating of the interior space [11]. He et al. investigated the thermal performance of DSF systems in Hangzhou under the climatic conditions of hot summers and cold winters. Compared to single-glazed curtain walls, DSFs reduced the cooling load in summer and the heating load in winter, except when the sun was present in winter. In summer, single-glazed curtain walls are more energy-efficient because they reduce the transmission rate of solar radiation, regardless of cavity ventilation [12]. Li Xue [13] compared the effects of sawtooth and flat DSF shapes, as well as the influence of window positioning on system performance. In response to the fire risks associated with DSF systems, which have not been addressed in this paper, Hang Dong Qi [14] investigated the potential risks of naturally ventilated DSF and provided recommendations. Mahboube [4] improved DSF supporting components and used electrochromic glass to study the annual electricity consumption of a building with a DSF system in a hot, humid area, reducing it by 3.5%. Therefore, there are numerous research directions for DSFs, all of which hold practical value.

Given the limited studies on the window arrangement of DSF, this paper focuses on comparing the effectiveness of DSF with different window opening angles in controlling indoor temperature, aiming to provide a reference for the practical application of DSF systems. This paper proposes a window arrangement scheme for an independent DSF system. Computational fluid dynamics (CFD) simulations and other methods are used to analyze the effects of DSF systems on indoor temperature at different window opening angles, determine the optimal opening angle, and construct an indoor temperature prediction model using the ET algorithm. The relationship between different window opening angles and indoor temperature is derived using VENT engineering. Orthogonal tests were performed to develop a feasible window arrangement, and the effects on DSF zones, indoor temperature, and air conditioning runtime were calculated using CFD methods. The study of window arrangements provides new insights and valuable references for DSF applications.

2. Study Subject

2.1. Overview of the Research Subjects

The DSF structure consists of a two-layer glass envelope, with an inner and outer glass layer separated by a ventilated cavity. This structure serves the dual function of thermal insulation and cooling in summer, and insulation in winter. This study focuses on the west room on the 14th floor of the training building at Xi’an University of Architecture and Technology (XAUAT), China, which is surrounded by space outside the DSF. The study targets the space with a sawtooth DSF in the middle of the west façade, with a glass thickness of 0.024 m. The area shown in Figure 2a is insulated along all boundaries except for the DSF area, and only the effect of the DSF on the space within the study area is investigated. The curtain wall is equipped with center-hung and top-hung casement windows on both the inner and outer façades, adjustable via servo-motor controlled angles. The walls and insulating glass of the serrated façade in the DSF are staggered components. The DSF structure comprises eight identical sets of serrated façades, as shown in Figure 2.

2.2. Meteorological Parameters

The DSF system in this study is located at 108.93° E, 34.3° N. Meteorological data were obtained from the Xi’an Meteorological Service. The average annual temperature is 13.5 °C, with the hottest month averaging 26.5 °C and reaching a maximum of 38 °C. In contrast, the average temperature of the coldest month is −10 °C, indicating a hot summer and cold winter climate, as shown in Figure 3 [15]. In this study, indoor temperature prediction simulations were performed using weather data from May to July. The data at 14:00 on May 10 is used to set the CFD simulation parameters for the DSF system, with wind speed replaced by the DSF inlet speed. Figure 3 shows the annual temperature variation curve for Xi’an.

3. Research Methodology

3.1. Indoor Temperature Prediction Model

The indoor temperature prediction model consists of two components: indoor temperature prediction and the extreme random tree (ET) model. Specifically, different angles and corresponding indoor temperatures, derived from VENT engineering simulations under the same weather parameters, are input into the ET algorithm as training data. Seventy percent of the data is used for algorithm learning and training, while the remaining thirty percent is used for model validation. This allows the prediction model to estimate indoor temperatures using meteorological data, thereby establishing the temperature prediction model. The ET-based indoor temperature prediction model can be used to evaluate the performance of the DSF system and predict indoor temperatures based on weather parameters. Figure 4 illustrates the flowchart of the main methodology in this study.

3.1.1. Weather Factor Screening

The Spearman correlation coefficient is a nonparametric statistical method used to assess the monotonic relationship between two variables. It does not rely on data distribution but instead uses the ranks (or order) of the data, thus eliminating the need for covariance treatment in the correlation analysis. In this study, several parameters potentially impacting the indoor environment were selected, and the Spearman correlation coefficient was applied to identify key factors influencing indoor temperatures for the first time. These factors include outdoor temperature (t_out_door), dew point temperature (dew_point_t), scattered radiation (scatter_r), direct radiation (direct_r), sky temperature (t_sky), ground temperature (t_ground), wind speed (wind_speed), relative humidity (relative_humidity), atmospheric pressure (atmospheric_p), window angle (in_door), and dry bulb temperature (dry_bulb).

3.1.2. Simulation of Room Temperature at Different DSF Angles

VENT Engineering is a commercial software developed by Greenbuild Swell, built on the AutoCAD platform. Greenbuild Swell software 2017 can simulate various building parameters, such as cooling and heating loads, energy consumption, lighting, ventilation, and acoustic environments, with detailed results. Related teaching materials, including books and videos, have also been published [16]. VENT Engineering uses the second-order upwind scheme to discretize the equations. This scheme satisfies the accuracy requirements for general fluid simulation calculations [17] and complies with the simulation algorithm standards outlined in the “(JGJ/T 309-2013) Building Ventilation Effects Testing and Evaluation Standards” [18,19].

The simulation process in VENT Engineering involves the following steps: first, project setup is performed, followed by the calculation of indoor and outdoor wind pressures, which are then applied to the windows and doors. Next, in accordance with the Design Code for Heating, Ventilation, and Air Conditioning of Civil Buildings [20], the air change rate was set to 0.5 air changes per hour, and the air conditioner was disabled during the simulation. This setup was used to model indoor temperatures from May to August and to calculate the average indoor temperatures at different window opening angles.

The simulation using VENT Engineering software 2018 involves the following steps:

First, establish the room structure based on the room parameters and create the geometric model of the interior space, defining the dimensions of the room, walls, windows, and other architectural elements;
Set up the project in the software and configure the environmental conditions, including external factors such as temperature, humidity, wind speed, and boundary conditions;
Select the radiation model as the calculation method;
Calculate the wind pressure for both indoor and outdoor environments;
Apply the calculated wind pressures to the doors and windows;
Set the air exchange rate to 0.5 air changes per hour, as specified in the Design Code for Heating, Ventilation, and Air Conditioning of Civil Buildings [16], while turning off the air-conditioning system throughout the simulation;
Based on the window type and opening angle, adjust the window sash to four positions: 0°, 20°, 50°, and 70°, which can be calculated individually;
Import the weather data into the Thermal Comfort software 2018 and define the room’s ventilation rate;
Perform the simulation and export the data;
Obtain room temperatures for different window opening angles under the same climate conditions.

3.1.3. ET Algorithm

The ET algorithm, similar to Random Forest (RF) [21], consists of multiple de-correlated trees. Unlike traditional machine learning methods, ET does not require data preprocessing and can generate valid results for large datasets in a short time [22]. The application of extreme random trees can provide decision support based on real data and models, helping the DSF system select the optimal window opening angle and improve indoor temperature prediction accuracy. A comparison of common machine learning methods is shown in Table 1.

The algorithm is implemented as follows: Given original dataset

S

with

N

samples and

M

features, a decision tree is constructed using the training dataset. The attribute with the lowest Gain_GINI value, along with its corresponding attribute value, is selected as the optimal split attribute and splitting value, respectively. The Gain_GINI value is calculated using the following formula [22]:

G I N I (S) = 1 - \sum P_{k}^{2}

(1)

Here,

P_{k}

represents the proportion of occurrences of the

k

th category in the classification results across all samples.

The generated tree is pruned with the validation dataset and the optimal subtree is selected. In the process of pruning, the loss function is calculated [21].

c_{α} (T) = C (T) + α | Τ |

(2)

Here,

α \geq 0, C (T)

represents the prediction error of the training data, and

| Τ |

denotes the complexity of the model. During each of the

k

iterations,

k

decision trees are generated as extreme random trees. These trees are then used to make predictions on test samples, and the final classification result is determined by aggregating the predictions of all base classifiers using a voting method. Figure 5 illustrates the flow of the ET algorithm to build the model.

The indoor temperatures at different window opening angles from the VENT Engineering simulation, along with the associated weather parameters, are exported as raw data. Seventy percent of the raw data is used for training the ET algorithm, while the remaining thirty percent is used for validation.

3.2. Modeling of Research Object

Converting the window area into a flow area for simulation maximizes the actual effect and improves simulation accuracy. Therefore, the outer and inner glazing areas of the double-skin façades are used in place of the window-opening area for different window-opening arrangements. The calculation formula and parameters for the gas flow area at different window opening angles are provided in Formula (3), and Table 2 and Table 3.

S = W * H * F C

(3)

where

W

represents the sash width in meters (m),

H

denotes the sash height in meters (m), and

F C

stands for the flow coefficient.

The modeling area is shown in Figure 6. In Figure 6b, the exterior is the glass curtain wall, and the pink and dark yellow areas represent the DSF windows. The yellow plane represents the internal glazed curtain wall, and the white areas on the plane represent the windows. A polyhedral mesh is used for the simulations due to its fewer cells, faster computational convergence, higher mesh quality, and clear advantages in fluid simulation. According to reference [23], the research object is the same, and only part of the structure is simplified differently, with a strong reference. The grid size of 0.05 m and a total of 3,260,000 grids are selected to ensure grid independence, calculation accuracy, and reduced computational time. The cavity interior is kept open to promote natural airflow.

Given the complexity of heat flow within the DSF, the analysis is simplified under the following assumptions:

(1): Secondary structures, such as columns, are excluded;
(2): Internal heat sources and air-conditioning effects are disregarded;
(3): The thermal and optical properties of all materials are assumed to remain constant;
(4): Heat flow through edges and seams is neglected;
(5): All curtain wall parameters conform to the Chinese National Standard for Lighting Design of Buildings [24];
(6): All boundaries of the study area are insulated, except for the DSF boundary, and only the DSF curtain wall is subjected to external influences.

Detailed information on the DSF curtain wall arrangement is presented in Table 4, and the overall model is illustrated in Figure 6.

3.3. CFD Simulation

CFD is currently the primary tool for solving engineering analysis problems in building and HVAC design [25]. Several approaches are employed in DSF studies, including measurement methods, CFD simulations, energy simulation software, and area modeling. Measurement methods provide high accuracy and detailed information but require a physical building and substantial financial and human resources for real-time monitoring, including costly equipment. Energy simulation software can accurately predict overall energy consumption but may produce errors when modeling airflow in complex geometries [26]. Region modeling cannot provide real-time temperature and velocity data, whereas CFD simulation divides the fluid domain into discrete meshes and solves the governing equations to calculate integrated results. CFD simulations can accurately model gas flow, calculate heat transfer in complex environments, and monitor temperature and velocity variations in real-time, providing reliable hydrodynamic calculations [27]. Therefore, Fluent software was selected for CFD simulations in this study.

The experimental simulation model was created using Unigraphics NX, with meteorological parameters selected for 14:00 on 10 May 2023. The ET model predicted the indoor temperature as 26 °C. At that time, the meteorological data showed an outdoor temperature of 26.4 °C, and the average value of 26.2 °C was selected as the air temperature for the simulation. Simulation experiments were conducted using these air temperatures to model airflow temperatures during the time period from 14:00 to 15:00.

In this study, the airflow in the DSF and the indoor cavity is more complex and irregular, so the RNG (Re-normalization Group) model was chosen. The DO model (Discrete Coordinate Radiation Model) is selected for heat transfer, and parameters such as longitude, latitude, date, and time of the DSF location are input into the solar calculator. The solar ray tracing method was employed to calculate solar radiation. The specific boundary conditions for the simulation are provided in Table 5.

4. Window-Opening Arrangements

Due to the presence of eight identical DSF sawtooth structures and a total of 32 windows, the number of possible combinations for a single arrangement is large. To simplify the arrangement process, a single DSF sawtooth structure was defined as the arrangement criterion. The window-opening arrangement was developed using orthogonal tests and applied uniformly to all eight DSF sawtooth structures.

The orthogonal test method was employed to select representative tests from numerous combinations, reducing both experimental and computational workload, and thereby improving the research outcomes [28]. The simplified structure is shown in Figure 1b, with the cavity at the center, windows No. 2 and No. 3 as upper windows, and windows No. 1 and No. 4 as lower windows. The orthogonal table was created and screened through orthogonal tests to identify eight combinations of window openings, each corresponding to a different window-opening arrangement.

The window-opening arrangements were simplified through orthogonal experiments and are detailed in Table 6. These arrangements provide the foundation for evaluating the performance of the DSF under identical weather conditions in the following sections.

Following the orthogonal tests, window-opening arrangements were formulated based on various window-opening combinations identified in the tests. These arrangements represent a broader range of combinations and guide further analysis of DSF performance under varying conditions.

5. Parametric Analysis

5.1. Screening of Meteorological Parameters

The screening results are shown in Figure 7 and Figure 8. Among these factors, atmospheric pressure is negatively correlated with indoor temperature. Since atmospheric pressure remained relatively stable, its effect on indoor temperature variations was not considered in the subsequent analysis. Relative humidity (relative_humidity), atmospheric pressure (atmospheric_p), and their associated factors exhibited weak correlations with each other. The correlation coefficients were predominantly negative, indicating negligible correlations, and were thus excluded from further analysis. Increases in wind speed and variations in solar radiation were both correlated with indoor temperature at different window opening angles. Strong correlations were observed between dry bulb temperature, sky temperature, and surface temperature, indicating a strong relationship between these factors: T_out_door, dew_point_t, scatter_r, direct_r, t-sky, t-ground, wind_speed, dry_bulb_t, window opening angle, and indoor temperature variations. Wind speed, dry bulb temperature, and window opening angle exhibited strong or moderate correlations with indoor temperature variations and were therefore considered key influencing factors in this study. Figure 8 presents the nine parameters with the strongest correlations, with larger correlation coefficients indicating stronger interactions between the variables. These nine weather parameters were ultimately chosen as the experimental parameters.

5.2. Room Temperature at Different Opening Angles

Figure 9 illustrates the variation in indoor temperature at different window opening angles from May to August. The average indoor temperatures from May to August, with the DSF set to 0°, 20°, 50°, and 70° opening angles, were 30.78 °C, 25.02 °C, 24.97 °C, and 26.45 °C, respectively.

To illustrate the variation in indoor temperature at different opening angles, data from the week of 1–7 August 2023, were selected. This week was chosen due to the high temperatures and significant temperature differences observed at various opening angles, which facilitated a clear visualization of these variations. Figure 10 illustrates the effect of different DSF opening angles on indoor temperature during a typical summer week. Based on Figure 10, it can be concluded that a 50° opening angle is the most effective in comparison to 0°, 20°, and 70°.

When the opening angle is 0°, the indoor temperature rises to 30 °C. This is because solar radiation enters the room through the façade without ventilation, preventing heat from escaping. When the opening angle is 20° or 50°, a decrease in indoor temperature is observed. This is because opening the window at an appropriate angle allows outdoor air to enter the ventilation chamber of the curtain wall, removing some of the heat and lowering the indoor temperature. In this study, the research object is subject to thermal insulation boundaries, except for the DSF, with a window-to-wall ratio of 25%. When the window-to-wall ratio exceeds 25%, larger window areas or higher ratios often result in significant fluctuations in indoor temperature, particularly in regions with large seasonal variations. Therefore, these factors must be fully considered during the application and performance evaluation of the DSF. When the opening angle is 70°, the indoor temperature increases. This occurs because, when the window is widely opened, outdoor air enters the room directly and cannot dissipate heat effectively through the cavity. Meanwhile, solar radiation enters the room directly, leading to an increased accumulation of heat.

Based on the graphs of indoor temperature at different opening angles, the optimal window opening angle for reducing indoor temperature is 50°. By combining the indoor prediction model with VENT Engineering’s indoor temperature simulation data, it can be concluded that, in order to reduce indoor temperature during the simulation period, the maximum thermal comfort temperature at a window opening angle of 50° is 27.8 °C. A temperature of 27.8 °C means that, during the simulation period, if the indoor temperature exceeds 27.8 °C, the DSF will not be able to control the room temperature below the maximum thermal comfort level, and air conditioning will need to be activated to cool the room. Based on the indoor temperature data for different opening angles, a 50° opening angle was found to be the most effective in reducing indoor temperature. Therefore, 50° was selected as the opening angle for the DSF in the CFD simulation experiments in this study.

The above steps serve as a general guide; specific procedures may vary depending on the software and simulation requirements. During the experiment, the wind direction was set perpendicular to the DSF window surface. According to the flow coefficient table, the coefficient is 0.33 for a 20° opening angle, 0.60 for a 60° opening angle, and 0.62 for a 90° opening angle. In this study, the flow coefficients are approximated as follows: 0.33 at 20°, 0.60 at 60°, and 0.62 at 90°, as the actual model can only be set to 0°, 20°, 50°, or 70° in practical applications. As linear angle adjustment is not possible, the simulation process was simplified by selecting only 0°, 20°, 50°, and 70° opening angles.

5.3. Indoor Temperature Prediction

Using the data in Figure 8, along with the corresponding indoor temperature data, the ET model was applied to identify the optimal opening angles for different arrangements. The accuracy of the different opening angles was evaluated, and the results show that the ET model demonstrated high accuracy in the experimental setup. Figure 11 illustrates the ET algorithm’s precise predictive capability for indoor temperature.

With 70% of the data used for training, the model achieves an accuracy of 96.3% when the opening angle is set to 0°. When the opening angle was set to 20°, the model achieved a prediction accuracy of 92.9%. When the opening angle was set to 50°, the model achieved a prediction accuracy of 95.8%. As the opening angle increased to 70°, the model’s prediction accuracy slightly decreased to 89.7%. This decrease can be attributed to the fact that, with fully open ventilation openings, outdoor air enters the indoor space directly without passing through the ventilated cavity.

5.4. Effect of Window Opening Arrangements on Indoor Horizontal Surface Temperatures

Figure 12 shows the temperature variation at a height of 1.2 m for different window-opening arrangements. All arrangements show a gradual increase in temperature. The peak temperature for window-opening arrangement 3 is 31.6 °C. The peak temperature for arrangement 2 is 26.74 °C, which is 2.7% lower than the 27.48 °C observed in arrangement 1. To characterize the effects of the different window-opening arrangements, the following section describes the velocity and temperature distributions for arrangements 1 and 2 at the 1.2 m plane of the room.

Figure 13 and Figure 14 show the temperature and velocity distributions (in °C) in the horizontal direction for window-opening arrangements 1 and 2. Both arrangements effectively reduce the indoor temperature. In window-opening arrangement 2, the temperature is higher near the exterior curtain wall and the metal plate. In window-opening arrangement 1, the average indoor temperature at a height of 1.2 m is higher than in arrangement 2. However, in arrangement 2, the indoor temperature near the interior curtain wall is excessively high. The velocity distribution shows that window-opening arrangement 1 provides adequate indoor ventilation, whereas arrangement 2 isolates outdoor heat and enhances thermal convection in the curtain wall cavity. Due to the orientation of the DSF windows, the airflow in arrangement 1 tends to move in the northeast direction.

5.5. Effect of Window-Opening Arrangements on Internal Curtain Wall Temperatures

Figure 15 shows the temperature profiles of the internal curtain wall for the eight DSF window-opening arrangements. Arrangements 7 and 8 had the fastest increase in exterior curtain wall temperature, reaching 44.7 °C and 44.9 °C, respectively. At 3:00 p.m., the internal curtain wall temperatures were 27.4 °C for arrangement 1, 34.2 °C for arrangement 2, and 44.9 °C for arrangement 8. The temperature in window-opening arrangement 2 was 24.8% higher than in arrangement 1 and 23.8% lower than in arrangement 8. Figure 16 shows the distribution of internal curtain wall temperatures for window-opening arrangements 1, 2, and 8.

Figure 16 shows the temperature distribution (in °C) of the internal curtain wall for window-opening arrangements 1, 2, and 8. Window-opening arrangement 1 shows a uniform temperature distribution, with a maximum temperature of 38 °C at the lower end. Window-opening arrangement 2 exhibits a maximum temperature of 85 °C at the upper end of the interior wall. Window-opening arrangement 8 has a maximum temperature of 63 °C.

At the end of the simulation, the highest temperature of the inner curtain wall and the lowest average room temperature occurred in window arrangement 2. The specific reasons are as follows: in window arrangement 2, the opening of the external window, without the protection of the external curtain wall, allows more direct radiation to reach the inner glass curtain wall, contributing to the temperature increase. Secondly, due to the closure of the internal curtain wall, convective heat transfer can only occur significantly with the air in the cavity. The reduced convective heat transfer on the interior surface of the internal curtain wall further exacerbates the temperature increase. However, the increase in the temperature of the curtain wall also indicates that heat entering the room is isolated, preventing external hot air and radiation from directly entering the room. As a result, the indoor air is not involved in reducing the temperature of the DSF cavity.

5.6. Influence of Average Indoor Temperature on Window-Opening Arrangements

Figure 17 shows the variation in average indoor temperature for different window-opening arrangements. Window-opening arrangement 3 exhibits the highest average indoor temperature while window-opening arrangement 2 exhibits the lowest. The average indoor temperature of window-opening arrangement 2 was 3.6%, 16.9%, 12.1%, 9.8%, 13.8%, 16.3%, and 14.5% lower than those of window-opening arrangements 1, 3, 4, 5, 6, 7, and 8, respectively.

5.7. Impact on Air Conditioning Operating Hours

Window-opening arrangement 2 demonstrates better overall performance than the other arrangements. Therefore, the CFD simulation time for this arrangement was extended to obtain the average indoor temperature profile between 15:00 and 17:00, as shown in Figure 18.

At 16:20, the DSF opening angle was set to 50°. At this time, the temperatures estimated by the meteorological data and the prediction model were 28.1 °C and 28 °C, respectively, confirming the reliability of the prediction model. In window-opening arrangement 2, the room temperature reaches the maximum comfort threshold of 27.8 °C at 16:20. Window-opening arrangement 1 reaches the critical threshold at 15:07, with the air conditioning system starting 1.22 h later than in window-opening arrangement 2. Window-opening arrangement 8, with the DSF fully closed, reaches the comfort threshold at 15:00, 1.33 h earlier than window-opening arrangement 2.

6. Discussion

Regarding the impact of different DSF opening angles on room temperature, among 0°, 20°, 50°, and 70°, the 50° opening angle yields the most optimal effect. Different opening angles significantly affect room temperature, though the relationship is nonlinear.

In terms of the indoor temperature prediction model, the prediction accuracies are 96.3%, 92.9%, 95.8%, and 89.7% for window angles of 0°, 20°, 50°, and 70°, respectively, resulting in an average accuracy of 93.67%. The algorithm requires extensive mathematical training to enhance accuracy and will need real indoor temperature data for future validation.

Impact on indoor temperature and DSF interior wall temperature: window arrangement 2 was most effective in reducing indoor temperatures during the hotter hours. The results for window arrangements 7 and 8 indicate that the condition of the interior curtain wall openings has minimal impact on indoor temperatures when the exterior curtain wall windows are closed.

Due to the numerous factors influencing the DSF system, varying parameter settings for its components can affect the system’s actual performance. The results of this study support practical applications and provide data for the use of curtain walls in regions with comparable climates and architectural structures. These curtain walls can be adjusted to an optimal angle to regulate indoor temperature through varying window openings.

7. Conclusions

7.1. Major Results

This paper investigates the performance of a high-rise building in Xi’an with a double-skin façade structure and varying window openings. VENT Engineering software 2018 is used to predict the indoor temperature for different window opening angles of the double-skin curtain wall. An indoor temperature prediction model is developed using the Extreme Stochastic Tree (ET) algorithm. Additionally, computational fluid dynamics (CFD) software Fluent 2023R1 is employed to assess the performance of the double skin façades, determine the optimal window opening arrangement, and evaluate its impact on air-conditioning operation time.

The average indoor temperature was lowest at the optimal opening angle of 50° among the four DSF opening angles studied.
The ET-based indoor temperature prediction model achieved an accuracy of 93.67%.
Compared with other window-opening arrangements, window-opening arrangement 2, with a 50° angle, reduces the overall average indoor temperature by 3.6%, 16.9%, 12.1%, 9.8%, 13.8%, 16.3%, and 14.5%, respectively. Different window-opening arrangements and angles generate distinct indoor temperature trends and should therefore be selected based on seasonal and climatic conditions.
Window-opening arrangement 2 outperforms others in terms of average indoor temperature at a height of 1.2 m, being 2.7% lower than the 27.48 °C of window-opening arrangement 1.
The average temperature of the DSF interior façade was 19.9% and 38.9% lower under window-opening arrangement 1 compared to arrangements 2 and 8, respectively.
With window-opening option 2 at the optimal opening angle, the air-conditioning system start-up is delayed by 1.22 h and 1.33 h, respectively, compared to options 8 (system fully closed) and 1 (system fully open).

7.2. General Conclusions

To optimize the performance of double-skin façade (DSF) systems, it is essential to carefully control the building’s window-to-wall ratio to prevent excessive glazing. In addition, strengthening the thermal insulation and heat retention properties of the building’s exterior walls, as well as improving airtightness, are critical. The orientation and structural parameters of the DSF should be designed according to the specific conditions of the building’s location to maximize its effectiveness. The optimal opening angle of the DSF is around 50°, indicating that DSF performance is not linearly dependent on the opening angle. The use of the Extreme Stochastic Tree (ET) algorithm allows for reliable prediction of indoor temperatures, in conjunction with weather parameters, aiding in decisions regarding the opening and closing of the DSF system. For buildings in similar climates, window-opening arrangement 2 can be optimized for effective indoor temperature control.

This study has some limitations. It mainly relies on simulations and lacks validation with real experimental data. Additionally, the relationship between the DSF opening angle and indoor temperature is not fully explored. The algorithm model requires further refinement, and the idealized conditions assumed in the CFD simulation may not fully reflect real-world scenarios. Furthermore, the study does not consider a wide range of DSF opening methods.

Author Contributions

Methodology, Q.S.; Validation, H.A.; Writing—original draft, J.S.; Writing—review & editing, Y.Y.; Supervision, L.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Shaanxi Provincial Key R&D Programme (2023-YBSF-381).

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

The authors declare no conflicts of interest.

References

Liu, X.; Shen, C.; Wang, J.; Zhang, C.; Shuai, Y. Static and dynamic regulations of photovoltaic double skin facades towards building sustainability: A review. Renew. Sustain. Energy Rev. 2023, 183, 113458. [Google Scholar] [CrossRef]
Allouhi, A.; El Fouih, Y.; Kousksou, T.; Jamil, A.; Zeraouli, Y. Energy consumption and efficiency in buildings: Current status and future trends. Clean. Prod. 2015, 109, 118–130. [Google Scholar] [CrossRef]
Pelletier, K.; Wood, C.; Calautit, J.; Wu, Y. The viability of double-skin façade systems in the 21st century: A systematic review and meta-analysis of the nexus of factors affecting ventilation and thermal performance, and building integration. Build. Environ. 2023, 228, 109870. [Google Scholar] [CrossRef]
Mahboube, R.; Fatemeh, A. Energetical effect of electrochromic glazing on the double-skin façade of the building in different climates. Energy Build. 2024, 316, 114344. [Google Scholar]
Zheng, C.; Chen, C.; Hong, X.; Zhang, W.; Yang, R.; Shi, F. Experimental evaluation of the thermal, lighting, and energy performances of a mechanically ventilated double-skin façade with Venetian blinds and a light shelf. Energy Build. 2024, 306, 113947. [Google Scholar] [CrossRef]
Duan, Z.; Zhang, M.; Li, Y. Research on the design of double-glazed curtain wall of office buildings in Xuzhou. Archit. Ski. 2023, 29, 86–89. [Google Scholar] [CrossRef]
Lahayrech, S.; Siroux, M.; El Maakoul, A.; Khay, I.; Degiovani, A. A Review: Ventilated Double-skin Façades. IOP Conf. Ser. Earth Environ. Sci. 2022, 1050, 012019. [Google Scholar] [CrossRef]
Nari, Y.; Yeonsook, H. Weather-based operation strategy for a dynamically compartmentalized double-skin façade system. Build. Environ. 2022, 226, 109755. [Google Scholar]
Jaewan, J.; Wonjun, C.; Hansol, K.; Jung-Ho, H. Load characteristics and operation strategies of building integrated with multi-story double skin facade. Energy Build. 2013, 60, 185–198. [Google Scholar]
Lin, Z.; Song, Y.; Chu, Y. Summer performance of a naturally ventilated double-skin facade with adjustable glazed louvers for building energy retrofitting. Energy Build. 2022, 267, 112163. [Google Scholar] [CrossRef]
Wonjun, C.; Jaewan, J.; Younghoon, K.; Jung-Ho, H. Operation and window-opening arrangements for multi-storey double skin facades during the heating season. Energy Build. 2012, 49, 454–465. [Google Scholar]
He, G.; Shu, L.; Zhang, S. Double skin facades in the hot summer and cold winter zone in China: Cavity open or closed? Build. Simul. 2011, 4, 283–291. [Google Scholar] [CrossRef]
Li, X. Dynamic Ventilation and Heat Transfer Characteristics of Double-Layer Breather Glass Curtain Wall in Xi’an. Master’s Thesis, Xi’an University of Architecture and Technology, Xi’an, China, 2021. [Google Scholar] [CrossRef]
Dong, Q.; Zhao, X.; Song, Y.; Qi, J.; Shi, L. Determining the potential risks of naturally ventilated double skin façades. Renew. Sustain. Energy Rev. 2024, 191, 114064. [Google Scholar] [CrossRef]
Gao, S.; Yang, L.; Zhao, S.; Zhai, Y. Study on influence of external walls and windows on human thermal comfort in office buildings. Build. Energy Effic. (Engl. Chin.) 2022, 50, 52–56. [Google Scholar]
Liu, Q.; Wang, D. Building Daylighting; Intellectual Property Press: Beijing, China, 2018. [Google Scholar]
Greenbuildersville. Vent User Manual, 2016 ed.; Green Buildersville: Beijing, China, 2016. [Google Scholar]
Yang, J.; Wang, Z.; Tian, Z.; Du, X. ‘Ventilation’ and ‘wind trapping’: A comparative study of ventilation effects in high-rise public housing in subtropical regions. In Proceedings of the 2017 National HVAC Simulation Academic Annual Conference, San Francisco, CA, USA, 7–9 August 2017. [Google Scholar]
JGJ/T 309-2013; Test and Evaluation Standard for Building Ventilation Effect. China Building Industry Press: Beijing, China, 2013.
GB50736-2012; Heating, Ventilation and Air Conditioning Design Code for Civil Buildings. Ministry of Housing and Urban-Rural Development of the People’s Republic of China: Beijing, China, 2012.
Lin, X.; Zhang, Z.; Lin, J.; Yue, X.; Huang, Y. Enhanced Accuracy for the Determination of Chromium in Steel by Laser-Induced Breakdown Spectroscopy (LIBS) and Extreme Random Tree (ERT). Anal. Lett. 2024, 57, 1829–1840. [Google Scholar] [CrossRef]
Li, J.; Gao, X.; Li, Z. Comparative study of machine learning algorithms for soybean protein localization predicting. Soybean Sci. 2022, 41, 337–344. [Google Scholar]
Xu, T. Recent advances in Rapidly-exploring random tree: A review. Heliyon 2024, 10, e32451. [Google Scholar] [CrossRef] [PubMed]
GB50033-2013; China’s National Standard for Lighting Design of Buildings. China Ministry of Housing and Urban-Rural Development of the People’s Republic of China: Beijing, China, 2013.
Meng, Y. Study on Ventilation Performance of Double Skin Facades. Master’s Thesis, Zhejiang University, Hangzhou, China, 2023. [Google Scholar]
Yuan, T.; Yang, X.; Xia, X. CFD numerical simulation of aerodynamic noise in close proximity asymmetric twin-tower linked structures. Build. Struct. 2022, 52, 89–94. [Google Scholar]
Liu, J.; Liu, H.; Wang, Z.; Zeng, Z. Study on bearing performance of defective reinforced concrete pipes based on orthogonal tests. Build. Struct. 2024, 54, 119–125. [Google Scholar]
Lin, G.; Jiang, P.; Cui, B.; Lin, A.; Wan, X. Predicting the performance of a functional ecological substrate via a generative model based on an orthogonal experiment. Bull. Eng. Geol. Environ. 2024, 83, 325. [Google Scholar] [CrossRef]

Figure 1. DSF system classification diagram.

Figure 2. DSF system diagram: (a) orientation of the research object on the 14th floor; (b) elevation structure diagram of DSF; (c) photograph of interior DSF; (d) DSF cavity diagram.

Figure 3. Average high and low temperatures in Xi’an.

Figure 4. Flowchart of the main methodology.

Figure 5. ET algorithm flowchart.

Figure 6. Overall plan structure of the model and modeling effects: (a) plan structure of the room and DSF; (b) modeling effects of the room and DSF.

Figure 7. Heat map of weather factors before processing.

Figure 8. Heat map of weather factors after processing.

Figure 9. Indoor temperatures at different opening angles from May to August: (a) 0° and 20° angles; (b) 50° and 70° angles.

Figure 10. Comparison of indoor and outdoor temperatures in a typical week in summer.

Figure 11. ET algorithm predicts different kinds of indoor temperature results: (a) the prediction results of the sample; (b) confusion matrix of ET algorithm.

Figure 12. The average temperature curves at 1.2 m height indoors under different window-opening arrangements.

Figure 13. Temperature cloud at 1.2 m height plane under window-opening arrangement 1 and window-opening arrangement 2: (a) window-opening arrangement 1; (b) window-opening arrangement 2.

Figure 14. A 1.2 m velocity cloud image under different window-opening arrangements: (a) window-opening arrangement 1; (b) window-opening arrangement 2.

Figure 15. Average temperature curves of internal curtain walls under different window-opening arrangements.

Figure 16. Temperature distribution of the interior curtain wall under window-opening arrangements 1, 2, and 8: (a) window-opening arrangement 1; (b) window-opening arrangement 2; (c) window-opening arrangement 8.

Figure 17. Average indoor temperature curves under different window-opening arrangements.

Figure 18. Average indoor temperature profile after extending the simulation time for window-opening arrangement 2.

Table 1. Comparison of common machine learning methods.

Methodologies	Advantage	Drawback
ET	No pre-processing of data, simple and effective categorization
neural network	Self-learning function and associative storage	Rely on all data
Decision tree	Intuitive decision-making rules	Prone to overfitting, difficulty in dealing with missing data
Support Vector Machine (SVM)	Not dependent on all data	Inefficient for large number of prediction samples

Table 2. Table of flow coefficients for suspended windows.

Angle	20°	30°	45°	60°	90°
Flow coefficient	0.33	0.38	0.52	0.60	0.62

Table 3. Equivalent flow areas table for windows.

Size	Width/m	High/m	Area/m²	Equivalent Area/m²
Inner window	1.20	0.90	1.08	0.60
External window	1.07	0.96	0.96	0.54

Table 4. Table of DSF glass curtain wall material parameters.

Window-to-Wall Ratio	Transmission of Visible Light (%)	Visible Light Reflectance (%)	Heat Transfer Coefficient (W/m²)	Sclar Coefficient (SC)	Sclar Heat Gaim Ocefficient (SHGC)
0.6	57	16	1.65	0.44	0.38

Table 5. Simulation boundary conditions.

Name	Boundary Condition	Parameterization
External curtain wall	Mixed heat transfer	Thickness 0.024 m; Heat transfer coefficient 1.64 W/(m²K)
Inner wall	System Coupling	Thickness 0.024 m
Alloy plate	Mixed heat transfer	Heat transfer coefficient 41.64 W/(m²K)
Entrances	Pressure inlet	Average wind speed of 3 m/s
Exits	Free-flowing	Free-flowing
Outer Space Boundary	Temperature boundary	Consistent with outdoor temperatures

Table 6. Window-opening arrangements table.

Number	Window 1	Window 2	Window 3	Window 4
Arrangement 1	Open	Open	Open	Open
Arrangement 2	Open	Open	Close	Close
Arrangement 3	Open	Close	Open	Close
Arrangement 4	Open	Close	Close	Open
Arrangement 5	Close	Open	Open	Close
Arrangement 6	Close	Open	Close	Open
Arrangement 7	Close	Close	Open	Open
Arrangement 8	Close	Close	Close	Close

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Sun, Q.; Song, J.; Yu, Y.; Ai, H.; Zhao, L. A Study of the Impacts of Different Opening Arrangements of Double-Skin Façades on the Indoor Temperatures of a Selected Building. Buildings 2024, 14, 3893. https://doi.org/10.3390/buildings14123893

AMA Style

Sun Q, Song J, Yu Y, Ai H, Zhao L. A Study of the Impacts of Different Opening Arrangements of Double-Skin Façades on the Indoor Temperatures of a Selected Building. Buildings. 2024; 14(12):3893. https://doi.org/10.3390/buildings14123893

Chicago/Turabian Style

Sun, Qing, Junwei Song, Ying Yu, Hongbo Ai, and Long Zhao. 2024. "A Study of the Impacts of Different Opening Arrangements of Double-Skin Façades on the Indoor Temperatures of a Selected Building" Buildings 14, no. 12: 3893. https://doi.org/10.3390/buildings14123893

APA Style

Sun, Q., Song, J., Yu, Y., Ai, H., & Zhao, L. (2024). A Study of the Impacts of Different Opening Arrangements of Double-Skin Façades on the Indoor Temperatures of a Selected Building. Buildings, 14(12), 3893. https://doi.org/10.3390/buildings14123893

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu