Parametric Design and Spatial Optimization of East–West-Oriented Teaching Spaces in Shanghai

Abstract

The goal of the current study was to determine the appropriate spatial shapes for classroom occupants while saving energy. The research used parametric design and Genetic Algorithm (GA) to achieve this. Four recognized performance indicators, Energy Use Intensity (EUI), Useful Daylight Illuminance (UDI), Daylight Factor (DF), and Daylight Autonomy (DA), were used as the evaluation indexes for the research. The tests took place in six east–west-oriented classrooms at Shanghai University, China. The methodology was based on four steps: (1) parametric 3D modeling by Rhino and Grasshopper; (2) using building performance simulation tools; (3) running algorithm optimization; (4) outputting the useful results. The results proved that the methodology worked successfully in reducing energy consumption: optimized classrooms could be reduced by 7.5~14.5%, and classrooms with east directions were generally 4.8~8.3% more efficient than west-facing ones. The indoor lighting environment was also significantly improved, being slightly better than north–south-oriented classrooms in terms of the UDI index (60~75%) and inferior (but still high) in terms of the DF (4.0~7.0%) and DA (60~80%) indexes. The conclusion can help save design time in the early design process of teaching spaces.

1. Introduction

Since the industrial revolution, human beings have been obsessed with regulating their interior environments through building HVAC systems to achieve a stable constant temperature and humidity. However, the excessive pursuit of a comfortable indoor environment can lead to an explosive increase in energy consumption. In ref. [1], the authors evaluated the current status and future prospects of energy consumption simulation in China, and these findings examined the direction of what we should do in early design. Of course, the implementation of relevant national standards was an excellent starting point for building energy efficiency. There are two things that need to be done. One is to further investigate architectural space alternatives based on the requirements. The other is to suggest some unreasonable features of the standards in light of the current situation. Therefore, the trade-off between indoor thermal comfort, building energy efficiency, and lighting environment must be investigated.

In ref. [2], Arens E. et al. proposed the ASHRAE standard 55 by conducting tests of occupants’ thermal comfort and thermal sensation in three existing buildings based on ISO standard 7730 and European standard CEN 15251. The three classes did not exhibit different comfort/acceptability outcomes as given by occupants, despite the high energy costs of class A where environmental controls were more stringent. The PMV-PPD theory was therefore found to be flawed in that it was not sufficiently precise [3,4]. The authors suggested incorporating their feedback into building control, operation, and the design process. Besides, a classification of occupant-based aspects of energy use was also required [5].

In ref. [6], the biggest difference between thermal comfort and thermal sensation was that the former was influenced by a great number of psychological factors. However, being thermally neutral at any time is not all good. Moderate cold and other passive strategies are still very important in the environment, which will bring pleasure and physical simulation that would otherwise degrade several body functions [7,8,9]. Furthermore, for the purpose of energy-saving building design, we typically keep the window-to-wall ratio (WWR) around 40% [10]. This is because a WWR of 40% is an important inflection point for cooling energy consumption. Instead of increasing with the growth of WWR in the south and west directions, heating energy consumption also shows a significant decreasing trend. By using building performance simulation (BPS) tools such as Ecotect and CFD, the researchers developed a green design approach for primary and secondary school buildings in a cold climate zone [11]. They applied suitable strategies to environmental optimization and applied them to the practice of a school building in Tianjin. Fan and Lv [12], by simulating the combined use of air-conditioning and fans in schools during the transition season in Shanghai, aimed to investigate the relationship between the feasibility of improving indoor comfort and the rate of energy savings. It was found that the combined use of electric equipment could be in accordance with indoor comfort requirements and that this strategy could save 23.2% of energy during the transition season. The characteristics and usage patterns of schools dictated that the use of air conditioners alone was not conducive to either cooling or building energy efficiency [13]. It can be said that one still has to rely on a series of passive measures to solve the widespread problem of poor thermal performance in classrooms. In ref. [14], the authors summarized the frequency of use of passive green building design through practice. The significance of different types of natural ventilation in architectural design can be seen in ref. [15], and these conclusions are of great significance for passive energy-saving design. Moreover, night ventilation can be used to release the heat stored during the day, and mechanical ventilation can also be used to introduce fresh outdoor air to satisfy indoor demand.

A good indoor lighting environment is especially important for students. In ref. [16], the scholars evaluated the thermal, acoustic, and lighting comfort conditions of classrooms in Italy in terms of subjective and objective degrees, which provided a reference for later spatial optimization. Some studies [17,18] found that higher lighting conditions in schools had a protective effect on myopia. In his book [19], Cohn suggested that the best lighting for classrooms would be through glass skylights in the roof. Of course, there are also some practical activities to explore the lighting environment of teaching space, as close as possible to the outdoor environment, which can prevent the occurrence and development of myopia in the form of special classrooms [20]. Additionally, by comparing the standards of different developed countries, the focus of school lighting design was summarized in ref. [21]. It was suggested that landscape windows should be considered separately from light windows, and examples were used to prove this point.

These analyses suggest that the thermal comfort, passive strategies, construction, HVAC systems, lighting environment, building energy efficiency, and occupant features have a significant role in influencing the interior environment. The primary purpose of the classroom is mainly to help the teacher complete their teaching work. Therefore, continuous focus on what they are seeing and hearing during class is crucial for students. It is a fact that the quality of natural light is far superior to that of artificial light, due to the high demand for daylight in these classrooms and the fact that they are general performance spaces [22]. Therefore, a balanced relationship between a proper indoor lighting environment and energy consumption control should be considered in this study.

Some scholars have been researching digital architecture since the 1960s. From Computer-Aided Design [23] to Evolution Architecture [24] to Shape Grammars [25], digital technology has set off a wave of exploration of architectural design theories and methods. Architecture design gradually shifted from top–down to bottom–up energy-saving design theory, then to digital building energy-saving design with the development and improvement of BPS software technology [26,27,28]. It can take advantage of the complex calculation and arithmetic power of computers to break the previously experience-oriented design. Once the combined with the previous theories, it may provide architects with more opportunities to explore energy-efficient design. In reviews [29,30,31,32,33], a simulation framework was developed to support decision-making in early design. The workflow was documented in detail: target selection, statistical methods, meta-modeling, CAD-BPS interoperability, parametric modeling, and optimization algorithms. It seems necessary to first discuss the overall design of the school from a macro perspective, and it is also vital to study the plan of teaching functional unit space on top of that. Ma Chengye [34] conducted research on the generative design of primary and secondary schools with lighting based on the multi-agent system. He used python based on Rhino and Grasshopper platforms to calculate the spatial forms of educational buildings which are suitable for the Chinese context. We found that the layout created using this technique was equivalent to the scheme designed using standard methods in terms of logic when we looked at the plan and creation of a school as an example.

However, there were still flaws in the indoor teaching units’ area, and this literature set the groundwork for this study. This study, therefore, focused more on the spatial design of teaching space. In contrast to previous literature, this work proposed a novel multi-objective optimization workflow based on coupling several platforms to address issues with the east–west-oriented teaching environment, with a particular emphasis on the influence of spatial form on the indoor lighting environment and considering the overall control of energy consumption.

This paper is structured as follows. To begin, the study topic is defined, and the major challenges in the spatial environment of educational buildings with an east–west orientation are discussed. Secondly, a research method is proposed to solve these issues by parametric 3D modeling (Rhino and Grasshopper). Then, we analyze the advantages and disadvantages of teaching space by BPS tools (Ladybug Tools) and optimize them (Wallacei X). Finally, reliable conclusions are drawn for spatial forms.

2. The Study Area

This study was conducted in Shanghai, China. Shanghai has a population of 24.87 million as of 2020 [35], and is located at 31°10′ N, 121°28′ E with around 4 m elevation above sea level. The climate of Shanghai is a humid subtropical climate (Cfa), where summers are hot and rainy, while winters are cold and damp with northwesterly winds from Siberia. The annual temperature averages 17.0 °C and the relative humidity is 80.2% throughout the year (according to a weather station in Xujiahui) [36]. The EnergyPlus Weather (EPW) file shown in Figure 1 can be used to analyze the current situation in the city of Shanghai [37].

Shanghai’s four seasons are well-defined, with sufficient solar radiation. The building in Figure 2 is the subject of this study. It is called the 4th Academic Building and is located in Shanghai University. It has a reinforced concrete structure built in the 1990s, with fair fenestration on the facades and no sun visors for shading. Additionally, it is an east–west-oriented building facing 20° west of north. Its dimensions are 20 m length, 55 m width, and 25 m height. We ran the LB Incident Radiation simulation using a sky matrix from the Cumulative Sky Matrix component, which computed the incident radiation on geometry. After running an annual simulation of solar radiation, it could be found that the south and west directions had stronger solar radiation, 736.45 kWh/m² and 613.71 kWh/m², respectively, while the east-facing direction of building was weaker, at only 490.97 kWh/m².

The city receives 1895 h of direct radiation annually. According to the Benefit/Harm Radiation chart in Figure 1 and the Building Solar Radiation chart in Figure 2, it is clear that solar radiation from the south is the most helpful radiation; however, it is not always appropriate for teaching space. We also need to limit solar radiation absorption from the east and west, especially in the afternoons when solar radiation from the west is very strong. Otherwise, the building’s interior area would collect a lot of solar radiation during the afternoon hours, making the internal climate too hot to dwell in and causing terrible glare problems.

3. Methodology

The workflow of this study is mentioned as follows (Figure 3). The details of each step are described in the next section. The next chapter is the core chapter in which useful data results can be seen. Finally, reliable conclusions are drawn by aggregating and analyzing these data.

• parametric 3D modeling by Rhino and Grasshopper tools;
• setting environmental parameters and running simulation by BPS software (Ladybug and Honeybee);
• confirmation of fitness optimization targets and selection of genes;
• muti-objective algorithm optimization based on genetic algorithm (Wallacei X);
• results selection and phenotype output.

3.1. Classroom Classification and Parametric Modeling

According to the previous studies in ref. [16], the researchers selected seven different types of classrooms for the study in order to make the research more comprehensive. Thus, six representative classrooms were selected at Shanghai University (in an east–west-orientation building). It can be classified by classroom volumes: classrooms A and B are small-sized; classrooms C, D, and E are medium-sized; and classroom F is large-sized. The majority of east–west-oriented classroom scenarios may be covered by these three types of representative classrooms. So, these classrooms, with different facades, sizes, orientations, etc., have sample roles. The classroom plans and actual pictures are shown in Figure 4.

The fenestrations of classrooms are roughly divided into three parts: east only, west only, and bilateral window openings. Small-sized classrooms have only single-sided window openings (east or west side). Medium-sized classrooms have mainly single-sided window openings, but they also have bilateral window openings (west and north sides). The large-sized classroom has bilateral window openings (west and south sides), which especially ensures that the south facade must be involved.

Table 1. Characteristics of the investigated classrooms.

Classroom	A	B	C	D	E	F
Height (m)	3.4	3.4	3.4	3.8	3.8	4.0
Width (m)	6.80	6.80	10.70	7.42	7.42	10.81
Length (m)	7.11	7.11	7.08	10.40	10.40	17.42
Floor area (m2)	47.89	47.89	75.23	76.59	76.59	178.95
Volume (m3)	162.83	162.83	255.78	291.04	291.04	715.80
N° of doors	1	1	2	2	2	4
Total doors surface (m2)	1.89	1.89	3.78	3.78	3.78	6.13
Orientation	West	East	West, North	West	East	West, South
N° of windows	3	3	6	5	5	8
Total windows surface (m2)	10.92	10.92	16.29	16.22	16.22	35.26
Material of windows	Glass-metal
Ceiling material	Plaster
Floor material	Marble
Wall surface material	Plaster
Surface color	White, gray, beige		White, gray, brown	White, gray, beige
Building service systems	HVAC system and fans
Type of electric light	Fluorescent lamp
Shading	Curtains (no external visors)

contains all of the information on the experiment classrooms, including the size, orientation, number of windows, building service systems, electric lighting, and so on.

It should be mentioned that the structure under investigation lacks outdoor sunshades for the following reasons. As a result, this factor was not taken into account in the next optimization design. Firstly, as it serves as a school, the building is primarily used in the spring and fall, excluding the scorching summer and chilly winter, and thus the lack of exterior sunshades has relatively little bearing on the interior. Classrooms are also not used all day, usually only between 8:00 a.m. and 5:00 p.m. Secondly, trees outside the building can provide some shade and block some of the direct sunshine. This is particularly helpful in the late afternoon when the angle of sunlight coming from windows facing west is low. In addition, users can use curtains and interior translucent awnings for glare control. Finally, the maintenance cost of outdoor shading structures should not be underestimated. Therefore, it is possible to disregard outdoor shades for east–west-oriented teaching spaces.

3.2. Calibration of Environmental Parameters and Calculation of Indicators

In this step, it is indisputable that environmental parameters are assigned to each component of the model by using Ladybug and Honeybee tools. (1) The reflectance of the wall was 50%. The reflectance of the roof ceiling was 80%, while the floor was 20%, and all boundary conditions were set to adiabatic. The window transmittance was 86%. (2) The program of room was selected as “2013: Secondary School: Classroom” that output from the “HB List Programs” component (ASHRAE 90.1 2013 | IECC 2015). (3) A HB Room was created from honeybee surfaces and, then, all the outdoor apertures of rooms were extruded to 0.24 m to create a HB Model. (4) When performing environmental simulations, local climatic factors must be taken into account. We downloaded the EPW file of Shanghai from the official website of EnergyPlus by Ladybug Tools, which can import the local data into Grasshopper for simulation.

Secondly, we come to annual daylight simulation. To begin with, we must first construct sensor grids from the floor of classrooms and convert them into radiance senor grids that can be incorporated to a HB Model. This task is required for annual daylight study [38]. We set a number of options such as the grid size (1 × 1 m), the sample points height at a working level of 0.75 m, the lower threshold (−lt = 100) and upper threshold (−ut = 2000) for useful daylight illuminance (UDI), and the threshold (−t = 300) for daylight autonomy (DA). We used the running schedule “School Secondary BLDG_LIGHT_CLASSROOM_SCH_2013” from the honeybee energy strands library in order to obtain more realistic simulation results. Then, the output terminals of DA and UDI_100-2000lx results were linked to the “LB Spatial Heat Map” component to visualize images where access is granted to Legend Parameters to recolor the mesh or limit threshold. Finally, data from the simulation outputs were summed up first, and then the average values were calculated and recorded. In comparison to DA and UDI simulation stages, it appears that similar operating methods for running a daylight factor (DF) simulation are simpler, in which case it does not consider the climate zone. Aside from that, the rest of the steps are the same.

Settings are more complicated regarding energy simulation. In addition to the above mentioned settings, here are some new parameters to add. We looked up face types for construction settings from the HB materials library. Some building constructs, such as walls and glasses, use the most prevalent forms in order to portray the universality of the outcomes. The windows in the classroom were covered with a typical single piece of 5 mm thick glass. Details of the construction material values are shown in

Table 2. Main construction material values.

Material	Brick	LW Concrete	50 mm Insulation	Wall Air gap	Gypsum Board	Concrete Floor	Ground Floor_R11
Roughness	Medium rough	Medium rough	Medium rough	Smooth	Medium smooth	Medium rough	Smooth
Thickness (m)	0.1	0.1	0.05	0.1	0.01	0.1	0.04
Conductivity (W/m∙k)	0.9	0.53	0.03	0.67	0.16	2.31	0.02
Density (kg/m3)	1920.0	1280.0	43.0	1.28	800.0	2322.0	16.0
Specific heat (J/kg∙K)	790.0	840.0	1210.0	1000.0	1090.0	832.02	1129.71
Thermal absorptance	0.9	0.9	0.9	0.9	0.9	0.9	0.9
Solar absorptance	0.65	0.8	0.7	0.7	0.5	0.7	0.7
Visible absorptance	0.65	0.8	0.7	0.7	0.5	0.7	0.8
R value (m2∙k/W)	0.11	0.19	1.67	0.15	0.08	0.04	1.92

Notes: All data are from HB Material Library. It is a dependable source of data, and used to ensure the authenticity of the experiment.

. What is more, detailed schedules for activity, occupancy, lighting, electric equipment, heating set point, and cooling set point were assigned according to national and local standards in refs. [39,40,41,42,43,44] to make the simulation results closer to the real data. We created the HB Model by using an ideal air system (ventilation should be used on the system) and HVAC system (VRF, ASHRAE 90.1 2013 | IECC 2015) for classrooms. We imported the EPW file of the city of Shanghai and selected the simulation output that needs to be calculated, such as zone energy use, HVAC energy use, gains and losses, surface temperature, and surface energy flow. An Open Studio Model (OSM) was created from an HB Model so that it may be converted to an IDF file and run through EnergyPlus to simulate building performance. Subsequently, the results were linked to the “HB Thermal Load Balance” component, and the “LB Monthly Chart” component was selected to visualize the data. Data are included in the Energy Use Intensity (EUI) part to form an EnergyPlus SQL file that can be read out for subsequent comparative analysis by the panel. Finally, it links to the “LB Monthly Chart” component to visualize the data as well. Detailed pictures and data can be found in

Table 3. Image results for six classrooms.

Indicators	UDI (%)	DA (%)	DF (%)	ECB (kWh/m2)
Classroom A
Classroom B
Classroom C
Classroom D
Classroom E
Classroom F

Abbreviations: UDI: useful daylight illuminance; DA: daylight autonomy; DF: daylight factor; ECB: energy consumption balance. Notes: test images of six experimental classrooms showed the current situation using HB-Radiance and HB-Energy component parts.

and

Table 4. Data recording of six classrooms’ indoor test results.

Indicators	UDI (%)	DA (%)	DF (%)	EUI (kWh/m2)	Test Points	Orientation
Classroom A	67.92	61.20	4.10	109.50	49	West
Classroom B	70.14	58.76	4.09	105.36	49	East
Classroom C	67.26	73.02	2.86	103.37	77	West and North
Classroom D	66.41	67.63	5.08	100.91	70	West
Classroom E	70.03	70.14	5.05	104.25	70	East
Classroom F	61.53	62.74	4.54	109.35	202	West and South

Notes: Average data recording of UDI, DA, and DF.

.

According to the local standards in Shanghai [44], the relevant dimensional requirements for normal classrooms (north–south orientation) were tested as given here for comparison. Other environmental design parameters were unchanged, and only the classroom dimensions were modified, H = 3.6 m, L = 9.4 m, W = 7.3 m. Phenotypes for normal classrooms are EUI: 111.24 kWh/m²; UDI: 65.34%; DF: 6.81%; DA: 81.00%. We would like to compare the situation of north–south-oriented classrooms with that of east–west-oriented classrooms as a comparative analysis later.

Several conclusions may be derived based on the lighting environment and energy consumption of the existing teaching space after simulations. Firstly, the UDI index tended to decrease slowly as the room became larger. One point worth noting is that classrooms with east window openings were better than those with west ones because of the glare problems from the west during the afternoon time. Furthermore, the UDI index remained at the lowest level in the large-sized classrooms despite the fact that they add extra windows on the south side to attempt to obtain a better interior lighting environment. However, the opposite is true for the DA and UDI indexes. For the DF index, there was little difference between east-oriented and west-oriented classrooms. The DF indexes in classrooms A and B basically met the minimum conditions of use (DF > 4.0%). Nonetheless, classroom C was too deep to obtain sunlight in the east and west directions, resulting in the lowest DF index (watch out for this situation when optimizing). Classroom D and E had fit volumes and the highest DF indexes, and classroom F seemed to have a good DF index in combination with the south window openings. Besides, the energy consumption of classrooms were acceptable, but the EUI indexes were all above 100 kWh/m². As a result, there is still potential for improvement.

By simulating the daylight and energy use of six experimental classrooms, we can learn about the qualities and flaws of each. When it comes to east–west orientation, the emphasis on lighting and energy consumption is differentiated for distinct sizes of teaching spaces. The results of this analysis will guide the optimal settings in the next section.

3.3. Shape Optimization with Evolutionary Algorithms

Grasshopper [45,46] is a popular parametric software that is built on the Rhino platform. It helps users build a one-step workflow by coupling with many plugins. Ladybug Tools are open-source plugins that unite commonly used BPS software, such as Radiance [47] and EnergyPlus [37]. There are a large number of studies about school building spaces through the use of BPS software. Furthermore, there is also a good amount of literature studying parametric design, mostly focusing on complex form generation and structures [48]. Optimization algorithms are known to be used mainly in industry and engineering areas in early stages [49]. However, it is not enough to develop further research on spatial morphology, and therefore a small number of scholars have emerged who have used optimization algorithms to solve architectural problems (they are relatively new).

Grasshopper also allows for single-objective and multi-objective optimization algorithm computations. Wallacei X is a genetic optimization plugin, such as the K-means as the clustering algorithm [50,51]. Genetic algorithm is a computer model of a biological evolutionary process that replicates Darwinian biological evolution’s natural selection and genetics principles (Figure 5 and Figure 6). In other words, it is a method of searching for optimal solutions by simulating natural evolutionary processes. Wallacei X runs Non-dominated Sorting Genetic Algorithm (NSGA-II) [52,53]: an evolutionary computation. It allows users to establish multiple optimization objectives by selecting genes and chromosomes based on the real situation so that each parameter can be adjusted in detail [54].

In this work, four fitness values were established to run optimization, UDI, DF, DA, and EUI, mainly from the perspective of lighting environment and energy consumption.

By utilizing Wallacei X to determine the Min EUI; Max UDI, DF, and DA indexes and output phenotype, we hope to discover suitable east–west-oriented teaching space types (Figure 3). We classified the teaching spaces into small, medium, and large sizes according to their volumes and actual usage, according to Section 3.1 and Section 3.2. The information of each index, priority, and classroom sizes are recorded in detail in

Table 5. Selection of test indicators for different types of classrooms.

Indicators	Min EUI	Max UDI	Max DF	Max DA	Priority of Energy	Priority of Daylight
Small classroom	√	√	√			√
Medium classroom	√	√	√	√	√	√
Large classroom	√	√	√	√	√

. The indoor lighting environment just exceeds the basic specification limits in small-sized classrooms in Table 4, and the actual experience is not satisfactory. Therefore, we must ensure it, as one of the main teaching spaces, can achieve a better quality of lighting environment where the daylight priority is set to be higher than the energy priority. Medium-sized classrooms are the most numerous and appropriate types for teaching and learning, and we must carefully balance energy consumption and lighting environment. It is believed that the optimization results of this kind of space will provide guidance results for their design. Large-sized classrooms are huge in volume, accommodating more people but requiring fewer light standards than the former. It is more reasonable to start with the priority of reducing energy consumption. Detailed information is set in a parametric way, such as classroom sizes, WWR, number of genes, steps, etc., in

Table 6. Optimized gene targets selection and step size setting.

Optimization Range	Small Classroom	Medium Classroom	Large Classroom	Step
Height (m)	3.2~4.2	3.2~4.4	3.4~4.4	0.1
Length (m)	6~7.5	7.5~12	12~20	0.1
Width (m)	6~7.5	7.5~12	10~15	0.1
Orientation (°)	−30°~30°	−30°~30°	−30°~30°	1
WWR (%)	West: 30~70	West: 30~70	West, South: 30~70	5
	East: 30~70	East: 30~70	East, South: 30~70
N° of Genes	5	5	6	-

Notes: The right range of parameters is the key to the experiment.

.

4. Results and Discussions

The data derived from the optimization constitute the core results of this paper. The following steps were performed on the data.

• Filter the optimization results that are calculated;
• Record a detailed analysis;
• Determine visual expression and phenotype.

Firstly, the extreme situations of each performance index were found as references. Secondly, we fuzzy checked the cases that roughly satisfied the conditions in the controllable range. Finally, the best-ranked genomes were selected from the entire population of the Pareto Front Sets (phenotype). Three types of classrooms are categorized and discussed in the following.

4.1. Results from Small-Sized Classroom Optimization

4.1.1. West Facade Fenestration for Small Classrooms

The optimization of small teaching spaces was first run through Wallacei X with a population size of 30 individuals per generation, for a total of 50 generations. To sum up, a total population size of 1500 genomes was evaluated. The population size and running time were as follows:

−

Generation size: 30, Generation count: 50

−

Population size: 30 × 50 = 1500 genomes

−

Time required for simulation: 14.5 h

Based on the higher priority of daylight for small-sized classrooms previously identified in Table 5, the Min EUI, Max UDI, and DF indicators were mainly considered here (the DA index was not included in the calculation to simplify the optimization process). All results were analyzed and filtered in detail and are shown in

Table 7. The case of small-sized classrooms with window openings on the west facade.

Small Classroom	Min EUI	Max UDI	Max DF	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	94.32	102.30	126.25	104.30	101.35	103.97
UDI (%)	69.65	76.19	54.35	65.20	68.34	66.47
DF (%)	2.31	3.23	8.61	5.52	4.21	4.86
Population	Gen.24 Ind.1	Gen.0 Ind.24	Gen.42 Ind.1	Gen.49 Ind.28	Gen.45 Ind.6	Gen.35 Ind.26
Diamond Fitness Chart

Notes: Extreme values of three performance indexes (Min EUI, Max UDI, and DF) and the better selection of Pareto Front Solutions for small-sized classrooms.

and

Table 8. Genomes in detail and phenotype output for small-sized classrooms with west facade fenestration.

Small Classroom	Min EUI	Max UDI	Max DF	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.2	4.1	4.2	3.2	3.3	3.3
Length (m)	7.5	6.8	7.4	7.5	7.5	6.8
Width (m)	7.5	7	6	7.5	7.4	7.3
Orientation (°)	23°	−20°	30°	23°	30°	30°
Plan
WWR (%)	30	30	70	65	50	50
Model

and Figure 7.

Some useful information can be read from the results. The height of a small classroom with a west-facing direction should be controlled at 3.2~3.4 m, and the aspect ratio is a square close to 1:1. In other words, the length and width should be close to 7.5 m (both sides range from 6.5 m to 7.5 m). The classroom orientation is optimal at 20~30°, and the WWR may control at 50~65%.

The Standard Deviation (SD) chart in Figure 7 shows the first to the last generation from red to blue. A good convergence boundary was formed after conducting the initial extensive search, and the UDI and DF indexes trendlines showed a certain upward trend at the later stage of the graph. The EUI indexes also stabilized above 100 kWh/m², converging towards 110 kWh/m². It showed that only the EUI index must be stable at around 100 kWh/m² for the DF index to exceed 4.0%. At this point, the data are as follows: EUI fitness ranking = 200 (population: 0 to 1499 ranking); Gen. 27, Ind. 12; EUI: 100.05 kWh/m²; UDI: 68.80%; DF: 4.06%. Phenotype: Height = 3.2 m, Length = 7.5 m, Width = 7.5 m, Orientation = 26°, and WWR = 50%. This data set is similar to the Pareto Fronts.

4.1.2. East Facade Fenestration for Small Classrooms

The population size and running time here were the same as above, with the exception that the classroom types were fenestrated differently. Detailed data are as follows in

Table 9. The case of small-sized classrooms with window openings on the east facade.

Small Classroom	Min EUI	Max UDI	Max DF	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	89.19	94.80	115.93	99.95	95.65	97.73
UDI (%)	68.83	76.72	55.57	63.52	67.77	66.15
DF (%)	2.31	2.78	8.50	6.09	4.62	5.20
Population	Gen.47 Ind.3	Gen.1 Ind.20	Gen.33 Ind.0	Gen.48 Ind.28	Gen.33 Ind.9	Gen.41 Ind.28
Diamond Fitness Chart

and

Table 10. Genomes in detail and phenotype output for small-sized classrooms with east facade fenestration.

Small Classroom	Min EUI	Max UDI	Max DF	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.2	3.5	4.1	3.3	3.3	3.3
Length (m)	7.5	6.3	7.5	7.5	7.5	6.7
Width (m)	7.5	6.6	6	7.4	7.5	7.5
Orientation (°)	−26°	20°	−28°	−29°	−25°	−24°
Plan
WWR (%)	30	30	70	70	55	60
Model

and Figure 8.

The height of a classroom should be kept at 3.2~3.4 m for a small east-oriented teaching space, and the length and width should still be close to 7.5 m (an aspect ratio close to 1:1), much like a square. However, about 6.5 × 7.5 m is also a viable option. The classroom’s best orientation is about −20~30°, which is the counterpart to the west one, and the WWR is controlled at 50~70%.

Although the test classrooms were mirrored spaces, the difference between east-oriented and west-oriented was significant. The energy consumption of the east one was obviously less than the west one, which could be reduced by about 8.4% on average. The data from the Min EUI demonstrated that pursuing a significant decrease in energy usage would result in a DF index that is insufficient to fulfill indoor demands. Therefore, the arrangement of windows can be made larger to further improve the quality of the indoor light environment. This will make the quality of lighting environment of east-directed classrooms better than west-directed ones. Most of the DF indexes could achieve more than 5.0%, or even reached 6.0% and above. However, the UDI indexes were not much different, all more than 60%. It could be shown that for the DF index to reach 4.0%, only the EUI index must stay steady at roughly 95 kWh/m². At this point, the data are as follows: DF fitness ranking = 280 (population: 0 to 1499 ranking); Gen. 49, Ind. 5; EUI: 95.02 kWh/m²; UDI: 67.75%; DF: 4.15%. Phenotype: H = 3.2 m, L = 6.4 m, W = 7.5 m, O = −29°, and WWR = 50%.

4.2. Results from Medium-Sized Classroom Optimization

4.2.1. West Facade Fenestration for Medium Classrooms

Here, after computing a population size of 1500 genomes according to the small classroom, a total of 17 h were computed. It was found that the DF index trendline did not conform to the expected results as we thought and there might be a convergence anomaly. Therefore, it was judged that the number of calculations was not enough to complete the genetic optimization for the medium-sized classroom. Therefore, the number of calculations was doubled, and another judging index (DA) was introduced to constitute four optimization fitness values. The population size and running time are as follows:

−

Generation size: 60, Generation count: 50

−

Population size: 60 × 50 = 3000 genomes

−

Time required for simulation: 47.5 h

The lighting environment and energy efficiency are equally high priorities in medium-sized classrooms. Such classrooms are the most numerous and most frequently used, and therefore require special attention. All results were analyzed and filtered in detail, and are shown in

Table 11. The case of medium-sized classrooms with window openings on the west facade.

Medium Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	85.20	97.44	105.87	107.00	93.86	97.91	97.79
UDI (%)	52.27	73.59	67.66	62.65	68.47	66.16	70.57
DF (%)	1.53	3.17	6.82	6.79	4.26	5.10	5.24
DA (%)	29.75	52.14	83.43	83.60	60.89	70.85	70.37
Population	Gen.32 Ind.0	Gen.0 Ind.42	Gen.0 Ind.18	Gen.1 Ind.33	Gen.0 Ind.48	Gen.0 Ind.26	Gen.0 Ind.6
Diamond Fitness Chart

and

Table 12. Genomes in detail and phenotype outputs for medium-sized classrooms with west facade fenestration.

Medium Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.2	4.1	4.2	3.7	3.7	3.3	3.9
Length (m)	12	8.3	11	11.5	9.7	11.7	10.7
Width (m)	12	8.6	7.9	7.7	10.4	9.6	10.6
Orientation (°)	−30°	−29°	−28°	13°	−18°	20°	−29°
Plan
WWR (%)	30	35	65	70	55	70	70
Model

and Figure 9. However, the results showed that the previously conjectured problem of insufficient generations was incorrect. All curve trendlines were similar to those of the last time, indicating that the iterations were successful and the number of generations available was roughly concentrated in the first 30 generations. However, the calculation time reached up to 47.5 h (2.8 times). Therefore, it is better to reuse the population size of 1500 genomes for the next calculation to save time.

Some data results could be obtained. The height of a west-oriented classroom of medium size was desirable at 3.7~4.0 m. The aspect ratio was slightly longer than the width of the rectangle, with a length range of 9.5~12 m and a width range of 8~11 m. The classroom orientation was optimal at −20~−30° or 10~20° (special cases but valuable), and the WWR is controlled at 55~70%.

The SD chart in Figure 9 was very different from the previous in Figure 8. However, it was accurate. The UDI SD value trendline showed a slight upward trend. Other than that, the other fitness values were in a flat state after five generations of running. A good convergence boundary was formed after conducting the initial extensive search. As the large fluctuations seemed to be concentrated in the first five generations, a novel phenomenon occurred here. In order to make a DF index > 4.0%, the DF fitness ranking should be higher than 2967 (a total population size of 3000, 0 to 2999 ranking). At this point, the data were as follows: DF fitness ranking = 2967; Gen. 0, Ind. 48; EUI: 95.16 kWh/m²; UDI: 69.07%; DF: 4.10%; DA: 57.05%. Phenotype: H = 3.7 m, L = 9.7 m, W = 10.4 m, O = −18°, and WWR = 55%. For medium-sized classrooms with a west-facing orientation, the overall energy consumption could be controlled within the range of 90 kWh/m² to 100 kWh/m², which reduced energy consumption by up to 8.5% compared with the original classroom. One set of results required attention to the Max UDI group: H = 4.1 m; L = 8.3 m; W = 8.6 m; EUI: 97.44 kWh/m². This phenotype was a relatively good solution, despite the DF index being 3.17%. However, the volume of this classroom was not considered medium-sized but could be included in the small ones.

4.2.2. East Facade Fenestration for Medium Classrooms

The population size and running time in medium classrooms were the same as in the small ones, with a population size of 30 individuals per generation and a total of 50 generations. There was just a different orientation of the classroom types. All results were analyzed and filtered in detail, and are shown in

Table 13. The case of medium-sized classrooms with window openings on the east facade.

Medium Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	87.08	90.05	102.35	102.35	88.39	93.13	89.40
UDI (%)	51.20	74.71	62.13	62.14	70.68	69.95	70.22
DF (%)	1.58	4.32	6.80	6.79	4.42	4.66	5.08
DA (%)	29.20	67.53	83.45	83.52	61.33	63.50	68.68
Population	Gen.49 Ind.2	Gen.0 Ind.11	Gen.1 Ind.20	Gen.2 Ind.24	Gen.1 Ind.21	Gen.0 Ind.3	Gen.0 Ind.26
Diamond Fitness Chart

and

Table 14. Genomes in detail and phenotype output for medium-sized classrooms with east facade fenestration.

Medium Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.2	3.4	4.2	4.2	3.6	4	3.3
Length (m)	11.8	11.8	11	11	11.9	8.8	11.7
Width (m)	12	7.9	7.9	7.9	10.4	10.3	9.6
Orientation (°)	14°	28°	−28°	−28°	21°	12°	20°
Plan
WWR (%)	30	50	65	65	60	60	70
Model

and Figure 10. The population size and running time were as follows:

−

Generation size: 30, Generation count: 50

−

Population size: 30 × 50 = 1500 genomes

−

Time required for simulation: 17.5 h

The height of a classroom facing east should be controlled at 3.6~4.0 m. It is a rectangular space that is longer than it is wide, with a length range of 10~12 m and a width range of 9.5~11 m. The classroom orientation was optimal at 15~25°, and the WWR is controlled at 50~70%.

In order to make DF > 4.0%, the DF ranking has to be after 1471 (population: 0 to 1499 ranking). At this point, the data were as follows: DF fitness ranking = 1471; Gen. 1, Ind. 23; EUI: 89.18 kWh/m²; UDI: 68.97%; DF: 4.10%; DA:53.56%. Phenotype: H = 3.4 m, L = 8.8 m, W = 10.3 m, O = 25°, and WWR = 60%. There were few results for DF greater than 4.0%, and even fewer exceeding 5.0%. Most of them were concentrated in the first two generations. Although increasing the height of classroom could increase illumination and improved the DF and DA indexes, the energy consumption increased significantly, by 13.7% when compared with the Pareto Front Solutions. Therefore, it is important to pay careful attention to setting the classroom height. The optimization results always ended up grouped towards the bottom Parallel Coordinate Plot chart, but it was the early results that satisfied the real criteria according to the combined analysis of Section 4.2.1 and Section 4.2.2. This might point to its shortcoming of not being able to define a fixed range for the target.

4.3. Result from Large-Sized Classroom Optimization

4.3.1. West and South Facade Fenestration for Large Classrooms

The calculation process was the same as before, but it took longer. All results were analyzed and filtered in detail, and are in

Table 15. The case of large-sized classrooms with window openings on the west and south facade.

Large Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	87.24	91.96	118.54	118.54	94.75	97.51	101.31
UDI (%)	64.73	69.00	44.95	44.95	62.22	60.47	58.39
DF (%)	2.64	3.59	10.55	10.55	5.05	6.16	7.13
DA (%)	50.74	61.37	89.29	89.29	67.82	73.71	79.99
Population	Gen.42 Ind.0	Gen.2 Ind.1	Gen.44 Ind.2	Gen.44 Ind.2	Gen.29 Ind.12	Gen.31 Ind.16	Gen.42 Ind.15
Diamond Fitness Chart

and

Table 16. Genomes in detail and phenotype outputs for large-sized classrooms with west and south facade fenestration.

Large Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.4	4.3	4.4	4.4	3.4	3.4	3.5
Length (m)	19.9	16.3	12	12	19.7	16.8	18.4
Width (m)	15	13.4	11.1	11.1	15	14.9	11.1
Orientation (°)	13°	−20°	11°	11°	24°	9°	11°
Plan
WWR (%)	South:30	South:30	South:70	South:70	South:70	South:70	South:65
	West:35	West:30	West:70	West:70	West:50	West:60	West:65
Model

and Figure 11. The population size and running time were as follows:

−

Generation size: 30, Generation count: 50

−

Population size: 30 × 50 = 1500 genomes

−

Time required for simulation: 19.5 h

For a large-sized teaching space, the height of a classroom should be controlled at 3.4~3.6 m. It is a rectangular space longer than it is wide, with a length range of 15~20 m and a width range of 11~15 m. The classroom orientation was optimal at 10~25°, and the WWR of the south and west is controlled at 50~70% and 40~65%, respectively.

The SD value trendline showed an upward trend. Except for the DA fitness objective, which showed a top convergence phenomenon, the other three fitness values were somewhat dispersed. In order to make a DF index > 4.0%, the DF fitness ranking should be higher than 199 (0 to 1499 ranking). At this point, the data were as follows: DF fitness ranking = 199; Gen. 33, Ind. 11; EUI: 91.52 kWh/m²; UDI: 63.81%; DF: 4.01%; DA: 60.51%. Phenotype: H = 3.4 m, L = 19.3 m, W = 14.9 m, O = 18°, and WWR of south = 60%, WWR of west = 35%. All indicators had a good situation when the EUI index was around 95 kWh/m². Overall energy consumption could be controlled within 100 kWh/m². As seen from Pareto Set 3, when DF > 7.0%, the EUI index must be greater than 100 kWh/m². This lacked competition compared with other results.

4.3.2. East and South Facade Fenestration for Large Classrooms

The calculation process was also the same as the previous one, except that the calculation time was shorter at 18.8 h. All results were analyzed and filtered in detail, and are shown in

Table 17. The case of large-sized classrooms with window openings on the east and south facade.

Large Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
EUI (kWh/m2)	86.78	91.63	119.07	119.07	94.13	96.42	100.01
UDI (%)	63.41	70.09	44.42	44.42	63.18	62.45	61.18
DF (%)	2.43	4.13	10.57	10.57	5.02	5.43	6.01
DA (%)	48.13	67.24	89.09	89.09	67.63	71.06	76.85
Population	Gen.48 Ind.2	Gen.0 Ind.1	Gen.49 Ind.1	Gen.49 Ind.1	Gen.46 Ind.5	Gen.22 Ind.7	Gen.39 Ind.22
Diamond Fitness Chart

and

Table 18. Genomes in detail and phenotype outputs for large-sized classrooms with east and south facade fenestration.

Large Classroom	Min EUI	Max UDI	Max DF	Max DA	Pareto Set 1	Pareto Set 2	Pareto Set 3
Height (m)	3.4	3.8	4.4	4.4	3.8	3.8	4.1
Length (m)	20	19.6	12.2	12.2	20	19.6	12.2
Width (m)	14.9	10.5	10.9	10.9	14.8	13.8	14.8
Orientation (°)	−23°	9°	−27°	−27°	−17°	−24°	−25°
Plan
WWR (%)	South:30 East:30	South:30 East:40	South:70 East:70	South:70 East:70	South:55 East:50	South:60 East:50	South:30 East:70
Model

and Figure 12.

The height of a classroom here should be kept at 3.6~4.2 m. It is a rectangular space with a length range of 16~20 m and a width range of 12~15 m. The classroom orientation was optimal at −15~−25°, and the WWR of the south and east can be controlled at 40~60% and 40~70%, respectively.

It could be seen that the overall optimization result of the teaching spaces with east orientations was better than the west ones in the SD chart. In order to make a DF index > 4.0%, the DF fitness ranking should be higher than 139 (0 to 1499 ranking). At this point, the data were as follows: DF fitness ranking = 139; Gen. 13, Ind. 4; EUI: 92.62 kWh/m²; UDI: 64.12%; DF: 4.04%; DA: 60.89%. Phenotype: H = 3.5 m, L = 19.8 m, W = 14 m, O = −19°, and WWR of south = 65%, WWR of east = 30%. Therefore, the energy consumption stabilized at around 95 kWh/m², which was a relatively better balance point. Moreover, the situation of Pareto Set 3 could be fully counted as one of the best phenotypes of the medium-sized classroom. It was 12.2 m long and 14.8 m wide.

4.4. Summary of Results

To sum up, all optimization results for three types of classrooms were determined. We can observe that energy consumption was indeed reduced while maintaining the indoor lighting environment with the statistics from Section 4.1, Section 4.2 and Section 4.3. Some desired data could be visually derived from

Table 19. All possible ranges of optimization results for rectangular classrooms.

Classroom Type	Small Classroom		Medium Classroom		Large Classroom
Orientation	West	East	West	East	West and South	East and South
Height (m)	3.2~3.4	3.2~3.4	3.7~ 4.0	3.6~4.0	3.4~3.6	3.6~4.2
Length (m)	≈7.5	≈7.5	9.5~12	10~12	15~20	16~20
Width (m)	≈7.5	≈7.5	8~11	9.5~11	11~15	12~15
Orientation (°)	20~30°	−20~−30°	−20~−30°, 10~20°	15~25°	10~25°	−15~−25°
WWR (%)	50~65	50~70	55~70	50~70	South: 50~70	South: 40~60
					West: 40~65	West: 40~70
Expected EUI (kWh/m2)	100~105	95~100	90~100	88~95	93~100	90~95
Expected UDI (%)	65~70	60~75	65~75	65~70	55~65	60~65
Expected DF (%)	5.0~5.5	4.5~6.5	4.0~5.5	4.5~5.5	5.0~7.0	5.0~6.0
Expected DA (%)	60~65	65~70	70~80	65~80	65~80	65~80

Notes: An overview of six situations for different kinds of east–west-oriented teaching spaces in a humid subtropical climate (Shanghai, hot summer and cold winter climate zone).

in early design stages, and this provided a horizontal comparison of all the results.

• The reduction of energy consumption (EUI):

  −

  Small Classroom, West: 4.3~9.5%, East: 5.4~10.9%;

  −

  Medium Classroom, West: 0.9~12.1%, East: 9.7~18.5%;

  −

  Large Classroom, West: 9.4~17.6%, East: 15.1~18.9%;

  −

  Average decrease, West: 4.9~13.1%, East: 10.1~16.1%;

  −

  Total, 7.5~14.6%.

• The energy savings rate of east-facing classrooms compared with west-facing classrooms:

  −

  Small Classroom: almost identical;

  −

  Medium Classroom: 7.9~9.8%;

  −

  Large Classroom: 1.6~6.8%;

  −

  Average: 4.8~8.3%.

• The enhancement of the lighting environment:

  −

  UDI index: the improvements of small and large sizes were generally within 5%, or roughly the same as the previous. Medium classrooms could reach about 10%;

  −

  DF index: 20~35% improvement for small classrooms, 9% for medium classrooms, over 10% for the large classroom;

  −

  DA index: 6~20% improvement for small classrooms, approximately 14% for medium classrooms, 3.6~27.5% for the large classroom.

• Comparison with normal classrooms with north–south orientations:

  −

  Energy saving: the energy consumption of east–west-oriented teaching spaces could be reduced by 5.9%~26.4% with the same environmental factors.

  −

  Lighting environment: slightly better than normal classrooms in terms of UDI index, and slightly inferior (but still high) in terms of DF and DA indexes.

The research was effective because it calculated the optimal equilibrium state of each index for three different types of classrooms and derived the corresponding phenotype ranges that served as reference values for the design of east–west-oriented teaching spaces. Therefore, it helped to reduce energy consumption while ensuring the quality of the indoor lighting environment. This allows the environmental quality to catch up with the normal level (north–south direction) when the east–west-oriented teaching space has to be used.

Second, the results demonstrated that the algorithm based on multi-objective optimization was up to the task. As for genetic algorithm features, Wallacei X was able to perform multi-objective optimization that selects a batch of Pareto Fronts as the optimal solutions. It could store and optionally read potentially good results and output phenotypes. The three major lighting environment evaluation indexes, UDI, DF, and DA, are recognized as preferred in assessing indoor lighting environments, while the EUI index is a highly used index in building energy-saving designs. Therefore, this optimization workflow was shown as valid by using these four fitness values. The current work can be seen as an exploration and a step forward in response to the national policy call for energy-saving techniques and emission reduction.

5. Limitation

This study yielded interesting findings. Compared with conventional north–south-oriented teaching spaces, the quality of indoor light environments of optimized east–west-oriented spaces was not worse, but even when compared to the normal classrooms.

However, Wallacei X did not fully meet the needs of usage, see Section 4.2. It could only set a single maximum or minimum fitness value and was not able to establish a specific range for the target. In other words, the software optimized what it thought was feasible, but did not yield the result we expected. This may be considered a drawback because a higher classroom light environment is not always better, being that it can impose thermal stress on the users and be accompanied by higher energy consumption. The ongoing development of new algorithms appears to be fairly promising in terms of completing this work in the near future.

Following the results, subsequent studies will be conducted for varied spatial forms (that is, the space will no longer be rectangular), aiming toward new developments in the study of such problems. Furthermore, such spatial formations are likely to aid in the analysis of the layout of single buildings and group assemblages in schools from a local perspective, which can fill the gaps in ref. [34]. This also makes it an important issue for further research by studying the relationship between the whole and the local in the design of schools.

To sum up, the results of this research should be expanded further to test the effects in different climatic zones. In order to achieve this, we may tailor the genomes of the optimization goals to different climatic conditions and local regulations. Moderate cold and other passive strategies, as previously mentioned in refs. [7,8,9], are still significant in the environment to further reduce the amount of energy by diminishing reliance on active equipment. The final goal will be the same: to ensure the environmental quality of teaching spaces with less energy consumption.

6. Conclusions

This paper investigated finding the right shapes of east–west-oriented teaching spaces for occupants while safeguarding the indoor light environment and energy saving. Based on the Rhino and Grasshopper platforms, a quantitative framework for detection was built. It used many open-source plugins such as Ladybug, Honeybee, and Wallacei X. Based on the results and discussion, the following conclusions can be drawn:

(1)

The energy consumption of the optimized classrooms was reduced. Teaching spaces with an east–west orientation could save 7.5%~14.6% of energy after optimization. On average, spaces with west- and east-facing directions could save 4.9%~13.1% and 10.1%~16.1%, respectively. Furthermore, the arrangement of windows in easterly directions always consumed less energy than westerly ones, using 4.8%~8.3% less on average. This may be attributed to the sunlight coming from the west in the afternoon.

(2)

Surprisingly, the results showed a high energy-saving ratio for the optimization of medium-sized classrooms with the maximum number and frequency of usage because of their reasonable sizes.

(3)

The indoor lighting environment also showed satisfactory results. Three indexes (UDI, DF, and DA) all had different degrees of improvement. The enhancement of the UDI index was generally within 5%, while the overall data were above 60%, and sometimes even up to 75%; most of the DF indexes exceeded 5.0% for a satisfying indoor light environment; the improvement of the DA index was large, with values above 50% and some reaching 80%.

(4)

Compared with the normal classrooms, the energy consumption was reduced by 5.9%~26.4%. This was slightly better than the normal classrooms in terms of UDI index and slightly inferior (but still high) in terms of DF and DA indexes. The values provided are typically not high because all climate zones in China must be considered for the normal classrooms.

To some extent, this study altered our impression of the usage of east–west-oriented rooms and broadened our awareness of the possible use of east–west-oriented teaching spaces. This project can be used as a prototype to build a workflow to optimize such problems in early design stage that can save time and derive design results more rationally. However, there are still some issues that need further study. For example, we should consider the influence of the floor number of the classroom, indoor acoustic environment, and occupancy features. Of course, if better insulation materials and window types are used, energy consumption will be lower. Other options such as double glazing, insulating glass, and low-e glass can be taken into account based on the actual circumstances.