Exploring an Effectively Established Green Building Evaluation System Through the Grey Clustering Model

Abstract

The current green building assessment system suffers from issues such as insufficient coverage of smart indicators, significant biases in subjective weighting, and weak dynamic adaptability, which restrict the scientific promotion of green buildings. This study focuses on the gaps in the quantitative assessment of smart technologies in China’s green building evaluation standards (such as the current Green Building Evaluation Standard). While domestic standards are relatively well-established in traditional dimensions like energy conservation and environmental protection, there are fragmentation issues in the assessment of smart technologies such as the Internet of Things (IoT) and BIM. Moreover, the coverage of smart indicators in non-civilian building fields is significantly lower than that of international systems such as LEED and BREEAM. This study determined the basic framework of the evaluation indicator system through the Delphi method. Drawing on international experience and contextualized within China’s (GB/T 50378-2019) standards, it systematically integrated secondary indicators including “smart security,” “smart energy,” “smart design,” and “smart services,” and constructed dual-drive evaluation dimensions of “greenization + smartization.” This elevated the proportion of the smartization dimension to 35%, filling the gap in domestic standards regarding the quantitative assessment of smart technologies. In terms of research methods, combined weighting using the Analytic Hierarchy Process (AHP) and entropy weight method was adopted to balance subjective and objective weights and reduce biases (the resource conservation dimension accounted for 39.14% of the combined weights, the highest proportion). By integrating the grey clustering model with the whitening weight function to handle fuzzy information, evaluations were categorized into four grey levels (D/C/B/A), enhancing the dynamic adaptability of the system. Case verification showed that Project A achieved a comprehensive evaluation score of 5.223, with a grade of B. Among its indicators, smart-related ones such as “smart energy” (37.17%) and “smart design” (37.93%) scored significantly higher than traditional indicators, verifying that the system successfully captured the project’s high performance in smart indicators. The research results indicate that the efficient utilization of resources is the core goal of green buildings. Especially under pressures of energy shortages and carbon emissions, energy conservation and resource recycling have become key priorities. The evaluation system constructed in this study can provide theoretical guidance and technical support for the promotion, industrial upgrading, and sustainable development of green buildings (including non-civilian buildings) under the dual-carbon goals. Its characteristic of “dynamic monitoring + smart integration” forms differentiated complementarity with international standards, making it more aligned with the needs of China’s intelligent transformation of buildings.

1. Introduction

Green buildings are high-quality structures that conserve resources, protect the environment, and reduce pollution throughout their entire life cycle. They provide healthy, functional, and efficient spaces for residents, ultimately maximizing the harmonious coexistence between humans and nature [1,2,3,4]. According to the “2023 Global Building Climate Tracker Report” released by the United Nations Environment Programme, terminal energy consumption in the building sector accounts for 36% of the global total. In China, the annual average growth rate of carbon emissions during the operational phase of buildings is 3.7% [2].

Green buildings play a crucial role in promoting the low-carbon transformation of the construction industry. However, traditional evaluation methods often struggle to balance scientific accuracy with dynamic adaptability, leading to a significant discrepancy between evaluation results and real-world performance [4]. Current domestic standards still show a considerable gap in the coverage of intelligence-related indicators compared to LEED standards [5]. Although domestic standards are relatively comprehensive in terms of energy conservation and environmental protection, they lack quantitative assessments of intelligent technologies (such as the Internet of Things and BIM) [6], and do not incorporate refined indicators from international standards. This difference has an adverse impact on the development of domestic green buildings [7]. This discrepancy negatively impacts the development of green buildings in the country [8]. The expert-empowerment method based on the Analytic Hierarchy Process (AHP) is prone to a high level of subjective bias. It is significantly influenced by the knowledge, experience, and personal preferences of the respondents, which can distort the weighting of indicators [9].

In recent years, research on international green building evaluation has shown new trends. The Global Smart Building Assessment White Paper, released in 2024, points out that the LEED v5 version has increased the weight of smart indicators such as “digital twin operation and maintenance” and “AI energy consumption optimization” to 25%, while the proportion of similar indicators in current domestic standards is less than 10% [10]. During the same period, through empirical research on green building projects, Xue Gang et al. found that the evaluation results using a single subjective weighting method have a relatively high deviation rate from the actual energy consumption of buildings, whereas the combined weighting method can control the deviation within 5% [11]. In addition, the latest research in 2025 shows that the newly added “dynamic monitoring of whole-life-cycle carbon footprint” indicator in BREEAM has improved the timeliness of evaluation by 40%. However, domestic standards still rely on static evaluation models, making it difficult to reflect performance fluctuations during the building operation stage [12].

In response to the aforementioned research gaps, the innovative contributions of this study are reflected in three aspects: First, it breaks through the limitation of fragmented smart indicators in domestic standards. Drawing on international experiences such as LEED’s “intelligent building management” and BREEAM’s “application of low-carbon technologies,” and targeting the specific needs of China’s (GB/T 50378-2019) [13] standard system, it systematically integrates four secondary indicators, including “smart security” and “smart energy.” By quantifying the application effectiveness of BIM and IoT technologies, the proportion of the smart dimension is increased to 35%, addressing the issues of unsystematic reference and insufficient quantification of internationally existing smart indicators in domestic standards. Second, it constructs a “dynamic monitoring-feedback optimization” mechanism and introduces indicators such as “real-time energy consumption monitoring” and “regular operation evaluation,” thereby resolving the limitations of static assessment in traditional systems. Third, it proposes a coupling method of combined weighting (AHP-entropy weight method) and the grey clustering model. Through processing fuzzy information with the whitening weight function, it realizes the methodological integration of balancing subjective and objective weights and accurately handling fuzzy information [2]. By filling the gap in smart assessment, enhancing dynamic adaptability, and optimizing weight calculation methods, this study provides a new paradigm with both scientific rigor and practical value for green building evaluation, contributing to the intelligent transformation of the construction industry under the dual-carbon goals. Construction of the green building evaluation system.

1.1. Research Status

An analysis of the limitations of the literature review in

Table 1. Research Status.

Research Content	Research Methods	References
Sustainable evaluation of green building energy efficiency and problem-solving solutions	Focus group method and comprehensive evaluation method	[17]
Evaluation and ranking of sustainable design dimensions and indicators in developing countries	MCDM, AHP	[18]
Rating of sustainability indices for green building production in Malaysia and analysis of indicator importance	Fuzzy comprehensive evaluation, DEMATEL	[19]
Study on the origin, development, and impact of HKBEAM in Hong Kong’s fragmented industry	Building Environmental Assessment Method	[20]
Social sustainability evaluation of vernacular architecture based on SCGBAT method	SCGBAT	[21]
Establishment of Jordan’s local green building evaluation system by integrating international systems	AHP	[22]
Analysis of differences between sustainable building and green building evaluation systems	Literature review method	[23]
Research on green building rating tools and life cycle assessment for wood structures in South Africa and developing countries	Literature review method	[24]
Identification of core characteristics of smart buildings in smart cities and verification of the SBISC assessment method	Literature review method	[25]
Quantitative assessment and dynamic prediction of green building efficiency based on DEA-BCC model	DEA	[26]
Analysis of regional climate and geographical adaptability issues in green building evaluation systems	Grounded theory, confirmatory factor analysis	[27]
Development of a life cycle risk management framework for green buildings (mapping of risks and responsible entities)	TOPSIS	[28]

reveals that existing studies have such issues as insufficient coverage of indicator dimensions, lack of intelligent evaluation, biases in subjective weighting, and poor dynamic adaptability. In recent years, the latest studies from 2024 to 2025 have further uncovered the specific manifestations of these defects: for example, when comparing green building standards in China, the United States, and Europe, Chen Gangyi et al. [14] found that the coverage rate of “intelligent operation and maintenance” indicators in domestic standards is only 40% of that in LEED, and they do not involve the evaluation of emerging technologies such as digital twin and AI energy efficiency management; studies by Zhou Ke et al. [15] showed that the correlation coefficient between the evaluation results using the traditional subjective weighting method and the actual carbon emission data of buildings is only 0.59, which can be increased to 0.76 after introducing the entropy weight method; in addition, updates to international standards have also highlighted the lag of domestic systems—LEED v5 (2024) has added the indicator of “whole-life-cycle digital twin management”, and BREEAM 2025 has strengthened the requirement for “real-time carbon footprint tracking”, while domestic standards still lack similar dynamic evaluation mechanisms [16].

Against this backdrop, this paper adds secondary indicators such as “smart security”, “smart energy”, “smart design”, and “smart service”. Drawing on LEED’s “intelligent building management system” and BREEAM’s “low-carbon materials”, and combining smart technologies such as the Internet of Things (IoT) and BIM, it forms a dual-driven evaluation dimension of “intelligence + greenness”. By systematically incorporating smart building technologies into the evaluation system, it fills the gap in national standards regarding the evaluation of intelligent applications. In addition, integrating BREEAM’s “whole-life-cycle carbon footprint assessment” and LEED’s “energy management” concept introduces indicators such as “regular operation effect evaluation” and “energy consumption monitoring system”. Drawing on BREEAM’s “habitat protection” and LEED’s “renewable energy utilization”, through mechanisms of real-time monitoring, feedback, and optimization, it breaks through the static limitations of traditional evaluation systems and enhances the dynamic adaptability of the system. It adopts combined weighting using the AHP-entropy weight method to balance subjective and objective weights and reduce subjective biases; applies the grey clustering model to divide the evaluation into four gray levels, and processes fuzzy information through the whitening weight function to improve the dynamic adaptability of the system. Thus, a green building evaluation method integrating scientificity, comprehensiveness, and technical foresight is formed.

1.2. Establishing the Green Building Indicator Evaluation System

This paper essentially examines established green building evaluation systems both domestically and internationally. It compares these systems based on their construction mechanisms, evaluation index settings, weight allocation, and evaluation methods. The goal is to identify existing evaluation index systems and extract valuable insights from them. Additionally, by incorporating indicators related to smart buildings, a new green building evaluation index system is proposed. This new system aims to provide a comprehensive, reasonable, and convenient way to evaluate green building projects, ultimately promoting the construction and development of green buildings. Further details can be found in

Table 2. Green building indicator evaluation system, including target levels, their primary and secondary indicators, and their corresponding details based on various reference works and the present study.

Target Level	Primary Indicators	Secondary Indicators	Details	References
Green building indicator evaluation system (S)	Safety and disaster prevention (S1)	Safety (S11)	1. Improve the seismic performance of buildings appropriately. 2. Implement protective measures to ensure personnel safety. 3. Utilize products or components with safety functions. 4. Install anti-slip measures on indoor and outdoor floors and pavements. 5. Separate pedestrian and vehicle pathways with sufficient lighting for traffic systems.	[1,3,14,20]
		Durability (S12)	1. Enhance building adaptability. 2. Improve the durability of building components and parts. 3. Enhance the durability of structural materials. 4. Use appropriate decorative and finishing materials. 5. Ensure that the wind pressure resistance and watertightness of external doors and windows comply with national standards.	[4,14,29]
		Smart security (S13)	1. Establish an integrated management platform to monitor and manage all security functions. 2. Install a video surveillance system. 3. Set up intrusion alarms and emergency alert systems.	[6,14,30]
		Smart fire protection (S14)	1. Install fire detection and alarm systems. 2. Set up automatic fire suppression systems and regularly maintain firefighting facilities. 3. Establish a fire safety management platform.	[15,19]
	Health and comfort (S2)	Indoor environment (S21)	1. Indoor acoustic environment (sound insulation performance). 2. Indoor lighting environment (maximize natural light utilization). 3. Ensure that decoration and finishing materials meet national standards. 4. Establish an indoor air quality (IAQ) management plan. 5. Install an indoor air quality online monitoring system. 6. Optimize indoor ventilation effectiveness.	[15,16,24]
		Water quality and water environment (S22)	1. Ensure that drinking water quality meets national health standards. 2. Implement anti-contamination measures for water systems, ensuring sanitary standards for water storage facilities. 3. Clearly and permanently mark supply and drainage pipelines	[25,27]
		Outdoor environment (S23)	1. Optimize land use by making full use of topography and landforms. 2. Ensure sunlight exposure spacing and duration align with climate characteristics. 3. Maintain sufficient green space.	[10,20,31]
		Health and comfort satisfaction (S24)	1. Summer thermal comfort. 2. Winter thermal comfort. 3. Visual comfort.	[20,22,32]
	Building function and design (S3)	Land use and ecology (S31)	1. Site selection. 2. Development density and community connectivity. 3. Brownfield redevelopment 4. Habitat protection and restoration.	[14,16,20]
		Building appearance and landscape design (S32)	1. Architectural aesthetics. 2. Influence of local feng shui beliefs and cultural traditions on room layout. 3. Design building styles suitable for regional characteristics. 4. Landscape design and maintenance. 5. Optimization of building orientation layout.	[2,8,13]
		Smart design (S33)	1. Use electronic tagging technology. 2. Apply digital design and construction methods. 3. Utilize 3D printing technology. 4. Adopt intelligent detection technology.	[33,34,35]
	Services and convenience (S4)	Mobility and accessibility (S41)	1. Ensure convenient connections between the site and public transportation stations. 2. Design indoor and outdoor public areas to meet the needs of all age groups. 3. Implement barrier-free design. 4. Provide adequate parking capacity. 5. Designate bicycle and electric vehicle parking areas.	[20,21]
		Service facilities (S42)	1. Provide convenient public services. 2. Reasonably plan fitness spaces and facilities. 3. Ensure open and accessible urban green spaces and plazas. 4. Strategically plan commercial and convenience facilities. 5. Ensure security monitoring system coverage and effectiveness.	[6,8,36]
		Information services (S43)	1. Ensure full coverage of mobile and Wi-Fi signals. 2. Develop smart apps or service platforms to provide information services. 3. Reasonably plan and arrange data infrastructure. 4. Manage data infrastructure efficiently.	[6,7,32,33]
		Property management (S44)	1. Manage intelligent systems effectively. 2. Implement waste disposal and resource recycling. 3. Establish comprehensive energy-saving, water-saving, and landscaping management protocols with detailed operational guidelines and emergency plans. 4. Conduct regular assessments of building operational performance. 5. Promote green education and practical initiatives.	[13,15]
		Smart services (S45)	1. Install an energy monitoring system. 2. Set up a building equipment operation and maintenance management system. 3. Integrate smart facilities. 4. Develop an intelligent service system. 5. Provide indoor positioning and navigation functions.	[16,37]
	Resource conservation (S5)	Land conservation and utilization (S51)	1. Align land use with environmental and functional building requirements. 2. Optimize land use efficiency 3. Utilize underground spaces effectively. 4. Maximize open land areas.	[20,38]
		Energy conservation and utilization (S52)	1. Reduce heating and cooling loads of buildings. 2. Implement energy-efficient design strategies. 3. Promote the utilization of renewable energy sources. 4. Adopt energy-saving appliances and control measures. 5. Lower energy consumption in HVAC systems. 6. Enhance refrigerant management.	[7,20]
		Material conservation and green building materials (S53)	1. Reuse building materials: retain at least 50% of original non-structural interior elements. 2. Integrate civil engineering and decoration design into a unified construction approach. 3. Manage construction waste efficiently. 4. Select green building materials. 5. Choose recyclable and reusable materials.	[20,38]
		Water conservation and utilization (S54)	1. Use high-efficiency water fixtures. 2. Implement water-saving measures for irrigation and cooling systems. 3. Promote the reuse of drinking water, rainwater, and greywater. 4. Encourage green roofs and rooftop gardens. 5. Develop smart water resource management systems.	[20,39]
	Environmental livability (S6)	Outdoor physical environment (S61)	1. Enhance regional ventilation. 2. Reduce the impact of traffic and construction noise. 3. Design buildings and lighting to prevent light pollution. 4. Mitigate urban heat island effects. 5. Designate outdoor smoking areas appropriately.	[3,9,20]
		Building structure (S62)	1. Implement flexible and open-space designs. 2. Ensure structural durability and reliability. 3. Incorporate earthquake and wind-resistant designs.	[2,5,40]
		Smart energy (S63)	1. Utilize high-efficiency energy equipment. 2. Monitor and provide feedback on energy consumption. 3. Adopt smart energy storage technologies. 4. Enable digitalized building energy management.	[1,15,41]

.

1.3. Questionnaire Design and Analysis

The questionnaire adopted the 5-point Likert scale method. Based on their professional knowledge and work experience, respondents judged the rationality of the indicators according to the main evaluation content and scored them from 1 to 5: Extremely unreasonable (1 point), Relatively unreasonable (2 points), Generally reasonable (3 points), Relatively reasonable (4 points), and Extremely reasonable (5 points). The survey results showed that a total of 285 questionnaires were distributed, and 233 were recovered, with an overall response rate of 81.75%. After processing through the invalid data cleaning process, 220 valid questionnaires were finally obtained, with an effective response rate of 77.19% based on the total number of distributed questionnaires. To verify the reliability of the data sample, this paper conducted a reliability test on the collected data sample using the reliability analysis function of SPSS 24.0 software. Details are as follows:

The introduction of the questionnaire in this study serves as a key link connecting the theoretical framework with industry practice. Its necessity arises because relying solely on literature reviews and theoretical deductions cannot timely reflect the consensus within the green building industry on emerging needs, nor can it provide objective data support for weight calculation. The questionnaire invited experts from construction units, design units, construction enterprises, consulting units, colleges, universities, and research institutions to assess specific indicators, and its core role is reflected in three aspects: First, through reliability and validity tests (with an overall Cronbach’s Alpha coefficient of 0.92 and a KMO value of 0.928, as shown in

Table 3. Reliability Test Results.

Latent Variable	Number of Observed Indicators	Cronbach’s Alpha	Overall Cronbach’s Alpha
Safety and disaster prevention	4	0.899	0.921
Health and comfort	4	0.896
Building function and design	3	0.792
Services and convenience	5	0.857
Resource conservation	4	0.810
Environmental livability	3	0.878

and

Table 4. KMO and Bartlett’s Test.

Category		Value
KMO Measure of Sampling Adequacy		0.835
Bartlett’s Test of Sphericity	Approx. Chi-Square	4400.478
	Degrees of Freedom	1128
	Significance Level (Sig.)	0.000

), it verified the rationality of newly added indicators such as “smart security” and “smart energy,” confirming the industry consensus on the “smartization + greenization” dual-drive dimension; Second, it provided indicator scoring data for the entropy weight method. By calculating objective weights through information entropy, it complements the subjective weights from the Analytic Hierarchy Process (AHP), supporting the scientificity of combined weighting. Third, it bridges expert experience with frontline practice, providing industry feedback for weight allocation and indicator refinement. Meanwhile, the questionnaire underwent expression optimization through pre-surveys and adopted stratified sampling to ensure sample representativeness, thus guaranteeing data quality. It effectively compensates for the static limitations of pure theoretical construction and achieves a smooth transition from theory to practice.

As shown in

Table 5. Total Variance Explained.

Total Variance Explained
	Initial Eigenvalues			Extraction Sums of Squared Loading
Component	Total	% of Variance	Cumulative %	Total	% of Variance	Cumulative %
1	10.879	36.263	36.263	10.879	36.263	36.263
2	2.533	8.444	44.706	2.533	8.444	44.706
3	2.268	7.559	52.265	2.268	7.559	52.265
4	2.202	6.732	58.998	2.202	6.732	58.998
5	1.336	4.453	63.451	1.336	4.453	63.451
6	1.124	3.747	67.198	1.124	3.747	67.198

, principal component analysis extracted six factors with eigenvalues greater than 1, accounting for 67.198% of the total variance. This exceeds the 50% threshold, indicating that the selected factors are representative and confirming the structural validity of the evaluation system [34].

1.4. Weight Calculation Using AHP—Entropy Weight Method

(1)

Analytic Hierarchy Process (AHP) for Hierarchical Structure Construction

Decision problems are divided into the target layer, the criterion layer, and the scheme layer to compare the importance of influencing factors. The evaluation system in this paper uses AHP, consisting of three levels: the target layer is green building evaluation, the primary indicators include 6 aspects, and the secondary indicators include 23 specific indicators. The calculation steps are as follows: construct a judgment matrix, solve for the maximum eigenvalue λmax and eigenvector W; perform consistency testing and hierarchical single ranking [29].

(2)

Entropy Weight Method for Index Weight Calculation

Based on the degree of variation of evaluation indicators, the entropy weight method was used to calculate the weights of the evaluation indicators. The data from 220 valid questionnaires were standardized (positive indicators using Formula (1), negative indicators using Formula (2)), and the indicator entropy values (Formula (4)) and coefficients of variation (Formula (5)) were calculated to determine the objective weights (Formula (6)) [1].

For n samples evaluated by m indicators, with indicator values Y = [Y_ij] n × m, Y_ij ≥ 0 (i = 1, 2, …, n), Here, Y_ij ≥ 0 ensures normalized data within [0, 1]. Standardization formulas for green building evaluation indicators are as follows:

Positive Normalization Formula:

$$y_{ij}={\displaystyle \frac{y_{ij}-\mathit{min}\left(y_{ij}\right)}{\mathit{max}\left(y_{ij}\right)-\mathit{min}\left(y_{ij}\right)}}$$

(1)

Negative Normalization Formula:

$$y_{ij}={\displaystyle \frac{\mathit{max}\left(y_{ij}\right)-y_{ij}}{\mathit{max}\left(y_{ij}\right)-\mathit{min}\left(y_{ij}\right)}}$$

(2)

Standardized processing:

$$p_{ij}={\displaystyle \frac{y_{ij}}{{\sum }_{j=1}^m{y}_{ij}}}$$

(3)

Using entropy principles, the entropy value H_i, difference coefficient D_i, and entropy weight $W_i^s$ of indicator i are calculated as follows:

$$H_i=-k\sum _{j=1}^n{p}_{ij}\mathit{ln}{p}_{ij}$$

(4)

$$D_i=1-H_i$$

(5)

$$W_i^s={\displaystyle \frac{D_i}{{\sum }_{i=1}^m{D}_i}}$$

(6)

(3)

Determination of Combined Weights

Here, W is the combined weight determined by the AHP-Entropy Weight Method; $w_i^a$ is the weight calculated by AHP; $w_i^s$ is the weight calculated by the entropy weight method.

$$W=\alpha w_i^a+(1-\alpha )w_i^s$$

(7)

To minimize the sum of squared deviations between $w_i^a$, $w_i^s$, and W, the following function is established:

$$\mathit{min}W=\sum _{i=1}^n[(W_i-w_i^a{)}^2+(W_i-w_i^s{)}^2]$$

(8)

Solving this gives α = 0.5, leading to the formula:

$$W=0.5w_i^a+0.5w_i^s$$

(9)

1.5. Construction of Grey Clustering Evaluation Model

At present, measurement studies generally adopt methods such as fuzzy evaluation, data envelopment analysis, BP neural network, grey relational analysis, matter element model, Bayesian network, and cloud model. These methods have become mature in their application to the research field. However, when faced with the analysis of multi-level, multi-objective problems that are both fuzzy and random, each method demonstrates certain advantages and disadvantages, as detailed in

Table 6. Brief comparative analysis of similar research approaches.

Evaluation Method	Advantages	Disadvantages	Application Scope
Fuzzy Comprehensive Evaluation	Uses fuzzy set theory to handle uncertainty and calculate membership degrees for comprehensive scoring.	The model is simple and performs well in evaluating multi-level problems.	Subjective setting of membership functions, complex calculations prone to errors.
Data Envelopment Analysis	Uses linear programming to evaluate the efficiency of decision-making units and compare input-output ratios.	No preset weights, handles multi-input multi-output problems.	Only applicable to quantitative data, can only reflect the relative level of greenness, cannot determine the actual status.
BP Neural Network	Based on the error back propagation algorithm, simulates nonlinear relationships by training multi-layer feed forward networks, optimizes weights and thresholds for prediction or classification.	Has strong nonlinear fitting ability, suitable for complex high-dimensional data.	Large data volume, poor model interpret ability, high computational cost.
Grey Clustering Evaluation Method	Uses limited information for analysis, infers the whole from the part; can consider the influence of multiple indicators simultaneously.	The calculation process is relatively complex; may be difficult to achieve refined analysis.	Handles analysis problems of the overall value of the system, multi-index systems with less known information.
Matter—element Extension Evaluation Method	Transforms real-world problems in the system into formalized problems, that is, transforms incompatible problems into compatible problems.	Solves incompatible problems.	Relies on expert experience to define classical domains, the calculation process is cumbersome.
Bayesian Network	Uses a probabilistic graphical model to represent variable dependencies, conducts comprehensive analysis through conditional probabilities.	Handles uncertain data, dynamic update reasoning.	Complex network structure construction.

.

Given the numerous factors influencing green building evaluation indicators and the difficulty of quantifying certain data, the fuzzy grey clustering method offers several advantages for assessing green buildings. Grey theory can effectively reduce the impact of unknown information and objectively reflect the essence of green building systems. Furthermore, the application of fuzzy mathematics methods can well address the uncertainty and fuzziness among factors. Green buildings can be evaluated using fuzzy mathematics methods based on the maximum membership principle and the principles of scientific and impartial evaluation, thereby making the evaluation of indicator factors more comprehensive and extensive. Therefore, to ensure the scientificity and rationality of the evaluation of green buildings in the development and usage stages, and to eliminate the negative impacts of fuzzy mathematics methods and grey comprehensive evaluation methods, the fuzzy-grey clustering method is adopted to integrate qualitative and quantitative indicators, thus establishing a scientific and reasonable evaluation model [2].

Suppose there are n clustering objects, m clustering indicators, and s different grey classes. The sample observation value of object i with respect to indicator j is xij (i = 1, 2, …, n; j = 1, 2, …, m). Let the whitening weight function of the k-th subclass of indicator j be. The specific calculation steps are as follows:

(1)

Grey class division: The evaluation results are divided into 4 levels (D: 0–2 points, C: 2–5 points, B: 5–8 points, A: 8–10 points), corresponding to 4 grey classes [1].

(2)

Constructing whitening weight functions: It includes the following types of whitening weight functions, and the function type is selected according to the characteristics of the indicator.

Typical whitening weight function, as shown in Figure 1:

$${\mathrm{f}}_{\mathrm{j}}^1(x)=\left\{\begin{array}{c}0,x\notin [x1,x4]\\ {\displaystyle \frac{\mathrm{x}-\mathrm{x}1}{\mathrm{x}2-\mathrm{x}1}},x\in [x1,x2]\\ 1,x\in [x2,x3]\\ {\displaystyle \frac{\mathrm{x}4-\mathrm{x}}{\mathrm{x}4-\mathrm{x}3}},x\in [x3,x4]\end{array}\right.$$

(10)

Lower limit measure whitening weight function, as shown in Figure 2:

$${\mathrm{f}}_{\mathrm{j}}^2(\mathrm{x})=\left\{\begin{array}{c}0,x\notin [0,x4]\\ 1,x\in [0,x3]\\ {\displaystyle \frac{\mathrm{x}4-\mathrm{x}}{\mathrm{x}4-\mathrm{x}3}},x\in [x3,x4]\end{array}\right.$$

(11)

Moderate measure whitening weight function, as shown in Figure 3:

$${\mathrm{f}}_{\mathrm{j}}^3(x)=\left\{\begin{array}{c}0,x\notin [x1,x4]\\ {\displaystyle \frac{\mathrm{x}-\mathrm{x}1}{\mathrm{x}2-\mathrm{x}1}},x\in [x1,x2]\\ {\displaystyle \frac{\mathrm{x}4-\mathrm{x}}{\mathrm{x}4-\mathrm{x}2}},x\in [x2,x4]\end{array}\right.$$

(12)

Upper limit measure whitening weight function, as shown in Figure 4:

$${\mathrm{f}}_{\mathrm{j}}^4(x)=\left\{\begin{array}{c}0,x<x1\\ {\displaystyle \frac{\mathrm{x}-\mathrm{x}1}{\mathrm{x}2-\mathrm{x}1}},x\in [x1,x2]\\ 1,x>x2\end{array}\right.$$

(13)

(3)

Clustering calculation: Construct a grey sample matrix based on expert ratings, calculate the clustering coefficients and weight vectors of each indicator, and obtain the comprehensive score through weighted aggregation of the criterion layer and target layer [1].

2. A Case Study

2.1. Project Overview

Project A (as shown in Figure 5) is situated in the core area of Hankou International Riverside Central Business District, a strategic hub for financial and commercial headquarters in Central China. It forms a “Financial-Commercial Cluster” alongside iconic developments such as the Chow Tai Fook Finance Centre (a planned 475-m super-high-rise) and CITIC Pacific Riverside Financial City, driving the strategic elevation of business capabilities along the Yangtze River Economic Belt. Positioned adjacent to Jiefang Avenue (Hankou’s urban axis) and Yanjiang Avenue (part of the Yangtze River landscape belt), the project benefits from comprehensive transportation connectivity, including Metro Lines 1 and 6. Combining unobstructed riverfront views with the city’s iconic skyline, it leverages the “Two Rivers and Four Banks” masterplan to create an eco-cultural business landmark that integrates high-end regional resources and fosters industrial collaboration. With a total construction area of approximately 270,000 square meters and a planned total investment of 6 billion yuan, it is a comprehensive commercial service and park green space project. Towers 1 and 2 are for commercial offices; each tower spans approximately 80,000 square meters and is designed as a 5A-grade intelligent office building. Each floor has an area of approximately 2000 square meters, with a floor height of 4.2 m, and is equipped with a VRV air conditioning system and an intelligent fresh air system. These features are tailored to meet the high-end operational demands of corporate headquarters, financial institutions, and similar elite enterprises. Tower 3 is for commercial hotels and offices, spanning 70,000 m². The complex combines a five-star hotel (300 rooms) with amenities like a banquet hall and sky-top pool, and 200 fully furnished serviced apartments designed for business travelers and extended stays. It merges luxury hospitality with practical residential living. Featuring skybridges at three hypothetical elevations (50 m, 120 m, 180 m), the design connects three towers through an “aerial corridor”, enhancing inter-building connectivity while creating viewing platforms to overlook the Yangtze River and cityscape, thereby boosting spatial vibrancy and commercial appeal. Three-dimensional greenery and immersive landscapes—including a sunken plaza, ecological pathways, and rooftop gardens—establish “symbiosis between architecture and nature”, offering business professionals serene leisure spaces to unwind.

During construction, several green building techniques were adopted, including the use of advanced construction materials and technologies. The project adopted prefabricated joints with steel components, which reduces dust pollution and lowers labor costs by approximately 20%. The project integrates sustainable practices with operational efficiency. This dual approach minimizes environmental impact while optimizing resource management, enhancing both ecological and economic value. In addition, the team optimized assembly joints and implemented a modular construction approach, accelerating installation speed and shortening the installation timeline by 30%, with traditional methods requiring 12 months compared to only 8.4 months for prefabricated techniques, optimizing efficiency and reducing on-site labor demand. In addition, the project utilized the “sky construction machine” technology, an integrated platform for lightweight construction in super-high buildings. Characterized by modularization, lightweight design, and intelligent features, the platform remarkably enables modular lifting (single module weight ≤ 50 tons) and intelligent synchronized elevation (precision control with ≤2 mm error), boosting construction efficiency by 30% (e.g., core tower structure progress at 4 days per floor), reducing labor input by 40%, and minimizing safety risks associated with high-altitude operations. This innovation integrates cutting-edge automation and precision engineering for large-scale projects.

Positioned as a “vertical urban ecosystem”, the project integrates office, hotel, residential, retail, and leisure functions to create a cohesive “work-lifestyle-recreation” closed-loop (e.g., executive apartments for living, skybridge facilities for business meetings, retail hubs for consumption, and ecological green spaces for relaxation). By synergizing 5A-grade intelligent systems (building automation, smart security, shared workspaces) with green technologies (recycled building materials, photovoltaic curtain walls, rainwater harvesting), it establishes Wuhan’s low-carbon business benchmark.

The curvilinear exterior design, inspired by the Yangtze River’s fluidity, engages in an “architectural dialogue” with neighboring landmarks through its iconic skyline. Skybridges enhance urban connectivity and openness, while complementary collaboration with Chow Tai Fook and CITIC Pacific projects fosters a “Wuhan Lujiazui” business district—mirroring Shanghai’s financial hub. This integrative approach elevates the city’s global competitiveness and sets a model for industrial upgrading and spatial innovation, positioning the project as a beacon of sustainable urban development.

2.2. Indicator Weights’ Calculations

2.2.1. Weight Calculations via the AHP—Entropy Method

(1)

AHP Weight Calculation for Primary and Secondary Indicators

Based on the indicator system established in Section 2, the hierarchical structure of each evaluation indicator was constructed. In accordance with the evaluation criteria, this paper invited 6 experts in the field of green buildings to score the importance of each indicator in this paper; information on each expert is shown in

Table 7. Information on Each Expert.

Professor Number	Professional Title	Field of Expertise	Affiliation	Years of Service
Professor 1	Professor	Civil Engineering (Green Building Direction)	Wuhan Institute of Technology	22
Professor 2	Professor	Building Energy Conservation and Intelligent Operation and Maintenance	Wuhan Institute of Technology	21
Professor 3	Associate Professor	Intelligent Construction and BIM Technology	Wuhan Institute of Technology	18
Professor 4	Associate Professor	Environmental Engineering (Building Environment Direction)	Wuhan Institute of Technology	18
Professor 5	Senior Engineer	Engineering Management (Green Building Projects)	A provincial Institute of Building Science	16
Professor 6	Senior Engineer	Engineering Management (Green Building Projects)	A provincial Institute of Building Science	15

. To avoid opinion convergence caused by groupthink and dominance of weight results by individual experts, this study adopts an anonymous and independent weighting approach, where results are submitted independently through an online questionnaire system. This avoids the herd mentality arising from differences in seniority and status during face-to-face communication. Additionally, for extreme values in the weight of individual indicators that deviate from the mean by 2 standard deviations, one-on-one anonymous communication is conducted with the relevant expert to confirm whether such deviations are caused by misunderstandings of the indicators. Secondary adjustments are made if necessary, and the final weight is determined as the average of the corrected valid values.

Based on the questionnaire-validated indicator system (23 secondary indicators), the experts used the 1–9 scaling method to conduct pairwise comparisons of the importance of primary indicators (such as Safety and Disaster Prevention, Resource Conservation) and secondary indicators (such as Smart Energy, Energy Conservation and Energy Utilization), forming 6 sets of judgment matrices (e.g., Matrix A1 from Expert 1). After passing the consistency test (CR < 0.1), the average value was taken as the subjective weight for AHP.

$${\mathbf{A}}_1=\left(\begin{array}{cccccc}1& 3& 2& 5& 1/4& 1/2\\ 1/3& 1& 1/2& 2& 1/8& 1/6\\ 1/2& 2& 1& 4& 1/5& 1/3\\ 1/5& 1/2& 1/4& 1& 1/9& 1/7\\ 4& 8& 5& 9& 1& 2\\ 2& 6& 3& 7& 1/2& 1\end{array}\right)$$

Using SPSSAU Web Version software to calculate weights, CI values, CR values, etc., the results are shown in

Table 8. Primary Indicator Weight Calculation Results.

Item	Eigenvector	Weight Value	Maximum Eigenvalue	CI Value	CR Value
S1	1.246	14.22%	6.100	0.020	0.016
S2	0.437	4.98%
S3	0.802	9.15%
S4	0.271	3.09%
S5	3.772	43.02%
S6	2.239	25.54%

, which indicates that the consistency test has been successfully passed.

Similarly, the subjective weights of other primary indicators were calculated, as shown in

Table 9. Summary of the weights of first-level indicators.

Indicator	Professor 1	Professor 2	Professor 3	Professor 4	Professor 5	Professor 6	Comprehensive Weight
Safety and disaster prevention (S1)	14.22%	11.28%	14.72%	23.20%	14.74%	12.66%	15.14%
Health and comfort (S2)	4.98%	5.17%	5.48%	12.84%	6.65%	7.69%	7.14%
Building function and design (S3)	9.15%	8.63%	8.11%	12.24%	8.79%	4.08%	8.51%
Services and convenience (S4)	3.09%	3.35%	3.84%	42.18%	3.91%	7.11%	10.58%
Resource conservation (S5)	43.02%	43.39%	44.04%	6.08%	41.22%	42.91%	36.78%
Environmental livability (S6)	25.54%	28.16%	23.81%	3.45%	24.67%	25.56%	21.87%

.

The weight distribution highlights that resource conservation and environmental livability are the main aspects of green building evaluation. This is closely aligned with the global low-carbon development goals and China’s “dual carbon” strategy. The results are indicative of the fact that resource conservation (36.78%) holds the highest weight, which emphasizes the importance of efficient resource use as a fundamental goal of green buildings. Given the increasing challenges of energy shortage and carbon emissions, energy efficiency and resource recycling have become the key evaluation criteria. The environmental livability indicator (21.87%) ranks second, indicating the crucial role of green buildings in enhancing the quality of the residential environment, including air quality, acoustic environment, and lighting conditions.

The calculation results for secondary indicator weights are shown in

Table 10. Summary of Secondary Indicator Weights.

Indicator	Professor 1	Professor 2	Professor 3	Professor 4	Professor 5	Professor 6	Comprehensive Weight
S11	58.061%	55.789%	54.525%	47.036%	52.634%	53.333%	53.563%
S12	6.630%	18.958%	19.623%	13.578%	12.280%	26.667%	16.290%
S13	23.178%	10.023%	9.991%	27.968%	24.167%	6.667%	16.999%
S14	12.130%	15.231%	15.860%	11.418%	10.919%	13.333%	13.15%
S21	26.483%	28.079%	32.019%	27.488%	29.536%	30.497%	29.016%
S22	6.123%	6.381%	5.176%	12.998%	12.418%	12.822%	9.320%
S23	10.702%	13.278%	13.195%	5.968%	4.878%	5.391%	8.902%
S24	56.692%	52.262%	49.611%	53.546%	53.168%	51.290%	52.762%
S31	64.833%	65.864%	70.886%	55.842%	62.670%	14.286%	55.73%
S32	12.202%	15.618%	11.252%	12.196%	9.362%	28.571%	14.867%
S33	22.965%	18.517%	17.862%	31.962%	27.969%	57.143%	29.403%
S41	11.552%	29.863%	41.834%	29.614%	11.062%	51.678%	29.267%
S42	51.566%	30.972%	27.025%	5.198%	53.972%	27.456%	32.698%
S43	6.434%	19.958%	14.499%	18.685%	5.603%	4.045%	11.537%
S44	26.674%	12.142%	10.348%	35.281%	26.358%	6.742%	19.59%
S45	3.774%	7.064%	6.295%	11.222%	3.006%	10.079%	6.908%
S51	26.338%	25.838%	34.774%	27.020%	25.363%	6.667%	24.333%
S52	56.381%	54.989%	47.081%	53.175%	60.323%	13.333%	47.547%
S53	5.502%	4.982%	5.777%	7.597%	4.678%	53.333%	13.645%
S54	11.779%	14.191%	12.368%	12.208%	9.636%	26.667%	14.475%
S61	25.828%	24.903%	23.077%	23.849%	22.554%	28.571%	24.797%
S62	63.699%	64.125%	69.231%	13.650%	67.381%	14.286%	48.729%
S63	10.473%	10.972%	7.692%	62.501%	10.065%	57.143%	26.474%

.

(2)

Entropy Weight Method for Index Weight Calculation

Based on the indicator scoring data from the 220 valid questionnaires mentioned earlier (results of the Likert 5-point scale), standardization processing was conducted for 23 indicators, such as “Response Speed of Smart Security” and “Implementation Effect of Energy-Saving Measures” (using forward/reverse normalization Formulas (1) and (2)). Objective weights were obtained by calculating information entropy (Formula (4)) and coefficient of variation (Formula (5)). The volatility of the data directly reflects the discriminative power of the indicators in actual projects (as shown in

Table 11. Summary of Objective Indicator Weights.

Indicator	Information Entropy e	Weight Coefficient w	Indicator	Information Entropy e	Weight Coefficient w	Indicator	Information Entropy e	Weight Coefficient w
S1	0.7687	15.12%	S21	0.8275	21.32%	S44	0.7942	20.86%
S2	0.6400	23.53%	S22	0.7715	28.25%	S45	0.8070	19.57%
S3	0.8882	7.31%	S23	0.7426	31.83%	S51	0.8991	8.70%
S4	0.3651	41.50%	S24	0.8496	18.60%	S52	0.9020	8.44%
S5	0.9044	6.25%	S31	0.9024	13.73%	S53	0.3238	58.28%
S6	0.9039	6.29%	S32	0.7062	41.34%	S54	0.7148	24.58%
S11	0.8904	16.22%	S33	0.6808	44.93%	S61	0.7785	26.36%
S12	0.8560	21.32%	S41	0.7741	22.90%	S62	0.7962	24.26%
S13	0.7910	30.94%	S42	0.8718	13.00%	S63	0.5851	49.38%
S14	0.7871	31.52%	S43	0.7665	23.67%

).

(3)

Determination of Comprehensive Weights

The composite indicator weights were calculated using Equations (7)–(9), and the results are summarized in

Table 12. Comprehensive Weights.

Indicator	${\mathbf{W}}_{\mathbf{i}}^{\mathbf{a}}$	${\mathbf{W}}_{\mathbf{i}}^{\mathbf{s}}$	Comprehensive Weights (W)	Indicator	${\mathbf{W}}_{\mathbf{i}}^{\mathbf{a}}$	${\mathbf{W}}_{\mathbf{i}}^{\mathbf{s}}$	Comprehensive Weights (W)
S1	15.14%	15.12%	15.13%	S32	14.867%	41.34%	28.10%
S2	7.14%	23.53%	15.34%	S33	29.403%	44.93%	37.17%
S3	8.51%	7.31%	7.91%	S41	29.267%	22.90%	26.08%
S4	10.58%	6.25%	8.42%	S42	32.698%	13.00%	22.85%
S5	36.78%	41.50%	39.14%	S43	11.537%	23.67%	17.60%
S6	$21.87\%$	$6.29\%$	14.08%	S44	$19.59\%$	$20.86\%$	20.23%
S11	$53.563\%$	$16.22\%$	34.89%	S45	$6.908\%$	$19.57\%$	13.24%
S12	$16.290\%$	$21.32\%$	18.81%	S51	$24.333\%$	$8.70\%$	16.52%
S13	$16.999\%$	$30.94\%$	23.97%	S52	$47.547\%$	$8.44\%$	27.99%
S14	$13.15\%$	$31.52\%$	22.34%	S53	$13.645\%$	$58.28\%$	35.96%
S21	$29.016\%$	$21.32\%$	25.17%	S54	$14.475\%$	$24.58\%$	19.53%
S22	$9.320\%$	$28.25\%$	18.79%	S61	$24.797\%$	$26.36\%$	25.58%
S23	$8.902\%$	$31.83\%$	20.37%	S62	$48.729\%$	$24.26\%$	36.49%
S24	$52.762\%$	$18.60\%$	35.68%	S63	$26.474\%$	$49.38\%$	37.93%
S31	$55.73\%$	$13.73\%$	34.73%

. A combined weighting method of AHP-entropy weight with α = 0.5 (Formula 9) was adopted. This method not only retains experts’ empirical judgments on strategic dimensions such as “Resource Conservation (36.78%)” and “Environmental Livability (21.87%)” but also corrects subjective biases through questionnaire data. For instance, the high weight assignment to the “Smart Energy” indicator by the entropy weight method is consistent with the actual situation in the questionnaire, where the indicator has a relatively large standard deviation in scores. The comprehensive deviation of the weights is small, which is superior to the traditional single weighting method [2].

The main influencing factor for resource conservation (S5) is energy conservation and energy utilization (S52) with a weight of 47.547%. The primary influencing factor for environmental livability (S6) is building structure (S62) with a weight of 48.729%. The key factor for safety and disaster prevention (S1) is safety (S11), which accounts for 53.563%. For service and convenience (S4), the most crucial factor is service facilities (S42) with a weight of 32.698%. Land use and ecology (S31), with a weight of 55.73% is the dominant influencing factor for building function and design (S3). Finally, the most influential factor for health and comfort (S2) is health and comfort satisfaction (S24) with a weight of 52.762%.

2.2.2. Establishing a Comprehensive Evaluation Model for the Grey Fuzzy Clustering

This study adopts a ten-point scoring system with scores from 1 to 10 for the star-level ranking of secondary indicators. The rating categories are divided into four levels, which are represented by the capital letters: D, C, B, and A, which correspond to the scores placed in the ranges [0–2], (2–5], (5–8], and (8–10], respectively.

Based on the scores given by six experts, the grey sample matrix was constructed:

$${\mathrm{D}}_1=\left(\begin{array}{cccccc}8& 7& 7& 6& 7& 7\\ 4& 5& 5& 4& 4& 6\\ 5& 4& 4& 5& 5& 3\\ 5& 5& 4& 4& 4& 6\\ 6& 5& 5& 2& 4& 5\\ 3& 5& 4& 6& 6& 4\\ 4& 6& 3& 5& 4& 5\\ 7& 7& 8& 6& 5& 7\\ 8& 5& 6& 7& 6& 7\\ 4& 5& 5& 5& 4& 6\\ 6& 5& 6& 6& 5& 5\\ 4& 5& 7& 6& 4& 7\\ 7& 6& 5& 3& 8& 5\\ 4& 5& 3& 5& 4& 4\\ 4& 3& 4& 5& 4& 3\\ 3& 4& 3& 4& 5& 5\\ 5& 5& 4& 5& 5& 3\\ 8& 8& 7& 8& 9& 3\\ 4& 5& 5& 4& 3& 7\\ 3& 4& 4& 3& 4& 5\\ 4& 5& 5& 4& 5& 4\\ 7& 7& 8& 3& 7& 3\\ 3& 4& 3& 7& 3& 8\end{array}\right)$$

(1)

Construction of the whitening weight functions

Based on the evaluation indicator scoring criteria and the foundation of grey theory, this paper uses the whitening weight function to construct grey class levels, with 4 grey classes set. After reviewing relevant literature [2], the thresholds of the whitening weight function are finally determined as 1, 3.5, 6.5, and 9. If the expert ratings fall within the range of 0–3.5, they belong to the first grey class; ratings between 3.5 and 6.5 are categorized as the second grey class; ratings from 6.5 to 9 belong to the third grey class; and finally, ratings between 9 and 10 are classified as the fourth grey class.

First grey class: ${\mathrm{f}}_{\mathrm{j}}^1(\mathrm{x})=\left\{\begin{array}{c}1;\mathrm{x}\in \left[\mathrm{0,1}\right)\\ {\displaystyle \frac{3.5-\mathrm{x}}{2.5}};\mathrm{x}\in [\mathrm{1,3.5}]\\ 0;\mathrm{x}\notin \left(\mathrm{0,3.5}\right)\end{array}\right.$,

Second grey class: ${\mathrm{f}}_{\mathrm{j}}^2(\mathrm{x})=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{x}-1}{2.5}};\mathrm{x}\in [\mathrm{1,3.5})\\ {\displaystyle \frac{6.5-\mathrm{x}}{3}};\mathrm{x}\in [\mathrm{3.5,6.5}]\\ 0;\mathrm{x}\notin \left(\mathrm{1,6.5}\right)\end{array}\right.$,

Third grey class: ${\mathrm{f}}_{\mathrm{j}}^3(\mathrm{x})=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{x}-3.5}{3}};\mathrm{x}\in [\mathrm{3.5,6.5})\\ {\displaystyle \frac{9-\mathrm{x}}{2.5}};\mathrm{x}\in [\mathrm{6.5,9}]\\ 0;\mathrm{x}\notin \left(\mathrm{3.5,9}\right)\end{array}\right.$,

Fourth grey class: ${\mathrm{f}}_{\mathrm{j}}^4(\mathrm{x})=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{x}-6.5}{2.5}};\mathrm{x}\in [\mathrm{6.5,9})\\ 1;\mathrm{x}\in \left[\mathrm{9,10}\right]\\ 0;\mathrm{x}\notin \left(\mathrm{6.5,10}\right)\end{array}\right.$

(2)

The grey clustering weight coefficients and weight vectors of the secondary indicators were calculated

Taking the secondary indicator safety S11 as an example, the grey clustering coefficients and weight vectors were calculated:

When e = 1, ${\mathrm{f}}_{\mathrm{S}11}^1\left(\mathrm{x}\right)={\mathrm{f}}_{\mathrm{S}11}^1\left(8\right)$ + ${\mathrm{f}}_{\mathrm{S}11}^1\left(7\right)$ + ${\mathrm{f}}_{\mathrm{S}11}^1\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^1\left(6\right)+{\mathrm{f}}_{\mathrm{S}11}^1\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^1\left(7\right)$ = 0

When e = 2, ${\mathrm{f}}_{\mathrm{S}11}^2\left(\mathrm{x}\right)={\mathrm{f}}_{\mathrm{S}11}^2\left(8\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(6\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(7\right)$ = 0.17

When e = 3, ${\mathrm{f}}_{\mathrm{S}11}^3\left(\mathrm{x}\right)={\mathrm{f}}_{\mathrm{S}11}^3\left(8\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(6\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(7\right)$ = 3.43

When e = 4, ${\mathrm{f}}_{\mathrm{S}11}^4\left(\mathrm{x}\right)={\mathrm{f}}_{\mathrm{S}11}^4\left(8\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(6\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(7\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(7\right)$ = 1.4

The grey evaluation coefficient for indicator S11 is:

${\mathrm{f}}_{\mathrm{S}11}^1\left(\mathrm{x}\right)+{\mathrm{f}}_{\mathrm{S}11}^2\left(\mathrm{x}\right)+{\mathrm{f}}_{\mathrm{S}11}^3\left(\mathrm{x}\right)+{\mathrm{f}}_{\mathrm{S}11}^4\left(\mathrm{x}\right)$ = 5, and the weight vector is [0,0.034,0.686,0.28]. Similarly, the grey evaluation coefficients and weight vectors for other secondary indicators were obtained, as provided below:

$$A_{\mathrm{S}1}=\left(\begin{array}{cccc}0& 0.034& 0.69& 0.28\\ 0& 0.61& 0.39& 0\\ 0.03& 0.66& 0.31& 0\\ 0& 0.61& 0.39& 0\end{array}\right)A_{\mathrm{S}2}=\left(\begin{array}{cccc}0.1& 0.48& 0.42& 0\\ 0.04& 0.53& 0.43& 0\\ 0.03& 0.61& 0.36& 0\\ 0& 0.11& 0.69& 0.2\end{array}\right)$$

$$A_{\mathrm{S}3}=\left(\begin{array}{cccc}0& 0.14& 0.7& 0.17\\ 0& 0.51& 0.49& 0\\ 0& 0.33& 0.67& 0\end{array}\right)A_{\mathrm{S}4}=\left(\begin{array}{cccc}0& 0.39& 0.56& 0.07\\ 0.03& 0.33& 0.51& 0.13\\ 0.03& 0.72& 0.25& 0\\ 0.07& 0.77& 0.17& 0\\ 0.07& 0.71& 0.22& 0\end{array}\right)$$

$$A_{\mathrm{S}5}=\left(\begin{array}{cccc}0.03& 0.61& 0.36& 0\\ 0.03& 0.13& 0.33& 0.5\\ 0.03& 0.58& 0.36& 0.03\\ 0.07& 0.77& 0.17& 0\end{array}\right)A_{\mathrm{S}6}=\left(\begin{array}{cccc}0& 0.67& 0.33& 0\\ 0.07& 0.27& 0.47& 0.2\\ 0.1& 0.54& 0.23& 0.13\end{array}\right)$$

(3)

Criterion-level clustering evaluation results

The evaluation results of the criterion layer are obtained by multiplying the weight values of each indicator by the grey clustering weight vector. Take the calculation of S1 clustering evaluation as an example:

Clustering evaluation for S1:

$$\begin{array}{c}{B}_{\mathrm{S}1}=W_{\mathrm{S}1}\ast A_{\mathrm{S}1}=\left(\begin{array}{cccc}0.35& 0.19& 0.24& 0.22\end{array}\right)\times \left(\begin{array}{cccc}0& 0.034& 0.69& 0.28\\ 0& 0.61& 0.39& 0\\ 0.03& 0.66& 0.31& 0\\ 0& 0.61& 0.39& 0\end{array}\right)\\ =(0.007 0.420 0.476 0.098)\end{array}$$

Similarly,

Clustering evaluation for S2: B_S2 = $\left(\begin{array}{cccc}0.035& 0.30& 0.56& 0.11\end{array}\right)$;

Clustering evaluation for S3: B_S3 = $\left(\begin{array}{cccc}0& 0.834& 0.66& 0.094\end{array}\right)$;

Clustering evaluation for S4: B_S4 = $\left(\begin{array}{cccc}0.031& 0.505& 0.408& 0.063\end{array}\right)$;

Clustering evaluation for S5: B_S5 = $\left(\begin{array}{cccc}0.036& 0.401& 0.318& 0.242\end{array}\right)$;

Clustering evaluation for S6: B_S6 = $\left(\begin{array}{cccc}0.061& 0.441& 0.372& 0.132\end{array}\right)$.

(4)

Target-level clustering evaluation results

$$\begin{array}{c}{B}_{\mathrm{S}}=\left(\begin{array}{cccccc}0.151& 0.153& 0.079& 0.084& 0.392& 0.141\end{array}\right)\ast \left(\begin{array}{cccc}0.007& 0.420& 0.476& 0.098\\ 0.039& 0.380& 0.507& 0.072\\ 0& 0.314& 0.630& 0.060\\ 0.035& 0.550& 0.371& 0.048\\ 0.038& 0.50& 0.311& 0.151\\ 0.063& 0.477& 0.340& 0.120\end{array}\right)\\ =(0.034 0.456 0.399 0.111)\end{array}$$

Based on the principle of maximum membership degree, and using the formula K = B_S ∗ $\left(\begin{array}{cccc}1& 3.5& 6.5& 9\end{array}\right)$^T to calculate the comprehensive evaluation value of Project A, the quantified comprehensive evaluation result can be obtained in the following:

$$K=\left(\begin{array}{cccc}0.034& 0.456& 0.399& 0.111\end{array}\right)\ast {(1 3.5 6.5 9)}^{\mathrm{T}} = 5.223$$

2.3. Discussion

2.3.1. Case Discussion

This study takes Project A as the verification object. As a typical non-domestic green building, its specific overview and technical applications provide practical support for the effectiveness of the evaluation system in this study. With a total construction area of approximately 270,000 square meters and a total investment of 6 billion yuan, Project A integrates composite functions such as commercial offices (5A intelligent office buildings), five-star hotels, and serviced apartments, serving as a benchmark project for the “vertical urban ecosystem” in the region. During construction and operation, the project has integrated multiple green and smart technologies: in the construction phase, prefabricated connection technology (reducing welding pollution and lowering labor costs by 20%) and the “airborne building construction machine” super high-rise construction platform (improving efficiency by 30% and reducing labor consumption by 40%) were adopted; in the operation phase, it is equipped with a 5A intelligent system (building automation, smart security, shared office) and green technologies (photovoltaic curtain walls, rainwater recycling system, vertical greening), forming a functional closed loop of “work–life–leisure”. These characteristics make it an ideal case for verifying the “greenization + smartization” dual-drive evaluation system.

The comprehensive score of Project A is 5.223 (Grade B), with a significant gap from Grade A (8–10 points). Through the disassembly and analysis of the grey clustering results of secondary indicators, it is found that the core shortcomings are mainly concentrated in the following types of indicators, whose poor performance directly restricts the improvement of the overall score:

(1)

As the primary indicator with the highest weight (39.14%), Resource Conservation S5 has a significant impact on the overall score through the performance of its subordinate secondary indicators. The clustering results show:

Water Conservation and Water Resource Utilization S54: With a combined weight of 19.53%, its grey clustering weight vector is (0.07, 0.77, 0.17, 0), meaning 77% belongs to the second grey class (Grade C, 2–5 points) and only 17% belongs to the third grey class (Grade B). Specific shortcomings are reflected in “the utilization rate of the rainwater recycling system is less than 30%” and “intelligent water resource monitoring only covers public areas, not extending to hotel rooms or office units,” which is mismatched with Wuhan’s climate characteristics of “hot summer and cold winter with concentrated rainy seasons,” resulting in low efficiency of water resource recycling.

Material Conservation and Green Building Materials S53: With a combined weight of 35.96%, its clustering weight vector is (0.03, 0.58, 0.36, 0.03), with over half belonging to Grade C. The main issues are “the coverage rate of integrated civil construction and decoration is only 60%” and “the recycling rate of construction waste is less than 25%,” failing to meet the high-level requirements for green building material application.

(2)

Functional defects in the service and convenience dimension. Although Service and Convenience S4 has an overall low weight (8.42%), the poor performance of its subordinate indicators reflects insufficient implementation of smart services:

Intelligent Services S45: With a combined weight of 13.24%, its clustering weight vector is (0.07, 0.71, 0.22, 0), with 71% belonging to Grade C. Specific manifestations include “incomplete coverage of the indoor positioning and navigation system (only 50% of public areas completed)” and “the equipment operation and maintenance management system has not implemented AI fault early warning, still relying on manual inspections,” which is inconsistent with its positioning as a “5A intelligent office building.”

Information Services S43: With a combined weight of 17.60%, its clustering weight vector is (0.03, 0.72, 0.25, 0), with 72% belonging to Grade C. Shortcomings lie in “data infrastructure has not achieved inter-tower linkage” and “user activity on the smart service platform is less than 40%,” meaning the practicality of information services has not been fully exerted.

(3)

Partial deficiencies in the environmental livability dimension. The non-smart indicators in Environmental Livability S6 (weight 14.08%) show weak performance. Outdoor Physical Environment S61: With a combined weight of 25.58%, its clustering weight vector is (0, 0.67, 0.33, 0), with 67% belonging to Grade C. The main problems are “heat island intensity control has not met the design target (actual measurement is 1.2 °C higher than the surrounding area)” and “energy consumption of the vertical greening irrigation system is relatively high,” resulting in the incomplete release of the actual effectiveness of ecological design.

From the perspective of the resource conservation dimension, the resource conservation weight of Project A is as high as 39.14%, which is directly related to the technologies actually adopted by the project, such as photovoltaic curtain walls (renewable energy utilization), rainwater recycling systems (water resource recycling), and integrated civil engineering and decoration (material conservation). Kamal A [16] et al.’s research on super high-rise buildings in high-energy-consumption climate zones shows that the cooling load of buildings in hot-summer and cold-winter regions accounts for 62% of total energy consumption, requiring enhanced weighting of energy-saving indicators. Project A’s air-conditioning system and energy-saving control measures optimized for Wuhan’s climate characteristics exactly confirm the core position of resource conservation indicators in regional adaptability evaluation. Yadegaridehkordi E [18] pointed out in his research that compared with the generally low resource conservation weight (below 30%) in green building studies in Southeast Asia, the high weight result of this project is more in line with the actual pressure of an average annual increase of 3.7% in carbon emissions during the operation phase of Chinese buildings, highlighting the local adaptability of the evaluation system.

2.3.2. Model Discussion

This study refers to LEED’s four-level certification system and divides the grey clustering model into 4 grey classes, achieving more precise interval division. The four-classification realizes compatibility with international standards, facilitating comparisons of cross-border projects. If divided into 3 classes, the large span would lead to information loss, making it difficult to reflect scoring differences; if divided into 5 or more classes, it would increase model complexity and reduce the interpretability of results.

The methodological value of this case study is reflected in the following theoretical aspects: First, through the AHP-entropy weight coupling mechanism, it not only retains experts’ subjective judgments on indicators based on their experience but also uses the entropy weight method to ensure the objectivity of weights, controlling the comprehensive weight deviation within 15%. This is superior to the single AHP method compared by Zhang G et al. [39] (with a deviation of 25%), and their idea of optimizing energy consumption through the PSO-SVM algorithm provides methodological support for weight calculation in this study. Second, the four-level division of grey clustering in this study is consistent with the grey class interval model proposed by Wenbin H et al. [40]. Their study screened thermal disturbance parameters through grey relational analysis and established an interval grey class model to describe building thermal processes, proving that a four-level division can balance accuracy and operability. Compared with the 3-class division, this study reduces information loss of the “indoor thermal comfort” indicator through the whitening weight function (e.g., the clustering deviation of indicator S24 in Wuhan’s summer decreased from 12% to 5%).

In terms of regional adaptability limitations, the application of Project A shows that “smart energy” (weight 37.93%) and “energy conservation and energy utilization” (weight 27.99%) are core driving indicators. However, Wuhan belongs to a north subtropical monsoon humid climate with hot summers and cold winters, so the indicator system may be inapplicable in different regions or regulatory environments, requiring targeted adjustments to indicator weights. Sui X et al. [38] studied the TABS system in Xi’an and found that 8-h intermittent cooling in humid and hot climates saves 31% more energy than 24-h continuous operation, while Wuhan needs to strengthen nighttime cooling in summer. This confirms the insufficient regional adaptability of this study—the smart energy indicator (37.93%) in the Wuhan project does not consider differences in usage time and electricity prices across different climate zones.

2.3.3. Sensitivity Analysis

(1)

Variation of core indicator weights with α

To verify the rationality of the preference coefficient α between subjective weights (AHP) and objective weights (entropy weight method) in the combined weights, this study selects typical values within the interval α ∈ [0, 1] (0, 0.2, 0.4, 0.5, 0.6, 0.8, 1.0) for sensitivity analysis. The combined weights under different α values are calculated using Formula 9, and then the comprehensive score of Project A is recalculated based on the grey clustering model to analyze the influence law of α values on the weights of core indicators and evaluation results. The top 3 core indicators with the highest weight proportion in the evaluation system (Resource Conservation S5, Smart Energy S63, Smart Design S33) are selected, and the results of the variation in their combined weights with α are shown in the

Table 13. Variation of Core Indicator Weights.

α Value	Combined Weight of Resource Conservation S5	Combined Weight of Smart Energy S63	Combined Weight of Smart Design S33
0.0	41.50%	49.38%	44.93%
0.2	40.55%	45.30%	43.93%
0.4	39.60%	41.22%	42.93%
0.5	39.14%	37.93%	37.17%
0.6	38.67%	34.64%	31.41%
0.8	37.74%	28.06%	20.00%
1.0	36.78%	26.47%	29.40%

below:

As can be seen from the table above, as α increases (with the proportion of subjective weights increasing), the weight of Resource Conservation S5 shows a slow downward trend (from 41.50% to 36.78%), while the weights of Smart Energy S63 and Smart Design S33 show significant downward trends (from 49.38% to 26.47% and from 44.93% to 29.40% respectively). This indicates that the objective weights (entropy weight method) assign higher priority to smart indicators, whereas the subjective weights (AHP) place greater emphasis on the importance of the traditional resource conservation dimension.

(2)

Variation of Project A’s comprehensive score with α

Based on the combined weights under different α values, the grey clustering comprehensive score of Project A is recalculated, and the results are shown in the

Table 14. Variation of Project A’s Comprehensive Score.

α Value	Comprehensive Score	Evaluation Grade	Score Deviation from the Reference Value (α = 0.5)
0.0	5.682	B	+0.459
0.2	5.517	B	+0.294
0.4	5.356	B	+0.133
0.5	5.223	B	0
0.6	5.193	B	−0.03
0.8	4.985	B	−0.238
1.0	4.763	B	−0.46

below:

The comprehensive score of Project A shows a linear downward trend as α increases (with the proportion of subjective weights rising), decreasing from 5.682 under pure objective weights (α = 0) to 4.763 under pure subjective weights (α = 1). However, the evaluation grade remains stable at Level B. The score deviation is smallest within the interval α = 0.4–0.6, indicating that results within this interval have low sensitivity to the value of α, and α = 0.5 lies at the midpoint of the stable interval.

(3)

Results of Sensitivity Analysis

Impact of α value on weights: Objective weights (α = 0) highlight the importance of smart indicators (smart energy, smart design) more significantly, while subjective weights (α = 1) focus more on the traditional resource conservation dimension. In contrast, α = 0.5 achieves balanced consideration of both types of indicators, which is consistent with the dual-drive evaluation orientation of “greenization + smartization.”

Verification of result robustness: Although changes in α values lead to score fluctuations (with a maximum deviation of 0.516), the evaluation grade remains unchanged. This indicates that the system has certain fault tolerance for differences in subjective-objective preferences, and the selection of α = 0.5 will not affect the reliability of evaluation conclusions.

Methodological rationality: The interval 0.4–0.6 where α = 0.5 is located is a stable result interval, with minimal score deviation within this range. This further verifies the rationality of deriving α = 0.5 through the method of sum of least squares deviations, while avoiding the extremization flaws of single subjective or objective weighting.

3. Conclusions

This study constructed a green building evaluation system integrating the dual-driven dimensions of “intelligence + greenness”. Through combined weighting using the AHP-entropy weight method and the grey clustering model, a dynamic assessment of building sustainability and intelligent performance was realized, as shown in in Figure A1 of Appendix A.

The main conclusions are as follows:

(1)

Innovation in the Evaluation System: Based on drawing on smart-related assessment concepts from international frameworks such as LEED and BREEAM, and in response to the contextual needs of China’s GB/T standards, four secondary indicators, including “smart security” and “smart energy,” have been systematically integrated. This has increased the proportion of the smart dimension to 35%, filling the gap in domestic standards regarding the quantitative assessment of Internet of Things (IoT) and BIM technologies. Dynamic indicators such as “energy consumption monitoring system” and “regular operation evaluation” have been introduced, breaking through the limitations of traditional static evaluation and forming a four-in-one framework of “resource-environment-smart-security.” This design is fully consistent with the requirements of “real-time energy consumption monitoring” and “regular operation evaluation” specified in the Zero-Carbon Building Assessment Standard (Trial), and the proportion of the smart dimension is higher than that in the current national standards (less than 10%).

(2)

Methodological optimization: Combined weighting using the AHP-entropy weight method (α = 0.5) balances subjective and objective deviations, improving evaluation accuracy by 12% compared with the single AHP method. The grey clustering model processes fuzzy information through a four-level gray division and whitening weight function, making the quantitative result of Project A’s comprehensive score (5.223, Grade B) more consistent with its actual performance. This method is recommended in Chongqing’s Low-Carbon Building Evaluation Standard for balancing subjective and objective weights, echoing the trend of dynamic assessment in international standards (e.g., LEED v5).

(3)

Value of case verification: Results of Project A show that resource conservation (39.14%) and smart energy (37.93%) are core indicators, confirming the key role of energy conservation and intelligent technologies under the “dual-carbon” goals.

The current research has three limitations: First, the issues of parameter sensitivity of the whitening weight function and the lack of an optimization mechanism are prominent. As the core tool for grey clustering models to handle fuzzy information, the setting of turning point parameters of the whitening weight function (such as thresholds for grey class classification: 1, 3.5, 6.5, 9) directly affects the calculation results of indicator membership degrees. In this study, parameters are mainly determined based on expert experience and industry consensus, lacking data-driven optimization verification. On the one hand, different types of indicators show significant differences in parameter sensitivity: the membership calculation of quantitative indicators is less affected by the deviation of turning points, while qualitative indicators, due to higher ambiguity in scoring, may have a high membership deviation rate when parameters are fine-tuned. For example, when the score of the “smart design” indicator is 7, if the turning point is adjusted from 6.5 to 7.0, its weight belonging to the third grey class will drop from 0.686 to 0.42, directly affecting the judgment of the evaluation grade. On the other hand, parameter settings do not consider differences in project types. Commercial office buildings and residential buildings have different threshold requirements for the “smart security” indicator, but this study uses unified parameters, which may lead to a disconnection between evaluation results and actual functional needs.

Second, the regional and type limitations of case samples are significant, restricting the promotion adaptability of the evaluation system. This study only uses the commercial office complex in Hankou International Riverside Business District of WH City as a verification case, with dual limitations. From the regional perspective, Wuhan has a north subtropical monsoon humid climate, and its climate characteristic of being hot in summer and cold in winter makes the weight settings of indicators such as “smart energy” (weight: 37.93%) and “energy conservation and utilization” (weight: 27.99%) in the evaluation system biased towards monitoring summer cooling efficiency and winter insulation performance. However, this weight distribution is difficult to adapt to other climate zones. In severe cold regions, building heating energy consumption accounts for over 60%, requiring the strengthening of sub-indicators, such as “intelligent heating system regulation”, which are not included in this system. In arid regions, the priority of water resource recycling is higher, and the weight of the “water conservation and water resource utilization” indicator (19.53%) needs to be further increased; otherwise, the ecological value of the project may be underestimated. From the perspective of building types, the case is a non-civilian building integrating offices and hotels, but its evaluation system does not fully consider the particularities of different building functions. Residential buildings pay more attention to refined indicators of “health and comfort satisfaction” (current weight: 35.68%), such as real-time monitoring of indoor formaldehyde. Industrial buildings need to strengthen characteristic indicators like “reuse of green building materials” and “industrial solid waste treatment”. However, the indicator settings and weight allocation in this system do not reflect such functional differences, which may reduce the evaluation accuracy for residential and industrial buildings.

Third, the standardization of long-term dynamic monitoring indicators in the operation stage and the ability of data integration are insufficient. Although this study introduces dynamic indicators such as “real-time energy consumption monitoring” and “regular operation evaluation”, it has not established a sound standardization mechanism and data integration framework. First, the indicator monitoring standards are not unified. In this case, the “energy consumption monitoring frequency” is set as once a day, but green building standards vary across regions, resulting in a lack of comparability in the evaluation of cross-regional projects. The “regular operation evaluation cycle” is set as once a quarter, but the threshold standards for evaluation indicators are not clarified, making it difficult to quantify project performance fluctuations. Second, data sources are scattered and integration is low. In this case, energy consumption data comes from the building automation system, and smart security data comes from an independent monitoring platform; real-time interconnection between the two types of data has not been achieved, leading to a lag in the “safety-energy” collaborative evaluation. At the same time, there is a lack of in-depth integration with Building Information Modeling (BIM), making it impossible to achieve full lifecycle data traceability through digital twin technology, which weakens the timeliness of dynamic evaluation.

Future research can advance in three aspects: (1) Optimize parameters of the whitening weight function by integrating BP neural networks to reduce deviations from subjective settings; (2) Expand samples to cold, arid and other climate zones, and establish a mechanism for regionally differentiated weight adjustment (e.g., referring to technical requirements for different climate zones in Shanghai’s Green Building Regulations); (3) Introduce Building Information Modeling (BIM) and digital twin technologies to realize real-time update of dynamic evaluation data throughout the whole life cycle, enhancing the system’s ability to track long-term operational performance.