The Diffusion Mechanism of Blockchain Technology for Power Sector Carbon Emission Data Supervision from the Perspective of Sustainable Development

Abstract

Accurate power-sector carbon emission data (PS-CED) are critical for ensuring sustainable practices in carbon trading and effective emission reductions. However, conventional centralized reporting systems are susceptible to data tampering, duplicate accounting, and inefficient manual verification, hindering the achievement of sustainability goals. Blockchain technology (BCT) provides transparency, immutability, and automated compliance, offering significant potential for improving the sustainability of PS-CED supervision. Despite this, its diffusion in the sector faces challenges such as data heterogeneity, security concerns, institutional differences, and resource limitations. This study integrates the Technology Acceptance Model (TAM) and the Theory of Planned Behavior (TPB) to develop a diffusion framework for BCT adoption in PS-CED supervision with a focus on sustainability. Using partial least squares structural equation modeling (PLS-SEM) and fuzzy-set qualitative comparative analysis (fsQCA), the study examines both linear effects and multiple adoption configurations. The results indicate that adoption willingness mediates the effects of perceived usefulness and ease of use, while perceived regulatory norms underscore policy pressure as a crucial external driver for fostering sustainability. Configurational analysis reveals heterogeneous diffusion patterns, with high adoption performance driven by technological capability combined with regulatory enforcement, and low performance linked to weak technological engagement and structural constraints. Based on these findings, a strategic framework is proposed to support differentiated and phased BCT adoption across organizational contexts to enhance sustainability in carbon emission supervision. This paper clarifies the diffusion mechanisms and provides practical guidance for scaling blockchain-based PS-CED supervision to promote sustainability.

1. Introduction

As the major contributor to global carbon emissions, the accuracy and supervisory efficiency of power-sector carbon emission data (PS-CED) are critical not only to the fairness of carbon trading markets but also to the achievement of long-term sustainable development goals [1,2]. According to the International Energy Agency (IEA), power sector emissions accounted for approximately 42% of global emissions by 2024 [3], making the sector a focal point for global climate governance and sustainable energy transition. The increasingly stringent greenhouse gas reduction targets and carbon neutrality commitments place higher demands on sustainable data governance, particularly in terms of transparency, credibility, and real-time monitoring. Accurate, credible, and traceable PS-CED constitute the institutional and technical foundation for sustainable carbon allowance allocation, carbon price formation, and the effective functioning of carbon markets, thereby enhancing market liquidity, regulatory effectiveness, and public confidence [4]. However, conventional centralized carbon emissions reporting systems typically rely on manual accounting and multi-level aggregation, which undermines the sustainability of regulatory systems by increasing the risks of data tampering, duplicate accounting, and prolonged verification cycles [5,6,7]. Blockchain technology (BCT), characterized by distributed ledgers, tamper-resistant records, and smart contract–enabled automation, offers a promising technological pathway to support sustainable PS-CED supervision [8,9]. By facilitating multi-source data integration, traceability, and automated compliance mechanisms, BCT can significantly enhance the transparency, reliability, and long-term sustainability of PS-CED supervision frameworks [10].

Although BCT holds considerable promise for enabling high-quality supervision of PS-CED, multiple barriers remain to its large-scale deployment. Existing studies are largely fragmented, typically focusing on isolated factors such as technological maturity, organizational readiness, or policy incentives, while systematic investigations into multi-factor diffusion mechanisms and their interactive or configurational effects are still lacking. In practice, PS-CED supervision is characterized by high data heterogeneity, frequent real-time reporting requirements, and cross-organizational coordination, which jointly impose substantial demands on data integration, security protection, and regulatory compliance [11]. Moreover, emissions supervision involves multiple stakeholders, including power generation enterprises, dispatch authorities, verification agencies, and trading platforms, intensifying concerns over information security, privacy protection, and regulatory accountability [12,13]. Limited awareness of BCT’s operational complexity, uncertainty regarding its practical effectiveness, constraints in human and financial resources, and divergences in institutional norms and technical standards further weaken incentives for large-scale adoption [14,15,16]. As a result, promoting the high-quality diffusion and scalable application of BCT for PS-CED governance has emerged as a critical challenge for policymakers, industry practitioners, and researchers [17].

To address this gap, this paper develops a theoretical model of BCT diffusion for PS-CED supervision by integrating the Technology Acceptance Model (TAM) and the Theory of Planned Behavior (TPB). This integrated framework systematically links technological cognition, behavioral intention, and regulatory influence, thereby extending the literature on carbon emissions data governance in the power sector. Methodologically, this paper advances existing research by jointly applying partial least squares structural equation modeling (PLS-SEM) and fuzzy set qualitative comparative analysis (fsQCA). In contrast to prior studies that predominantly rely on linear modeling approaches, the combined method enables the identification of both net effects and configurational pathways. From the practical perspective, this paper proposes a differentiated strategic framework for BCT diffusion. By clarifying how specific configurations of technological, organizational, and regulatory factors shape diffusion outcomes, the findings provide actionable insights for policymakers, regulators, and power sector organizations seeking to promote scalable and context-specific deployment of blockchain-based supervisory systems.

The structure of this paper is organized as follows. Section 1 introduces the background and motivation. Section 2 reviews the relevant literature and develops the hypotheses. Section 3 presents the research design and data analysis methods. Section 4 reports the empirical results of the PLS-SEM and fsQCA. Section 5 discusses the diffusion mechanisms and strategies for BCT. Section 6 summarizes the conclusions and limitations.

2. Theoretical Foundation and Hypotheses Development

2.1. Theoretical Foundation

The diffusion of BCT for PS-CED supervision is essentially a sustainable technology adoption and governance issue, situated at the intersection of digital innovation, environmental regulation, and institutional coordination. Existing studies on sustainable information systems and environmental data governance emphasize that the adoption of emerging digital technologies is not only driven by technical performance, but also shaped by organizational capabilities, regulatory pressure, and long-term sustainability objectives, such as transparency, and resilience of governance systems [18,19,20]. Within this stream of research, the TAM has been widely applied to explain the adoption of digital and environmental technologies, including energy management systems, smart grids, and blockchain-enabled sustainability applications. Prior studies demonstrate that Perceived Usefulness (PU) and Perceived Ease of Use (PEU) significantly influence users’ willingness to adopt technologies that enhance environmental performance, data transparency, and operational efficiency [21,22,23]. In the context of carbon emissions management, PU is often associated with improvements in data accuracy, traceability, and verification efficiency, while PEU reflects the degree to which complex technologies can be integrated into existing regulatory and operational processes without excessive learning costs. However, existing TAM-based studies have been criticized for focusing primarily on individual cognitive evaluations, while paying limited attention to institutional constraints, regulatory compliance, and collective action problems that are central to sustainable environmental governance [24,25].

To address these limitations, the TPB provides a complementary perspective by explicitly incorporating social and institutional factors into technology adoption decisions. TPB posits that Attitude (AT), Subjective Norm (SN), and Perceived Behavioral Control (PBC) jointly determine behavioral intention [26]. In sustainability-oriented contexts, TPB has been extensively employed to explain environmentally responsible behaviors, low-carbon technology adoption, and organizational compliance with environmental regulations. In particular, SN captures the influence of policy mandates, regulatory expectations, and industry norms, while PBC reflects the availability of organizational resources, technical expertise, and implementation capacity [27,28,29].

In the specific context of PS-CED supervision, BCT diffusion faces multiple sustainability-related challenges, including high data heterogeneity, cross-organizational coordination costs, information security concerns, and stringent regulatory accountability requirements. These challenges cannot be fully explained by technological perceptions alone. Instead, they require a framework that integrates technological cognition with institutional pressure and execution capability. By combining TAM and TPB, this study captures both the efficiency-oriented logic of digital innovation and the governance-oriented logic of sustainable regulation, thereby providing a more comprehensive theoretical foundation for analyzing BCT diffusion in PS-CED supervision [30,31,32]. Accordingly, the integrated TAM-TPB framework adopted in this study enables a systematic examination of how technological characteristics, organizational readiness, and regulatory environments jointly shape the sustainable diffusion of blockchain-based supervisory systems in the power sector.

2.2. Hypotheses Development

Perceived Operational Ease (POE) is derived from the PEU dimension in the TAM, which posits that technologies requiring less effort to learn and operate are more likely to be accepted and adopted. In the context of power-sector carbon emission data supervision, POE reflects the extent to which BCT reduces operational complexity associated with data entry, format conversion, verification procedures, and cross-departmental coordination [20,25]. When blockchain systems are perceived as compatible with existing supervisory workflows and easy to operate, organizational resistance to technological change is reduced, thereby strengthening behavioral intention toward adoption. At the same time, lower operational barriers facilitate effective system use, enabling organizations to translate technical feasibility into improved supervisory performance. Accordingly, POE is expected to positively influence both adoption willingness and diffusion outcomes.

H1: Perceived Operational Ease positively affects the Adoption Willingness of Blockchain Technology in the power-sector carbon emission data supervision.

H2: Perceived Operational Ease positively affects the Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

Perceived Enabling Utility (PEU) corresponds to PU in the TAM and represents the extent to which a technology enhances task performance and goal achievement. In PS-CED supervision, PEU captures the functional value of BCT in ensuring data immutability, enabling traceable auditing, and supporting automated compliance through smart contracts [26,27,28]. These capabilities directly address the limitations of traditional centralized reporting systems and enhance regulatory efficiency and reliability. According to the TAM, when organizations perceive clear performance gains from a technology, they are more likely to develop strong adoption intentions and commit resources to its implementation. Therefore, PEU is expected to positively influence adoption willingness and, through effective application, diffusion performance.

H3: Perceived Enabling Utility positively affects the Adoption Willingness of Blockchain Technology in the power-sector carbon emission data supervision.

H4: Perceived Enabling Utility positively affects the Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

Adoption Willingness (AW) functions as a central mediating mechanism in the diffusion of BCT at the organizational level. According to the TPB, behavioral intention represents the most immediate antecedent of actual behavior, particularly in contexts requiring sustained effort and coordinated action. In PS-CED supervision, the adoption of BCT entails not only technical deployment but also organizational commitment, resource allocation, process restructuring, and cross-departmental coordination [29,30,31]. Therefore, favorable perceptions of operational ease or enabling utility alone are insufficient to guarantee effective diffusion. AW reflects the collective readiness of organizations to translate technological cognition into concrete implementation and routine use. Consequently, AW is expected to exert a direct influence on diffusion outcomes and to mediate the effects of POE and PEU on BCT diffusion performance.

H5: Adoption Willingness positively affects the Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

H6: Adoption Willingness mediates the relationship between Perceived Operational Ease and Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

H7: Adoption Willingness mediates mediate the relationship between Perceived Enabling Utility and Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

Given the stringent policy compliance requirements in PS-CED supervision, Regulatory Norm Perception (RNP) represents a form of subjective norm as conceptualized in the TPB. A high level of regulatory norm perception signals policy legitimacy and compliance necessity, thereby directly motivating organizations to adopt BCT as a solution aligned with regulatory requirements [32]. Subjective Compliance Cognition (SCC), which is closely related to perceived behavioral control in the TPB, reflects the extent to which organizations perceive their capability to effectively implement and operate BCT systems. High levels of SCC indicate the presence of adequate resources, sufficient technical competence, and clear procedural arrangements. These conditions strengthen organizational confidence in executing adoption decisions and contribute to the achievement of diffusion outcomes [33]. Accordingly, both RNP and SCC are expected to exert direct effects on BCT diffusion performance.

H8: Regulatory Norm Perception positively affects the Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

H9: Subjective Compliance Cognition positively affects the Blockchain Technology diffusion outcome in the power-sector carbon emission data supervision.

By integrating TAM and the TPB, this paper develops a theoretical hypothesis model comprising a perception layer (POE, PEU), a performance layer (BDO, AW), and a regulatory layer (RNP, SCC). The model aims to systematically elucidate the core mechanisms underlying BCT diffusion in PS-CED supervision. As shown in Figure 1.

3. Methodology

3.1. Research Design

The technical route of this study consists of three main steps, as illustrated in Figure 2. First, drawing on the TAM and the TPB as the theoretical foundations, this study formulates hypothesized diffusion pathways of BCT in PS-CED supervision. To ensure that the research framework is grounded in actual engineering practice and regulatory contexts, fieldwork was conducted between August and September 2024 through group-based surveys and on-site visits to three representative power infrastructure projects in Liaoning Province: the Chaoyang Baoguo 220 kV transmission line project, the Murong 220 kV substation project, and the Songtao 220 kV transmission line project. The selection of three projects is methodologically appropriate for the exploratory purpose of this study. These projects are all large-scale 220 kV power transmission and substation projects, which are key components of regional power supply systems. Such projects involve complex construction processes, multiple participating stakeholders, and intensive data generation related to construction progress, safety management, and carbon emission monitoring. As a result, they constitute typical application scenarios for PS-CED supervision and provide a realistic context for examining the adoption and diffusion of BCT in supervisory practices. Access to these projects was obtained through institutional cooperation with power sector organizations, which enabled in-depth field investigation and reliable data collection. In October 2024, the research team held weekly expert meetings with scholars from Liaoning Technological University, Shenyang Jianzhu University, and the State Grid Liaoning Economic and Technical Research Institute to assess the scientific validity, clarity, and measurability of the questionnaire items.

Based on insights obtained from the field investigations and a systematic review of the relevant literature, an indicator system for BCT diffusion in PS-CED supervision is developed. As shown in Appendix A, the indicator system comprises six latent dimensions: PEU, POE, RNP, SCC, AW, and BDO. The selection of these latent variables was guided by both theoretical relevance and empirical applicability. Specifically, PEU and POE originate from TAM and reflect users’ cognitive evaluation of BCT. RNP captures external institutional and regulatory influences emphasized in TPB. SCC reflects the compatibility and collaborative requirements of BCT implementation in complex supervisory systems. AW represents stakeholders’ cognitive awareness of BCT and PS-CED supervision. BDO captures the actual diffusion outcomes and behavioral responses to BCT adoption. Together, these dimensions comprehensively represent individual, organizational, and institutional factors influencing BCT diffusion in PS-CED supervision. The observed variables corresponding to each latent variable were selected based on three criteria: (1) consistency with established measurement scales in prior studies on technology adoption and governance innovation; (2) relevance to the operational characteristics of PS-CED supervision identified during fieldwork; (3) clarity and measurability to ensure respondents’ understanding. The numbers 1–13 shown in Appendix A refer to the key literature sources on which the corresponding indicators are theoretically grounded. These references provide the conceptual and empirical basis for the construction of the measurement items used in this study. Second, PLS-SEM is employed to examine the driving mechanisms of BCT diffusion in PS-CED supervision. Finally, informed by the PLS-SEM results, fsQCA is applied to explore how different configurations of influencing factors jointly shape multiple diffusion pathways in PS-CED governance.

3.2. Data Collection and Analysis

Data are collected using a structured and closed-ended questionnaire. Informed consent was obtained from all participants prior to their participation in the survey. Participation was voluntary, and all responses were collected anonymously. The measurement items are developed based on the indicator system described in Section 3.1 and Appendix A, which is constructed through a systematic review of the relevant literature and insights obtained from field investigations, rather than through the direct adoption of a single standardized scale. During the pretest stage, a practical sample size ranging from 5 to 30 respondents is generally considered acceptable for assessing the clarity and rationality of questionnaire items [34]. Accordingly, this study invites 2 professors from Liaoning Technological University of Engineering and Technology, 2 professors from Shenyang Jianzhu University, 2 experts from the State Grid Liaoning Economic and Technical Research Institute, and 10 doctoral students to participate in the pretest. The invited experts and doctoral students possess relevant academic or professional backgrounds related to this study. The participating professors and experts have long-term research or practical experience in areas such as digital construction management, power system engineering, energy governance, and information technology applications, including blockchain-related or data governance topics. The doctoral students are engaged in research fields associated with infrastructure management, energy systems, or digital technologies and have received formal training in research methods and emerging information technologies. Their combined expertise enabled them to assess the conceptual consistency, contextual relevance, and comprehensibility of the questionnaire items in relation to BCT and PS-CED supervision.

Expert feedback is consolidated to further validate the rationality and clarity of the questionnaire items. The questionnaire consists of three parts, as detailed in the Appendix B. The first part collects respondents’ demographic and professional information, including educational background, gender, and years of service. The second part measures respondents’ attitudes toward 19 indicators across 6 dimensions using a five-point Likert scale anchored from very unimportant (1) to very important (5). The third part solicits open-ended comments. The relevant collected questionnaire data are presented in the Supplementary Materials. Between November 2024 and January 2025, the research team distributed 300 questionnaires to personnel associated with the field projects and received 273 responses, of which 251 met the criteria for inclusion in the final analysis, as shown in

Table 1. Basic information of respondents.

Variable	Categorization	Number	Percentage	Variable	Categorization	Number	Percentage
Gender	Male	136	54.2%	Position	Technical staff	93	37%
					General Manager	84	33.5%
	Female	115	45.8%		Senior Manager	74	29.5%
Work Experience	1–4 years	84	33.5%	Educational Background	Bachelor	97	38.6%
	5–7 years	96	38.2%		Postgraduate	69	27.5%
	≥8 years	71	28.3%		Other	85	33.9%

.

3.3. PLS-SEM Method

PLS-SEM integrates the analysis of both measurement and structural models and is particularly suitable for theory testing and prediction-oriented research involving complex causal relationships. In this study, PLS-SEM is employed to examine the net effects and mediating mechanisms among latent variables influencing BCT diffusion in PS-CED supervision. Scale reliability and convergent validity are assessed using composite reliability and average variance extracted (AVE), while discriminant validity is evaluated through the Fornell–Larcker criterion [35]. Path coefficient estimation enables the identification of direct and indirect causal relationships among latent constructs, thereby revealing the average effects of individual factors and their mediating pathways [36]. In addition, the coefficient of determination (R²) is used to evaluate the explanatory power of the structural model with respect to BCT diffusion outcomes. Predictive relevance (Q²), together with root mean square error (RMSE) and mean absolute error (MAE), is applied to assess the model’s predictive accuracy and generalizability to new samples [37]. SmartPLS 4.0 is used to conduct the PLS-SEM analysis. Overall, PLS-SEM provides a rigorous quantitative assessment of the dominant linear relationships and mediating mechanisms underlying BCT diffusion in PS-CED supervision.

3.4. fsQCA Method

While PLS-SEM focuses on estimating average net effects based on additive and symmetric assumptions, it may overlook the configurational and non-linear nature of causal relationships in complex socio-technical systems. To address this limitation, fsQCA is employed as a complementary method. Grounded in set theory and Boolean algebra, fsQCA calibrates variables into fuzzy membership scores and identifies combinations of causal conditions that are sufficient or necessary for achieving high or low levels of BCT diffusion [38,39]. The primary rationale for integrating fsQCA with PLS-SEM lies in its ability to capture causal complexity, including equifinality, asymmetry, and substitution effects among conditions. fsQCA enables the identification of multiple alternative diffusion pathways, in which different combinations of technological, organizational, and institutional factors can lead to similar outcomes. This configuration-oriented perspective allows for the exploration of how the absence of a specific factor identified as significant in PLS-SEM may be compensated by other conditions, thereby uncovering heterogeneous diffusion mechanisms that cannot be fully explained by net-effect models alone [40].

By combining PLS-SEM and fsQCA, this study achieves a more comprehensive understanding of BCT diffusion in PS-CED supervision. PLS-SEM establishes the dominant causal relationships and mediating structures at the aggregate level, while fsQCA reveals multiple equivalent and context-sensitive diffusion pathways at the configurational level. The dual-method approach not only enriches the findings obtained from each method individually, but also generates novel insights into differentiated promotion strategies for BCT diffusion under varying practical conditions [41].

4. Results

4.1. Normality, Bias and Robustness Analysis

This paper conducts a series of tests to address issues related to normality, non-response bias, and the robustness of the questionnaire-based data. As shown in

Table 2. Normality and biasness analysis.

Normality
Latent Variable	Kolmogorov–Smirnov			Shapiro–Wilk
	Statistic	Df	Sig.	Statistic	Df	Sig.
PEU	0.246	251	0	0.86	251	0
POE	0.174	251	0	0.937	251	0
AW	0.218	251	0	0.907	251	0
RNP	0.177	251	0	0.941	251	0
SCC	0.168	251	0	0.941	251	0
BDO	0.218	251	0	0.919	251	0
Non-response Bias
Latent Variable	First Thirty			Last Thirty
	Mean	SD		Mean	SD
PEU	3.491	0.712		3.642	0.704
POE	3.891	0.638		3.976	0.775
AW	3.659	0.645		3.833	0.607
RNP	3.614	0.634		3.502	0.653
SCC	3.707	0.521		3.568	0.627
BDO	3.688	0.707		3.577	0.743

, the Shapiro–Wilk test and Kolmogorov–Smirnov test are used to assess the distributional characteristics of the data. Significance values below 0.05 indicate that the normality assumption is violated [42], which is consistent with the application conditions of PLS-SEM. To examine potential non-response bias, the mean values and standard deviations of the first 30 respondents are compared with those of the last 30 respondents [43]. No statistically significant differences are observed between the two groups, suggesting that non-response bias is unlikely to affect the representativeness of the sample. Furthermore, the questionnaire design was grounded in established literature and refined through expert evaluation, which helps mitigate potential systematic bias associated with self-reported survey data and supports the reliability of the subsequent empirical analysis.

To address potential bias arising from the questionnaire-based data and to enhance the comprehensiveness of the overall data analysis, analysis of variance (ANOVA) is conducted to examine whether respondents’ demographic characteristics systematically influence the survey outcomes [44]. The questionnaire includes four fixed-category demographic variables (gender, position, work experience, and educational background), while all dependent variables are quantitative, making ANOVA an appropriate method for group comparison. SPSS 25.0 is used to perform the analysis. As shown in

Table 3. The overall ANOVA results.

Latent Variable	Gender		Position		Related Work Experience		Educational Background
	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
PEU	0.961	0.328	0.349	0.706	0.085	0.919	0.025	0.975
POE	2.383	0.124	0.706	0.611	0.919	0.979	0.975	0.949
AW	0.504	0.478	0.494	0.887	0.021	0.631	0.052	0.884
RNP	1.365	0.244	0.611	0.582	0.979	0.905	0.949	0.918
SCC	0.833	0.362	0.120	0.831	0.461	0.864	0.123	0.987
BDO	1.133	0.288	0.887	0.352	0.631	0.866	0.884	0.831

, the p-values of all demographic variables across the six latent constructs exceed the 0.05 significance threshold. This result indicates that respondents’ demographic characteristics do not exert statistically significant effects on the key constructs. Consequently, concerns regarding demographic-related bias in the questionnaire responses are alleviated, and the empirical results can be interpreted as reflecting substantive relationships among the theoretical constructs rather than artifacts of sample composition. This finding further supports the robustness and overall validity of the subsequent structural model analysis.

4.2. PLS-SEM Results

4.2.1. Measurement Model Assessment

The measurement model assessment serves as a critical step in ensuring that the questionnaire-based data provide a reliable and unbiased basis for evaluating the proposed model. SMARTPLS 4.0 is used to assess the reliability and validity of all latent constructs, and the estimated measurement model is presented in Figure 3, which depicts the relationships between latent variables and their observed indicators through standardized factor loadings. As illustrated in Figure 3, each indicator loads primarily on its corresponding latent construct, indicating a clear measurement structure without evident cross-loading issues. This graphical representation is supported by the statistical results reported in

Table 4. Convergent validity.

Latent Variable	Cronbach’s Alpha	Composite Reliability (rho_a)	Average Variance Extracted	Observed Variable	Factor Loadings	VIF
AW	0.726	0.734	0.647	AW1	0.834	1.549
				AW2	0.831	1.520
				AW3	0.744	1.311
BDO	0.747	0.758	0.664	BDO1	0.821	2.056
				BDO2	0.848	1.910
				BDO3	0.773	1.214
SCC	0.835	0.839	0.752	SCC1	0.880	2.133
				SCC2	0.881	2.012
				SCC3	0.840	1.778
POE	0.733	0.761	0.658	POE1	0.894	2.056
				POE2	0.847	1.910
				POE3	0.676	1.214
PEU	0.786	0.792	0.609	PEU1	0.817	1.964
				PEU2	0.728	1.716
				PEU3	0.776	1.666
				PEU4	0.797	1.626
RNP	0.819	0.822	0.734	RNP1	0.848	2.133
				RNP2	0.852	2.012
				RNP3	0.871	1.778

, where the Cronbach’s alpha and composite reliability (CR) values for all constructs exceed the recommended threshold of 0.7, confirming internal consistency reliability. The AVE values are all above 0.5, demonstrating adequate convergent validity, as each construct explains a substantial proportion of variance in its indicators. In addition, all variance inflation factor (VIF) values remain below 3 [45], suggesting that multicollinearity among indicators is negligible and that the estimation of the measurement model is not affected by redundancy-related bias. Discriminant validity is evaluated using the Fornell–Larcker criterion [46]. As shown in

Table 5. Fornell–Larcker criterion.

Variable	BI	DD	PBC	PBU	PU	SN
AW	0.848
BDO	0.744	0.817
SCC	0.529	0.667	0.867
POE	0.713	0.577	0.558	0.855
PEU	0.727	0.619	0.624	0.782	0.840
RNP	0.517	0.598	0.723	0.707	0.611	0.856

, the square root of AVE for each construct exceeds its correlations with other constructs, indicating that the latent variables capture conceptually distinct dimensions. Taken together, the consistency between the visual pattern observed in Figure 3 and the reliability and validity statistics confirms that the measurement model is well specified and suitable for subsequent structural model analysis.

4.2.2. Structural Model Assessment

The structural model assessment evaluates the explanatory and predictive performance of the hypothesized causal relationships among latent variables. As shown in Figure 4 and

Table 6. Quality of the structural model.

Variable	R2	Q2 Predict	RMSE	MAE	Exogenous Variables	f2
AW	0.812	0.804	0.447	0.307	PEU	0.774
					POW	0.140
BDO	0.864	0.845	0.397	0.282	PEU	0.348
					POE	0.012
					RNP	0.060
					SCC	0.311

, the R² for AW and BDO reach 0.812 and 0.864, respectively, indicating that the model explains more than 80% of the variance in both endogenous constructs. These results demonstrate strong explanatory power of the proposed structural model in the context of PS-CED supervision. Predictive relevance is further confirmed by the Q² predict values of 0.804 for AW and 0.845 for BDO, both of which are well above zero, suggesting robust out-of-sample predictive capability [45]. The pattern of path coefficients in Figure 4 highlights the central role of AW as a mediating construct linking technological, organizational, and institutional factors to diffusion outcomes. In particular, PEU exerts a substantial effect on AW, which in turn strongly influences BDO, indicating that usability primarily affects diffusion through enhancing cognitive awareness rather than through a direct pathway. Effect size (f²) analysis further supports this interpretation: PEU shows a large effect on AW and a moderate effect on BDO, whereas other predictors such as POE and RNP exhibit relatively small individual effect sizes. These smaller f² values do not imply theoretical insignificance, but instead reflect the inherent complexity of BCT diffusion in PS-CED supervision, where outcomes are shaped by the joint influence of multiple interacting factors rather than by a single dominant driver. This complexity is also consistent with the balanced RMSE and MAE values, which indicate stable predictive accuracy across the model. Taken together, the structural relationships visualized in Figure 4 and the associated diagnostic indicators confirm that the proposed model provides a reliable and well-specified representation of the causal mechanisms underlying BCT diffusion.

To further validate the hypothesized relationships, bootstrapping with 5000 resamples is applied to examine the significance of the path coefficients. As shown in

Table 7. Path coefficient results.

Hypotheses	Path Coefficient	T-Values	p-Values	Decision
AW → BDO	0.188	2.803	0.005	Supported
SCC → BDO	0.368	6.416	0.000	Supported
POE → AW	0.658	10.921	0.000	Supported
POE → BDO	−0.099	1.174	0.240	Not Supported
PEU → AW	0.280	4.439	0.000	Supported
PEU → BDO	0.220	2.481	0.013	Supported
RNP → BDO	0.366	5.818	0.000	Supported
POE → AW → BDO	0.124	2.831	0.005	Supported
PEU → AW → BDO	0.053	2.119	0.034	Supported

, AW, SCC, PEU, and RNP exhibit significant positive effects on BDO, supporting the corresponding hypotheses. POE and PEU also show significant positive effects on AW, confirming the motivational roles. In contrast, the direct path from PBU to diffusion outcome is not statistically significant. Overall, these results demonstrate that the structural model reliably captures the key mechanisms underlying BCT diffusion, while minimizing concerns related to model misspecification or spurious relationships.

4.2.3. Importance Performance Map Analysis (IPMA)

The IPMA complements the structural model results by jointly considering the relative importance and current performance of key antecedent variables influencing BCT diffusion in PS-CED supervision [47]. As illustrated in Figure 5 and summarized in

Table 8. IPMA results.

Variable	Importance	Performance
AW	0.188	64.001
SCC	0.368	55.125
POE	0.024	63.978
PEU	0.273	65.888
RNP	0.366	64.159

, importance is measured by the total effects of each construct on the target variable, while performance reflects the standardized average scores of the constructs. Figure 5 shows that SCC and RNP are positioned in the high-importance but relatively low-performance region. This discrepancy indicates that although organizational compatibility and regulatory support play a critical role in promoting BCT diffusion, their current implementation levels lag behind their strategic importance. From a managerial perspective, the low performance of SCC suggests constraints in operational coordination, technical preparedness, and system resilience within PS-CED supervision. Accordingly, priority actions should focus on enhancing system interoperability, strengthening platform node operation and disaster recovery mechanisms, and improving technical training and support. Similarly, the relatively low performance of RNP implies insufficient alignment between BCT applications and existing regulatory frameworks. This highlights the need to accelerate the integration of BCT systems with carbon market regulations, auditing procedures, and standardized data formats in order to reinforce regulatory legitimacy and compliance incentives. In contrast, constructs such as PEU and AW exhibit comparatively higher performance levels but lower relative importance, suggesting that further improvements in usability and cognition alone are unlikely to generate substantial gains in diffusion outcomes.

4.3. fsQCA Results

4.3.1. Calibration

The first step in fsQCA is to transform the raw data from PLS-SEM into fuzzy membership scores between 0 and 1. This process captures the degree to which cases belong to states of full membership, crossover, and full non-membership. In this paper, the 90%, 50%, and 10% quantiles of each variable distribution serve as the calibration thresholds for being fully-in, the cross-over point, and fully-out. The choice of these specific quantiles is justified because they effectively manage potential outliers and ensure that the membership scores reflect the actual diversity within the PS-CED supervision sample. This approach is widely recognized as a robust method when objective external standards for specific categories are not available. These thresholds represent the critical balance points between the presence and absence of causal conditions [43,48]. As shown in

Table 9. Descriptive statistics and calibrations.

Variable	N	Min	Max	Mean	S.D.	Fully-Out (10%)	Cross-Over (50%)	Fully-In (90%)
PEU	251	1	5	3.6434	0.712	−0.901	−0.182	1.933
POE	251	1	5	3.6799	0.725	−1.383	−0.040	1.833
AW	251	1	5	3.6985	0.713	−1.420	−0.069	1.834
RNP	251	1	5	3.7304	0.651	−1.124	−0.096	1.952
SCC	251	1	5	3.6587	0.705	−0.928	−0.466	1.911
BDO	251	1	5	3.7118	0.652	−1.080	−0.111	1.986

, the specific calibrated values for each influencing factor and diffusion outcome provide a practical and solid basis for the subsequent necessity and sufficiency analyses.

4.3.2. Necessary Condition Analysis

After calibrating the fuzzy set membership scores, the next step is a necessity analysis. This step identifies which causal conditions are indispensable for achieving a specific outcome when the preset consistency threshold is met. In this paper, the threshold is set at 0.9. Conditions with consistency values above 0.9 are considered strictly necessary. Conditions with consistency values between 0.8 and 0.9, combined with a coverage score higher than 0.75, are considered almost always necessary [49]. The results are reported in

Table 10. Results of Necessary Condition Analysis.

Configurational Constructs	High BDO		Configurational Constructs	Low BDO
	Consistency	Coverage		Consistency	Coverage
AW	0.841	0.760	AW	0.485	0.567
~AW	0.521	0.439	~AW	0.795	0.866
SCC	0.893	0.777	SCC	0.477	0.537
~SCC	0.469	0.409	~SCC	0.802	0.907
POE	0.839	0.775	POE	0.481	0.575
~POE	0.540	0.446	~POE	0.811	0.867
PEU	0.855	0.822	PEU	0.421	0.524
~PEU	0.505	0.403	~PEU	0.857	0.884
RNP	0.847	0.837	RNP	0.425	0.543
~RNP	0.538	0.420	~RNP	0.873	0.881

. High-level BCT diffusion does not depend entirely on a single factor. However, some variables demonstrate strong quasi-necessity. SCC shows the highest consistency, approaching the 0.9 threshold. This suggests that most cases of high-level diffusion involve respondents with strong system control perceptions. Other variables yield consistency values between 0.83 and 0.86, indicating a moderate tendency toward necessity. For low-level BCT diffusion, the absence of PEU, RNP, and SCC all show consistency above 0.85. This suggests that regulatory failure is often linked to weak perceptions of utility, regulatory norms, and control. Overall, multiple factors act together across dimensions to support the outcome.

4.3.3. Sufficient Conditions Analysis

Following the necessity analysis, this paper applies fsQCA to identify sufficient configurations that lead to BCT diffusion. The analysis uses a consistency threshold of 0.80 and a raw coverage threshold of 0.27 to ensure robust results [50,51]. As

Table 11. Sufficient configurations for predicting of BDO.

Configuration	High DD		Low DD
	1	2	3	4	5	6
AW	●	⊗	•		⊗	⊗
SCC	●	•	⊗	⊗		•
POE	●	⊗		⊗		•
PEU	●	●		⊗	⊗
RNP		●			⊗	⊗
Raw Coverage	0.695	0.285	0.438	0.665	0.698	0.306
Unique Coverage	0.454	0.043	0.034	0.038	0.060	0.007
Consistency	0.937	0.926	0.964	0.969	0.974	0.978
Overall Solution Coverage	0.739		0.802
Overall Solution Consistency	0.928		0.946

Note: • indicates that boundary conditions exist; ● indicates that the core conditions exist; ⊗ indicates that the condition is missing. A blank space indicates that the presence or absence of this condition has no effect on the result.

shows, the overall solutions for both high and low diffusion levels demonstrate high consistency and coverage, confirming the validity of these configurational pathways. For high-level diffusion, two practical strategic paths emerge. Configuration 1 represents a proactive strategy where internal readiness is the key. When stakeholders possess high awareness, strong willingness to adopt, and effective control over the system, technology diffusion reaches a high level through internal motivation. Configuration 2 represents a policy-led strategy. This path suggests that even if internal willingness or technical familiarity is low, high diffusion is still possible through strong regulatory support and essential platform utility. For practitioners, this means that policy mandates can compensate for a lack of initial stakeholder enthusiasm. Regarding low-level diffusion, the results highlight specific barriers that hinder progress. Configuration 3 shows that a gap between intention and execution occurs when stakeholders want to adopt but lack operational control. Configuration 4 identifies technology indifference as a major obstacle, where a lack of basic understanding and control prevents any motivation to move forward. Configurations 5 and 6 illustrate systemic failure caused by a total lack of institutional support and technological recognition. These findings suggest that managers must focus on both technical training and policy stability to avoid diffusion failure.

5. Discussion

5.1. Diffusion Effects of Multiple Key Factors

The effectiveness of BCT diffusion in PS-CED supervision is shaped by the interaction among multiple drivers and their alignment with organizational capacity, institutional frameworks, and user perceptions. From the technology and organizational perspective, PEU and SCC jointly constitute the core conditions for sustainable BCT operation. PEU exerts both a direct effect on BDO and an indirect effect through AW, and emerges as a core element in the fully empowered configuration in fsQCA, highlighting the central role of regulatory compliance capability [52]. Unlike prior studies emphasizing ease of interaction or deployment [9], this paper indicates that traceability, verifiability, and accountability of PS-CED are the primary determinants of diffusion effectiveness [53]. SCC reflects organizational control over node operation, permission management, and system recovery, serving as a key mechanism for translating technical capacity into stable organizational behavior [33]. Its high importance but low performance, together with its absence in the willing but unable configuration, explains why organizations remain cautious when control capacity and system reliability are uncertain [26,54]. Within the cognitive and institutional shaping dimension, AW functions as a mediating variable in the PLS-SEM model, linking constructs such as PEU and POE with BDO [31,55]. In PS-CED supervision, diffusion is driven less by functional convenience than by perceived compliance value. Regulatory norm perception functions as a critical external driver, compensating for limited operational capacity or technical familiarity and enabling top-down diffusion, particularly for resource-constrained or early-stage regulatory actors [32].

5.2. Clarification of Diffusion Mechanism

PLS-SEM captures the linear effects of key antecedents and the mediating role of behavioral intention, whereas fsQCA identifies distinct high and low performance pathways arising from interactions among technological, institutional, cognitive, and intentional conditions. The alignment between the comprehensive empowerment configuration and the significant PLS-SEM paths confirms that effective diffusion depends on coordinated multidimensional support rather than any single capability [56]. Accordingly, neither technical capacity nor institutional enforcement alone can ensure diffusion success. Diffusion requires the joint strengthening of system control capacity, institutional support, perceived utility, and adoption intention [28,57]. Moreover, the externally driven pathway indicates that in institutionally intensive regulatory contexts, BCT diffusion can occur through top-down enforcement even in the absence of strong intrinsic motivation [27,58]. In contrast, low performance configurations reveal that cognitive awareness without adequate technological control capacity leads to diffusion failure, highlighting the limits of intention-based explanations [59]. Collectively, these findings demonstrate the value of combining PLS-SEM and fsQCA for uncovering both enabling and constraining mechanisms, thereby offering clearer guidance for governance design and policy intervention.

5.3. BCT Diffusion Strategy Framework

Based on the clarified roles of key drivers and diffusion mechanisms identified through the PLS-SEM and fsQCA analyses, this study develops an integrated BCT diffusion strategy framework for PS-CED supervision. The framework is structured along three core dimensions that are widely emphasized in the literature on digital technology diffusion and governance innovation: organizational operational capacity, institutional and regulatory alignment, and user cognitive status. Together, these dimensions capture the technological, organizational, and institutional conditions shaping BCT diffusion in complex supervisory systems. As illustrated in Figure 6, the framework synthesizes these dimensions into three archetypal diffusion trajectories, each corresponding to a distinct configuration of conditions and strategic priorities. These trajectories do not represent sequential stages of diffusion, but rather alternative strategic logics through which organizations may advance blockchain-enabled supervision under different structural constraints, consistent with prior research on heterogeneous diffusion pathways and configurational governance mechanisms [17,60].

The first trajectory, referred to as endogenous empowerment, applies to organizations with strong digital infrastructure, high operational capacity, and relatively mature regulatory alignment. In such contexts, diffusion is primarily driven by internal capability enhancement and focuses on deepening the functional integration of BCT into end-to-end supervisory processes, including real-time carbon data registration, traceable verification, and compliant settlement. Strategic priorities emphasize improving platform utility, strengthening node governance, and leveraging smart contracts to enhance control capability and data credibility, which aligns with existing studies on internally driven digital transformation and blockchain-enabled process optimization [8,57,61]. The second trajectory follows an externally regulated diffusion logic, which characterizes organizations operating under strong institutional and regulatory pressure but with limited technical resources and cognitive readiness. In this case, diffusion is initially compliance-oriented and facilitated through external mechanisms such as standardized integration requirements, policy incentives, pilot demonstrations, and managed platform services. These instruments reduce entry barriers and support gradual capability accumulation, enabling a transition from passive compliance toward more active use. The third trajectory represents a coordinated intervention strategy for organizations facing multiple constraints, including weak adoption intention, limited operational capacity, and insufficient institutional understanding. Under such conditions, diffusion requires simultaneous interventions across multiple dimensions. Strategic priorities include enhancing cognitive awareness of BCT value, lowering technical deployment barriers, and clarifying institutional standards related to data fields, auditing procedures, and carbon-trading interfaces. Similar multi-pronged intervention approaches have been emphasized in the literature on early-stage technology diffusion and governance capacity building in regulated sectors [21].

It should be noted that the diffusion strategy framework presented in Figure 6 is developed based on empirical evidence from a limited number of representative projects. Accordingly, the identified diffusion trajectories are intended to illustrate typical strategic logics rather than to serve as universally prescriptive pathways. Their applicability may vary depending on organizational maturity, institutional settings, and regional regulatory conditions.

6. Conclusions

This study examines the diffusion of BCT in PS-CED supervision, addressing a critical challenge in the pursuit of sustainable energy governance and low-carbon transition. In the context of increasingly stringent requirements for data transparency, traceability, and accountability in carbon management, the effective diffusion of digital technologies such as BCT is essential for enhancing the credibility and long-term sustainability of carbon emission supervision systems.

The findings demonstrate that BCT diffusion in PS-CED supervision is driven by synergistic interactions among technological, organizational, cognitive, and institutional factors. Perceived functional value and operational control capability jointly emerge as the core drivers of diffusion, indicating that technological advantages alone are insufficient to support sustainable supervision outcomes. Instead, effective diffusion depends on the coordinated alignment of stable node operation, transparent permission management, adoption intention, and regulatory norm perception. This alignment links technological cognition with organizational practices and policy frameworks, thereby enabling reliable, trustworthy, and sustainable carbon data governance. Moreover, the study reveals pronounced configurational diversity and path heterogeneity in BCT diffusion. High-performance diffusion can be achieved through multiple pathways, including an internally driven empowerment path based on organizational capability synergy and an externally enforced path shaped by regulatory pressure and platform usability. In contrast, low-performance configurations expose persistent barriers such as the willing-but-unable dilemma, technological indifference, and compounded capability gaps. These findings underscore that sustainable digital transformation in carbon supervision does not follow a single linear trajectory, but rather depends on context-specific combinations of enabling conditions.

By integrating PLS-SEM and fsQCA analyses, this study proposes a BCT diffusion strategy framework that translates empirical insights into differentiated pathways for sustainable implementation. For organizations with strong digital capacity and auditing responsibilities, sustainability-oriented diffusion should emphasize deep operational integration of BCT into carbon data recording, verification, and quota management processes. For resource-constrained actors operating under regulatory pressure, standardized interfaces and compliance-oriented adoption provide a pragmatic entry point for improving data reliability and regulatory compliance. For participants lacking both willingness and capacity, coordinated interventions in training, operational support, and institutional guidance are necessary to gradually build the foundations for sustainable adoption.

This study deepens the understanding of BCT diffusion in carbon governance from a structured and multi-method perspective and demonstrates how digital technologies can support sustainable supervision practices in energy systems. Despite these contributions, several limitations should be acknowledged. The empirical analysis is grounded in data collected from a limited number of representative projects within a specific regional and institutional context. While these projects provide in-depth insights into BCT diffusion in PS-CED supervision, the identified diffusion mechanisms and strategic trajectories should be interpreted with caution when extended to broader contexts. The study also relies on self-reported survey data, which may be subject to perceptual and response biases. Moreover, the sample is regionally bounded and embedded within a particular regulatory and governance regime. Institutional arrangements, regulatory frameworks, and organizational cultures related to power systems and carbon emission governance vary substantially across regions, governance regimes, and energy markets. Such differences may influence both the relative importance of diffusion drivers and the feasibility of specific diffusion strategies, thereby limiting the direct transferability of the findings. In addition, the fsQCA results depend on calibration choices and threshold settings, which may affect configurational interpretations. Future research could address these limitations by incorporating multi-source and longitudinal data, expanding cross-regional and cross-institutional comparisons, and applying alternative calibration and robustness checks to further validate and refine the proposed diffusion mechanisms and strategy framework.