Digital Oasis: How Green Infrastructure Is Reshaping China’s Energy Resilience Landscape

Abstract

In the context of global energy transition and climate change, energy system resilience has become critical for countries worldwide. While green digital infrastructure—emerging from the integration of digitalization and low-carbon development—shows theoretical potential to strengthen energy resilience, empirical evidence remains limited. This study utilizes China’s 2015 Green Data Center Pilot Policy as a quasi-natural experiment to examine this relationship through comprehensive panel data from 30 Chinese provinces spanning 2011–2021. We developed an integrated energy resilience evaluation framework across four dimensions: economic resilience, engineering resilience, resource resilience, and ecological resilience, applying the CRITIC method to determine objective indicator weights. Our difference-in-differences analysis demonstrates that green digital infrastructure significantly enhances regional energy resilience, with pilot regions experiencing a 2.83% improvement compared to non-pilot areas. This impact shows regional heterogeneity, with stronger effects in economically developed areas with better digital foundations. We identify two primary mechanisms through which green digital infrastructure influences energy resilience: industrial structure optimization (particularly through service industry growth) and enhanced innovation capacity. These findings provide robust empirical support for green digital infrastructure’s role in strengthening energy system stability and adaptability, offering valuable policy insights for promoting both digitalization and low-carbon transition under global climate governance.

1. Introduction

Against the backdrop of global climate change and energy transition, energy security and resilience have become major challenges for countries worldwide. With the increasing frequency of extreme weather events [1], intensified geopolitical conflicts, and growing instability in energy supply chains, the vulnerability of traditional energy systems has become increasingly evident [2,3]. Meanwhile, as the wave of digital transformation sweeps across the globe, green digital infrastructure, as a product of the combination of digital technology and sustainable development concepts [4], has provided new possibilities for the intelligent management and resilience enhancement of energy systems while promoting high-quality economic development. In this context, exploring the relationship between green digital infrastructure and energy resilience has significant theoretical and practical implications, providing important reference value for guiding countries in addressing climate change, ensuring energy security, and achieving sustainable development goals.

In recent years, green infrastructure as a multifunctional concept has received widespread attention globally [5,6,7]. With the development of digital technology, green infrastructure has integrated with digitalization to form the new concept of green digital infrastructure, which emphasizes enhancing environmental adaptability and resource utilization efficiency through digital technology. Currently, academic research on green digital infrastructure mainly focuses on its environmental benefits, resource conservation, and ecological functions, but research on its role in energy systems is relatively limited. Studies have shown that the application of digital technology can significantly improve resource allocation efficiency and enhance the flexibility and reliability of system operation [8,9], but how these potential benefits specifically translate into improvements in energy system resilience still lacks systematic theoretical explanation and empirical support.

At the same time, energy resilience, as a key indicator for addressing uncertainty and vulnerability in energy systems, has gradually become a focus of attention for academics and policymakers. Research indicates that energy resilience is not only related to the stability of the energy system itself but also closely linked to economic development, social welfare, and environmental sustainability [10]. Scholars have explored concept definitions and measurement methods for energy resilience from different perspectives, including constructing comprehensive evaluation systems from dimensions such as energy access, energy diversification, and energy efficiency [11,12,13]. However, existing research mostly focuses on the impact of energy system characteristics or external factors on energy resilience, while research on how digitalization, especially green digital infrastructure, affects energy resilience remains insufficient.

Despite growing interest in both green digital infrastructure and energy resilience as separate fields, significant research gaps remain in understanding their interconnection. First, while the existing literature has examined the environmental benefits and resource conservation aspects of green infrastructure, research specifically investigating its impact on energy system resilience is notably scarce. Second, current energy resilience studies have primarily focused on conceptual frameworks and isolated case analyses rather than providing systematic empirical evidence across diverse regions. Third, the mechanisms through which digital technologies influence energy resilience remain underexplored, limiting our understanding of how to effectively leverage these technologies for enhancing energy system stability. This study addresses these gaps by providing the first comprehensive empirical analysis of the causal relationship between green digital infrastructure and energy resilience using a quasi-natural experimental design.

Based on the above background, this study attempts to answer the following key questions: does the development of green digital infrastructure significantly impact energy resilience? If so, what is the direction and strength of this impact? Through what mechanisms does green digital infrastructure affect energy resilience? Does this impact vary across different regions and conditions? Answering these questions not only helps to fill gaps in the existing literature but also provides a basis for formulating more targeted policies, helping countries to enhance the resilience and sustainability of energy systems while advancing digital transformation.

The main contributions of this research are reflected in the following aspects: first, by constructing an energy resilience evaluation system encompassing four dimensions—economic resilience, engineering resilience, resource resilience, and ecological resilience—we advance methodological approaches to energy resilience measurement, moving beyond existing frameworks that often focus on limited aspects of resilience. Second, leveraging China’s green data center pilot policy as a quasi-natural experiment and applying the difference-in-differences method, we provide the first causal empirical evidence on how green digital infrastructure development impacts energy resilience, establishing a clear relationship that has previously been theorized but not empirically verified. Third, we systematically identify and quantify the specific transmission mechanisms through which green digital infrastructure affects energy resilience, revealing industrial structure optimization and innovation capacity enhancement as key pathways. This mechanistic understanding was notably absent in the previous literature. Finally, through heterogeneity analysis, we reveal previously unidentified regional differences in how green digital infrastructure influences energy resilience. These contributions not only address critical knowledge gaps between green digital infrastructure and energy resilience but also offer novel methodological approaches for examining the digital–energy nexus in the context of global climate change mitigation.

2. Literature Review and Hypothesis Development

2.1. Policy Background

In recent years, global efforts to address climate change and promote sustainable development have intensified, with countries increasingly focusing on integrating digitalization and low-carbon strategies. China, as a major global emitter and a leader in digital infrastructure development, has prioritized green digital infrastructure in its policy agenda. In 2015, the Chinese government launched the Green Data Center Pilot Policy to promote energy-efficient and environmentally friendly data center construction across provinces. This policy aimed to reduce carbon emissions from digital infrastructure while enhancing energy resilience through technological innovation and structural optimization.

The policy’s pilot nature provided a quasi-natural experimental framework, enabling researchers to evaluate the impact of green digital infrastructure on regional energy resilience. By incentivizing the adoption of advanced energy management systems and renewable energy integration, the policy sought to address both immediate energy challenges and long-term sustainability goals. This study leverages the policy’s implementation as a key contextual backdrop, examining its role in strengthening energy system stability and adaptability through green digital infrastructure development.

2.2. Literature Review

Green infrastructure, as a multifunctional concept, has been widely studied and applied globally. Ying et al. (2022) pointed out that green infrastructure has become an effective tool for coordinating environmental, social, and economic development, and is an important strategy for achieving sustainable development goals [14]. Through analyzing 2194 relevant papers between 1995 and 2019, they found that green infrastructure research has shown significant growth in recent years, primarily focusing on ecosystem services, stormwater management, biodiversity conservation, and climate change response. These studies have provided an important reference for environmental sustainable development. Zareba (2014) analyzed the application of green infrastructure in contemporary research from multifunctional and multi-scale perspectives, emphasizing green infrastructure’s contribution to improving urban sustainability and resilience, especially its role in urban forestry and water resource protection [15]. Kim and Tran (2018) conducted content analysis of comprehensive plans in 60 U.S. cities and found that current urban planning fails to fully incorporate key green infrastructure principles, requiring improvement through detailed policies, action strategies, and implementation plans [16]. Jerome et al. (2019) proposed a quality assessment framework for green infrastructure in the UK’s built environment, emphasizing that high-quality green infrastructure should be based on multifunctional network construction and long-term management principles, while also considering health and wellbeing, water resource management, and nature conservation [17].

With the development of digital technology, green infrastructure has increasingly integrated with digitalization, forming the new concept of green digital infrastructure. Wang et al. (2024) showed that the application of digitalization and intelligent technology in green infrastructure construction can significantly improve urban flood control and disaster reduction capabilities, optimizing resource allocation [18]. Chen et al. (2022), through morphological spatial pattern analysis and landscape connectivity evaluation, studied the spatiotemporal characteristics of green infrastructure networks, providing methodological support for digital-technology-based green infrastructure optimization [19]. Ferreira et al. (2021) proposed a multi-criteria method for designing green infrastructure in Setubal, Portugal, effectively integrating the roles of ecological and social components in planning and policy-making processes [20]. These studies indicate that the introduction of digital technology is changing the planning and management methods of traditional green infrastructure, enhancing its ability to address climate change and environmental challenges.

Meanwhile, energy resilience, as a relatively new research field, has received increasing attention in recent years. Gatto and Drago (2020) classified energy resilience through bibliometric analysis methods, identifying seven main methods or strategies including social, economic, and environmental aspects [12]. Their research shows that the concept of energy resilience has gradually expanded from initial applications in technical and hard science fields to social sciences and sustainable development research, becoming a holistic approach centered on sustainable development. Dong et al. (2021) constructed a comprehensive energy resilience index based on three components (energy access, renewable energy, and energy efficiency) and studied the impact of geopolitical risk on energy resilience [21]. They found that energy resilience is positively correlated with carbon dioxide emissions, with this positive correlation stemming from the scale effect of energy resilience on carbon dioxide emissions completely offsetting the technology and composition effects. Zhou (2023) reviewed energy resilience in energy regions, pointing out that energy efficiency, flexibility, and robustness can promote system sustainability and decarbonization under low-impact high-probability events, while energy resilience is crucial for addressing high-impact low-probability events [22].

Energy resilience research also covers multiple levels and dimensions. He et al. (2019) proposed a multi-regional input–output linear programming model to study the energy resilience of multi-regional economies that considers sectoral and inter-regional dependencies in economic structures and can calculate the impact of random energy production disruptions on the entire multi-regional input–output system [23]. Li and Zhang (2023), based on a grey correlation projection model, conducted spatiotemporal measurements of energy resilience in 30 Chinese provinces, finding that the spatiotemporal pattern of China’s energy resilience underwent significant changes from 2005 to 2018, with energy endowment being the primary condition for ensuring regional energy resilience, while renewable energy development, energy investment, economic development, and policy coordination played crucial roles in ensuring regional energy resilience [24]. Sharmin and Dhakal (2022) developed a “composite energy resilience index” for Bangladesh containing 15 individual indicators, emphasizing that energy resilience critically depends on a range of affordability, sustainability, and availability issues, as well as aspects of energy system management quality [25].

Although green digital infrastructure and energy resilience have each become research hotspots, there is a lack of studies linking the two. Therefore, this paper’s systematic research on green digital infrastructure and energy resilience not only fills gaps in the existing literature but also provides new perspectives and methods for addressing global challenges such as climate change, energy security, and sustainable development. By systematically discussing the connection between the two, it helps to formulate more comprehensive and effective policies to enhance the sustainability and adaptability of energy systems.

2.3. Research Hypotheses

As global energy transition and digital trends converge, green digital infrastructure—a product combining digital technology with sustainable development concepts—plays an increasingly prominent role in building resilient energy systems. Green digital infrastructure includes energy-efficient data centers, smart grids, communication networks, and supporting software systems [26], which not only serve as carriers for digital economic development but also provide critical support for the intelligent and low-carbon transformation of energy systems. From a systems theory perspective, energy resilience manifests as the ability of energy systems to maintain basic functions and quickly restore equilibrium when facing external shocks [27], a characteristic highly consistent with the self-organization and self-adaptive properties of complex adaptive systems. Green digital infrastructure enhances the core characteristics of energy systems as complex adaptive systems by improving information processing capabilities and feedback mechanism efficiency. Existing research shows that the application of digital technology can optimize energy resource allocation, improve system operational efficiency, and enhance the dynamic balance adjustment capability between energy supply and demand [28], all of which are key elements in building resilient energy systems. Ecological resilience theory emphasizes a system’s ability to maintain structure and function when facing disturbances, while green digital infrastructure enhances the energy system’s adaptability to various disturbances by building smarter and more efficient energy infrastructure networks [29]. Based on the above theoretical analysis and existing research findings, this study proposes the first research hypothesis:

H1. 

In the Chinese context, green digital infrastructure development has a positive impact on energy resilience.

Industrial structure theory suggests that the advancement, rationalization, and low-carbon transformation of industrial structure are important indicators of high-quality economic development and key pathways to enhancing energy system resilience [30]. As an important component of new infrastructure, green digital infrastructure promotes industrial structure evolution toward a more low-carbon and efficient direction by digitally empowering traditional industries and fostering emerging industries. Green digital infrastructure not only reduces information acquisition and transaction costs, promoting optimal resource allocation across industries [31], but also improves energy utilization efficiency across various sectors through intelligent control and refined management. From an industrial ecology perspective, green digital infrastructure promotes upstream and downstream collaboration in industrial chains and cross-industry integration, enhancing the industrial system’s shock resistance [32]. Dynamic capability theory suggests that organizations respond to environmental changes by integrating, building, and reconfiguring internal and external resources, with green digital infrastructure serving as an important tool for enhancing this dynamic capability [33]. Through the optimization and adjustment of industrial structure, the energy system’s supply structure becomes more diversified, the demand structure becomes more rational, and the market structure becomes more perfected, thus exhibiting stronger resilience when facing external shocks. The structure–conduct–performance paradigm also supports this view, suggesting that industrial structure changes driven by green digitalization lead to behavioral pattern transformations, ultimately improving overall system performance, including enhanced resilience. Based on this, this study proposes the second research hypothesis:

H2. 

In the Chinese context, green digital infrastructure development positively impacts energy resilience through industrial structure optimization.

Innovation theory has long emphasized the core driving role of technological progress and innovation in economic and social development. Schumpeterian innovation theory states that innovation is the fundamental driving force of economic development, especially when facing resource constraints and environmental challenges, as innovation can break through development bottlenecks and create new possibilities. Green digital infrastructure creates favorable conditions for technological innovation in the energy sector by providing powerful computing capabilities, rich data resources, and efficient network connectivity. Open innovation theory suggests that innovation relies not only on internal R&D but also on integrating external knowledge and resources, with green digital infrastructure serving as an important platform for promoting knowledge flow and resource sharing [34]. In the energy sector, innovations supported by green digital infrastructure are primarily manifested in renewable energy technology advancement, energy storage technology breakthroughs, and demand-side response technology development, all of which collectively enhance the flexibility and adaptability of energy systems. From the perspective of technological evolution path dependency theory, green digital infrastructure promotes the development of energy technology systems toward cleaner and more efficient directions by changing innovation incentive mechanisms and reducing innovation costs [35]. Within the framework of innovation diffusion theory, green digital infrastructure accelerates the promotion and application of new energy technologies and energy-saving technologies, making energy systems more adaptable to external environmental changes in both structure and function [36]. National innovation system theory further indicates that green digital infrastructure, as an important component of innovation infrastructure, enhances interaction and collaboration among innovation entities [37] and improves the efficiency of the entire innovation system, thereby enhancing the resilience of energy systems in response to complex changes. Based on the above theoretical analysis, this study proposes the third research hypothesis:

H3. 

In the Chinese context, green digital infrastructure development positively impacts energy resilience through enhanced innovation capacity.

These three hypotheses form the theoretical foundation of our research framework, as illustrated in Figure 1. The framework depicts the proposed relationship between green digital infrastructure development and energy resilience (H1), along with the two potential mechanisms—industrial structure optimization (H2) and innovation capacity enhancement (H3)—that may mediate this relationship. This conceptual structure guides our empirical investigation, which employs actual data and rigorous econometric methods to test these hypothesized relationships. By systematically examining both the direct effects and underlying mechanisms, we aim to provide a comprehensive understanding of how green digital infrastructure influences energy resilience across different regional contexts in China. The subsequent empirical analysis sections will present evidence that either supports or refutes these hypotheses, contributing to both theoretical knowledge and practical policy formulation in this emerging field.

3. Research Design

3.1. Data Sources

This study focuses on 30 provinces, municipalities, and autonomous regions in mainland China, spanning from 2011 to 2021. Tibet, Taiwan, Hong Kong, and Macau were excluded from the research scope due to serious data deficiencies. The data used in this study primarily come from the National Bureau of Statistics database, EPS data platform, and statistical yearbooks of corresponding years for each province (municipality, autonomous region), as well as various industry and professional statistical yearbooks, including authoritative data sources such as “China Statistical Yearbook”, “China Electric Power Yearbook”, “China Environmental Statistical Yearbook”, “China Labor Statistical Yearbook”, and “China Energy Statistical Yearbook”. For occasional missing values in the dataset, this study employed linear interpolation or ARIMA interpolation methods to supplement the data, ensuring data completeness and reliability of the analysis results. Data related to green digital infrastructure development were primarily obtained from the “Notice on Announcing the List of National Green Data Center Pilot Regions” issued by the Ministry of Industry and Information Technology, the National Government Offices Administration, and the National Energy Administration in 2015. The construction of the energy resilience indicator system was based on statistical data across multiple dimensions. Data for measuring industrial structure optimization and innovation capacity as mechanism variables were likewise sourced from the aforementioned statistical databases and yearbooks, ensuring the consistency and authority of data sources. During data processing, this study fully considered regional differences and time trends, standardizing the original data to eliminate the impact of different indicator measurement unit differences on research results. Simultaneously, through the construction of panel data models, unobservable regional fixed effects and time fixed effects were effectively controlled, enhancing the robustness and reliability of the research conclusions.

Our study period from 2011 to 2021 was carefully selected based on several considerations. The starting year of 2011 corresponds to the beginning of China’s 12th Five-Year Plan period, which marked a significant shift in national policy emphasis toward ecological civilization and green development strategies. This period represents a critical turning point in China’s approach to energy and environmental governance, making it an appropriate baseline for examining subsequent policy impacts. The endpoint of 2021 was chosen to ensure data reliability and completeness. While Statistical Yearbooks for 2022 and 2023 have been released in some provinces, comprehensive and verified data across all indicators and regions in our framework were not consistently available beyond 2021 at the time of our analysis. Furthermore, this 11-year period provides sufficient pre-treatment observations (2011–2014) and post-treatment observations (2015–2021) to effectively implement our difference-in-differences methodology and capture both immediate and medium-term effects of the green data center pilot policy.

3.2. Variable Definitions

The measurement of green digital infrastructure development (GDID) is based on policy implementation. According to the “Notice on Announcing the List of National Green Data Center Pilot Regions” issued by the Ministry of Industry and Information Technology, the National Government Offices Administration, and the National Energy Administration in 2015, cities with pilot green data centers were designated as the experimental group, while non-pilot regions served as the control group. When region i is included in the green data center pilot at time t, GDID = 1; otherwise, GDID = 0. This measurement method based on policy impact aligns with the design concept of quasi-natural experiments, effectively avoiding endogeneity issues and enhancing the credibility of research conclusions.

The measurement of the dependent variable, energy resilience (ERES), adopts a comprehensive evaluation system with multiple dimensions and indicators. The research constructs a comprehensive energy resilience evaluation system with 18 indicators selected from four dimensions: economic resilience, engineering resilience, resource resilience, and ecological resilience, as shown in

Table 1. Comprehensive energy resilience evaluation system.

Primary Indicators	Secondary Indicators	Tertiary Indicators	Attribute
Economic Resilience	Economic Operation	Energy Industry Investment (CNY 100 million)	+
		Industrial Producer Price Index	+
		Total Energy Consumption/GDP (10,000 tons of standard coal/CNY 100 million)	−
	Economic Structure	Industrial Added Value/GDP (%)	+
		Total Import and Export/GDP (%)	+
	Economic Potential	Number of Patent Applications	+
Engineering Resilience	Supply and Operation Facilities	Power Generation Capacity (10,000 kW)	+
		Transmission Line Circuit Length (km)	+
		Total Gas Pipeline Length (km)	+
	Information Transmission	Internet Penetration Rate (%)	+
		Mobile Phone Penetration Rate (units/100 people)	+
Resource Resilience	Energy Production	Energy Production/Year-end Total Population (10,000 tons of standard coal/10,000 people)	+
		Energy Production Diversification Index	+
	Energy Consumption	Energy Production/Total Energy Consumption (%)	+
		Energy Consumption/Year-end Total Population (10,000 tons of standard coal/10,000 people)	−
Ecological Resilience	Ecological Security	Forest Coverage Rate (%)	+
	Environmental Pollution	Per Capita Carbon Emissions (10,000 tons/10,000 people)	−
		Incidents of Sudden Environmental Events (times)	−

. Economic resilience reflects the ability of economic systems to recover to a stable state after disturbances caused by changes in energy systems, including economic operation (energy industry investment, industrial producer price index, energy consumption per unit of GDP), economic structure (industrial added value as a proportion of GDP, total import and export as a proportion of GDP), and economic potential (number of patent applications). Engineering resilience embodies the engineering hardware guarantee for energy systems to recover to a stable state when facing external shocks, covering supply and operation facilities (power generation capacity, transmission line circuit length, total gas pipeline length) and information transmission (Internet penetration rate, mobile phone penetration rate). Resource resilience represents the supportive role of each region’s inherent resource endowment in restoring energy system stability, including energy production (per capita energy production, energy production diversification index) and energy consumption (energy self-sufficiency rate, per capita energy consumption). Ecological resilience reflects the ability of regional ecological environments to respond to energy system fluctuations under safe decarbonization requirements, including ecological security (forest coverage rate) and environmental pollution (per capita carbon emissions, incidents of sudden environmental events). As shown in Figure 2, the comprehensive energy resilience evaluation system integrates these four dimensions to capture both the static capacity and dynamic recovery potential of regional energy systems. The multidimensional approach allows for a more nuanced assessment of resilience than single-factor measures, reflecting the complex inter-relationships between economic structures, physical infrastructure, resource availability, and environmental constraints that collectively determine a region’s ability to maintain energy stability when confronted with various disturbances.

This study adopts the CRITIC comprehensive evaluation method to determine the weight of each indicator. The specific calculation steps are as follows. First, the original data are normalized using the range normalization method with the following formula:

$$\begin{array}{ll}& Y_{ij}={\displaystyle \frac{X_{ij}-min\left(X_i\right)}{max\left(X_i\right)-min\left(X_i\right)}}\left(\mathrm{F}\mathrm{o}\mathrm{r} \mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{s}\right)\\ & Y_{ij}={\displaystyle \frac{max\left(X_i\right)-X_{ij}}{max\left(X_i\right)-min\left(X_i\right)}}(\mathrm{For} \mathrm{negative} \mathrm{indicators})\end{array}$$

(1)

where $X_{ij}$ represents the actual value of each indicator, $Y_{ij}$ represents the standardized indicator value, $max\left(X_i\right)$ represents the maximum value of the i-th indicator, and $min\left(X_i\right)$ represents the minimum value of the i-th indicator.

Further calculate the standard deviation of each standardized indicator with the following calculation formula:

$${\sigma }_j=\sqrt{{\displaystyle \frac{\sum _{i=1}^n {\left(Y_{ij}-\overline{Y_j}\right)}^2}{n-1}}}, \left(j=\mathrm{1,2},\dots ,m\right)$$

(2)

Determine the symmetric matrix constituted by the linear correlation coefficients $r_{jk}$ between indicators, and calculate the conflict coefficient between indicator j and the decision situations determined by other indicators, determining the information entropy of each indicator, with the following calculation formula:

$$c_j={\sigma }_j\sum _{k=1}^m \left(1-r_{jk}\right)$$

(3)

Calculate the weight w_j by calculating the information redundancy as follows:

$$w_j={\displaystyle \frac{c_j}{\sum _{j=1}^m c_j}}$$

(4)

The comprehensive score s_i for each scheme is calculated as follows:

$$s_i=\sum _{j=1}^m w_j{Y}_{ij}$$

(5)

This method reflects the contrast intensity and conflict coefficient between indicators by calculating the column vector standard deviation and correlation coefficients of indicators, considering both the information volume of indicators and the correlation between indicators, making it more scientific and reliable than subjective weighting methods.

The mechanism variable industrial structure optimization (ISO) uses the proportion of tertiary industry added value to GDP as a measurement indicator. This indicator effectively reflects the degree of economic structure transition toward service-oriented and low-carbon directions, closely related to energy consumption intensity and patterns. The mechanism variable innovation capacity (IC) uses the natural logarithm of the total number of patent applications as a measurement indicator. The number of patent applications is an important indicator for measuring regional innovation activities and technological progress, reflecting the overall level of technological innovation in the region.

Considering that regional energy resilience is influenced by multiple factors, this study incorporates the following control variables to improve the accuracy and robustness of model estimation. The economic development level (PGDP) is measured by the natural logarithm of per capita regional gross domestic product, reflecting regional economic development conditions and resident income levels, which directly affect energy consumption structure and efficiency levels. The urbanization rate (UR) is represented by the proportion of urban population to total population; the urbanization process not only changes the energy demand structure but also affects the layout and efficiency of energy infrastructure. Industrial structure (IS) is represented by the proportion of secondary industry added value to GDP, measuring regional industrialization level; the industrial sector is typically a major energy consumer, and its development status directly affects energy resilience. The degree of openness (OPEN) is represented by the proportion of total imports and exports to GDP, reflecting the degree of regional participation in international division of labor and trade, with significant impacts on technology introduction and industrial upgrading. The selection of these control variables helps to control for various potential influencing factors, thereby more accurately identifying the impact of green digital infrastructure development on energy resilience.

3.3. Model Specification

This study employs the difference-in-differences (DID) method to examine the impact of green digital infrastructure development on energy resilience. We employ the difference-in-differences (DID) method as our primary empirical strategy for several compelling reasons. First, DID enables us to establish causal inference by comparing changes in energy resilience between regions that implemented the green data center pilot policy (treatment group) and those that did not (control group), both before and after policy implementation. This approach effectively controls for time-invariant unobservable regional characteristics and common time trends that might confound the relationship. Second, the quasi-random assignment of pilot regions provides an ideal setting for applying DID methodology as it helps to mitigate selection bias concerns.

To address potential endogeneity concerns in our DID estimation, we implement several strategies. First, we conduct parallel trend tests to verify that treatment and control groups followed similar trajectories before policy implementation, thereby validating a key assumption of DID. Second, we perform placebo tests using randomly assigned ‘fake’ treatment statuses to ensure that our results are not driven by chance. Third, we control for other contemporaneous policies that might affect energy resilience, such as smart city initiatives and low-carbon city policies, to isolate the impact of green digital infrastructure. Fourth, we employ alternative estimation methods and variable constructions to verify the robustness of our findings. These comprehensive approaches collectively address endogeneity concerns and strengthen the validity of our causal inferences.

The benchmark regression model is specified as follows:

$${ERES}_{it}={\beta }_0+{\beta }_1{GDID}_{it}+\gamma {\mathit{Controls}}_{it}+u_i+{\lambda }_t+{\epsilon }_{i,t}$$

(6)

where ${ERES}_{it}$ represents the energy resilience level of region i in period t; ${GDID}_{it}$ is a policy dummy variable that takes the value of 1 when region i is included in the green data center pilot in period t, and 0 otherwise; ${\mathit{Controls}}_{it}$ represents a series of control variables; $u_i$ represents regional fixed effects, used to control for characteristics of each region that do not change over time; ${\lambda }_t$ represents time fixed effects, used to control for time trends common to all regions; ${\epsilon }_{i,t}$ is a random disturbance term. The core coefficient ${\beta }_1$ in the model measures the average treatment effect of green digital infrastructure development on energy resilience; if ${\beta }_1$ is significantly positive, it indicates that the green data center pilot policy has a positive effect on enhancing energy resilience.

4. Empirical Analysis

4.1. Descriptive Analysis

This study first conducts descriptive statistical analysis of the collected data to preliminarily understand the basic characteristics of the sample data and the correlations between variables.

Table 2. Descriptive statistics.

Variable	N	Mean	S.D.	Min	Median	Max
ERES	330	0.4419	0.0604	0.2844	0.4407	0.6466
PGDP	330	10.7430	0.5960	9.4210	10.7130	12.1510
UR	330	0.5620	0.1230	0.3120	0.5500	0.8960
IS	330	0.4530	0.0770	0.1920	0.4470	0.5930
OPEN	330	0.3360	0.4030	0.0350	0.1960	1.8370
ISO	330	0.4380	0.0840	0.2870	0.4260	0.8290
IC	330	10.4950	1.3890	7.1670	10.5380	13.7440

presents the descriptive statistical results of the main variables, including the number of observations, mean, standard deviation, minimum value, median, and maximum value, providing fundamental support for subsequent empirical analysis. As shown in the table, the average energy resilience of China’s 30 provinces, municipalities, and autonomous regions during the sample period is 0.4419, with a standard deviation of 0.0604, indicating certain differences in energy resilience levels across regions. The minimum value is 0.2844 and the maximum value is 0.6466, showing a large range, which indicates unbalanced energy resilience development across regions in China.

Figure 3 illustrates the overall trend of China’s energy resilience from 2011 to 2021, as well as the evolution of energy resilience in the eastern, central, western, and northeastern regions. Overall, China’s energy resilience has shown a steady upward trend, rising from 0.415 in 2011 to 0.485 in 2021, an increase of 16.9%. From a regional perspective, the eastern region has consistently maintained a leading position in energy resilience, rising from 0.450 in 2011 to 0.514 in 2021, an increase of 14.2%; the central region’s energy resilience rose from 0.396 to 0.476, an increase of 20.2%, which is the largest growth; the western region’s energy resilience rose from 0.391 to 0.459, an increase of 17.4%; and the northeastern region’s energy resilience rose from 0.419 to 0.499, an increase of 19.1%. In terms of regional differences, the gap in energy resilience between the eastern region and other regions reached its maximum in 2014–2015 and has since narrowed, indicating a certain effectiveness of regional coordinated development strategies in the field of energy resilience. Notably, after 2016, the growth rate of energy resilience in various regions accelerated significantly, which may have a certain correlation with the implementation of the green data center pilot policy in 2015, providing preliminary evidence for the subsequent empirical analysis.

Figure 4 further illustrates the changes in energy resilience across provinces. As can be seen from the figure, Guangdong Province has the highest level of energy resilience, with a median of 0.607 and relatively small fluctuations, and eastern coastal provinces such as Jiangsu, Fujian, and Zhejiang also have relatively high levels of energy resilience, while western regions such as Ningxia, Qinghai, and Gansu have relatively low levels of energy resilience. These regional differences are closely related to each region’s level of economic development, industrial structure, resource endowment, and ecological environment conditions. Notably, western regions such as Chongqing, Sichuan, and Shaanxi have relatively high levels of energy resilience, which may be related to these regions’ increased investment in energy infrastructure, optimization of energy structure, and active promotion of digital transformation in recent years. At the same time, some provinces such as Hebei and Shandong have experienced rapid growth in energy resilience, with large differences between maximum and minimum values, indicating significant improvements in energy resilience in these regions during the sample period.

From a temporal evolution perspective, energy resilience levels across regions remained relatively stable from 2011 to 2013, experienced slight fluctuations from 2014 to 2015, showed an accelerated upward trend after 2016, and reached the highest value of the sample period in 2021. This trend is highly consistent with China’s energy transition and digital development process. Especially after the implementation of the green data center pilot policy in 2015, there was a notable improvement in overall energy resilience levels, providing intuitive evidence for this study’s exploration of the impact of green digital infrastructure development on energy resilience. It is worth noting that, in 2020, affected by the COVID-19 pandemic, the growth rate of national energy resilience slowed down but still maintained an upward trend, reflecting the resilience capability of China’s energy system in the face of external shocks. Overall, the results of the descriptive analysis indicate that China’s energy resilience level has shown a steady upward trend, but regional differences still exist, and the development of green digital infrastructure may have a positive impact on energy resilience.

4.2. Impact of Green Digital Infrastructure on Energy Resilience

To test the impact of green digital infrastructure development on energy resilience, this study conducts empirical analysis using a difference-in-differences model, with the benchmark regression results reported in

Table 3. Benchmark regression results.

Variable	(1)	(2)
	ERES	ERES
GDID	0.0137 **	0.0125 **
	(2.39)	(2.18)
PGDP		0.0246 ***
		(3.42)
UR		0.0835 **
		(2.16)
IS		−0.0318 *
		(−1.75)
OPEN		0.0097
		(1.42)
City	Yes	Yes
Year	Yes	Yes
N	330	330
R2	0.642	0.683

Note: t-values in parentheses; *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

. Column (1) only controls for regional fixed effects and time fixed effects, while column (2) adds control variables on this basis. In model (1), the coefficient of green digital infrastructure development (GDID) is 0.0137, significant at the 5% level, indicating that the green data center pilot policy can significantly enhance regional energy resilience levels without considering other control variables. After controlling for relevant factors, the coefficient of GDID in model (2) becomes 0.0125, still significant at the 5% level, further verifying that the positive impact of green digital infrastructure development on energy resilience is statistically robust. Specifically, the green data center pilot policy can increase the regional energy resilience index by approximately 0.0125 units on average, which represents an improvement of about 2.83% compared to the sample mean of 0.4419, having significant economic implications. This result supports the core hypothesis of this study, namely that the development of green digital infrastructure can enhance regional energy systems’ resilience capabilities, promoting the ability of energy systems to maintain stable operation and quickly recover when facing external shocks. These benchmark results provide strong support for Hypothesis 1, confirming that green digital infrastructure development has a positive impact on energy resilience. The significant coefficient of GDID across model specifications demonstrates that this positive relationship persists even after controlling for relevant economic and social factors.

4.3. Robustness Tests

To verify the reliability of the benchmark model results, this section conducts comprehensive robustness tests. First, the application of the difference-in-differences method requires that the treatment group and the control group have similar time trends before policy implementation, i.e., satisfy the parallel trends assumption. Specifically, using the implementation of the green data center pilot policy in 2015 as the baseline point, the related statistical results are shown in Figure 5. It can be found that, before the policy began, there were no obvious trends in the experimental and control groups, satisfying the parallel trends assumption. Furthermore, after policy implementation, a continuous improvement effect of the policy can be clearly observed.

Second, to rule out possible randomness, this study conducted a placebo test. We constructed a “virtual” policy shock by randomly assigning treatment and control groups. Theoretically, such a “virtual policy” under random assignment should not have a significant impact on energy resilience. Specifically, as shown in Figure 6, the real red dashed line is not something that can be obtained by chance through random assignment, further supporting the robustness of our previous conclusions.

Finally, considering that there may be other policy interferences during the sample period, such as smart city construction (SCITY) and low-carbon city policy (LCITY), these policies may simultaneously affect green digital infrastructure development and energy resilience, leading to biased estimation results. To eliminate these interferences, columns (1) and (2) in

Table 4. Robustness test results.

Variable	(1)	(2)	(3)	(4)
	ERES	ERES	ERES	ERES
GDID	0.0121 **	0.0118 **	0.0116 **	0.0129 **
	(2.11)	(2.06)	(2.07)	(2.25)
SCITY	0.0082 *
	(1.78)
LCITY		0.0096 *
		(1.85)
Controls	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
N	330	330	330	330
R2	0.692	0.694	0.675	0.688

*, **, indicate significance at the 10%, 5%, respectively.

report the results after excluding other policy interferences. The results show that, after controlling for SCITY and LCITY, the coefficients of GDID are 0.0121 and 0.0118, respectively, both significant at the 5% level, and the coefficient magnitudes are close to the benchmark model, indicating that the results of the benchmark model are not affected by other policy interferences and have good robustness.

In addition to the above tests, we also conducted a series of other robustness tests. First, considering that the subjectivity in determining weights during the construction of energy resilience indicators may affect the results, column (3) of Table 4 shows that we recalculated energy resilience indicators using the entropy method, with the coefficient of GDID being 0.0116, still significant at the 5% level. Additionally, we tried different estimation methods, such as using the Tobit model in column (4) of Table 4 to consider the interval restriction of the dependent variable, with the coefficient of GDID being 0.0129, significant at the 5% level, indicating that the conclusion of the positive impact of green digital infrastructure development on energy resilience has good robustness. Combining the results of various robustness tests, the conclusion that green digital infrastructure development has a positive impact on energy resilience drawn by this study has high credibility.

4.4. Heterogeneity Analysis

Does the positive impact of green digital infrastructure development on energy resilience differ across cities with different levels of economic development and digital foundation? The results are shown in

Table 5. Heterogeneity analysis results.

Variable	(1)	(2)	(3)	(4)
	High Economic Development	Low Economic Development	High Digital Foundation	Low Digital Foundation
	ERES	ERES	ERES	ERES
GDID	0.0183 ***	0.0092 *	0.0164 **	0.0085
	(2.72)	(1.68)	(2.35)	(1.44)
Controls	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
R2	0.709	0.652	0.712	0.645
Inter-group Difference	0.0091 **		0.0079 *

*, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

. First, the sample is divided into high-economic-development-level and low-economic-development-level groups based on the median of per capita GDP in each region in 2015. The results in columns (1) and (2) of Table 5 show that, in the high-economic-development-level group, the coefficient of GDID is 0.0183, significant at the 1% level, while, in the low-economic-development-level group, the coefficient is 0.0092, significant at the 10% level. The difference test between the two groups shows that the difference is significant at the 5% level, indicating that the enhancement effect of green digital infrastructure development on energy resilience is more pronounced in regions with higher levels of economic development. This may be because regions with higher economic development levels have stronger financial capabilities and technical abilities, allowing them to better leverage the advantages of green digital infrastructure, achieve intelligent and green transformation of energy systems, thereby enhancing energy resilience.

Finally, the sample is divided into high-digital-foundation and low-digital-foundation groups based on the median of internet penetration rate in each region in 2015. The results in columns (3) and (4) of Table 5 show that, in the high-digital-foundation group, the coefficient of GDID is 0.0164, significant at the 5% level, while, in the low-digital-foundation group, the coefficient is 0.0085, statistically insignificant. The difference test between the two groups shows that the difference is significant at the 10% level, indicating that the enhancement effect of green digital infrastructure development on energy resilience is more pronounced in regions with better digital foundations. This finding is consistent with intuition: regions with better digital foundations have more complete information infrastructure and higher levels of digital technology application, allowing them to better support the construction and operation of green data centers, achieve the refined management and intelligent scheduling of energy systems, thereby more effectively enhancing energy resilience. The heterogeneity analysis results offer important nuance to our understanding of Hypothesis 1, revealing that, while green digital infrastructure generally enhances energy resilience, this effect is contingent on regional characteristics. The positive impact is significantly stronger in regions with higher economic development levels and better digital foundations, highlighting the importance of considering regional context when implementing green digital infrastructure policies.

4.5. Mechanism Exploration

4.5.1. Industrial Structure Optimization Mechanism

After verifying the positive impact of green digital infrastructure development on energy resilience, this study further explores the mechanisms through which green digital infrastructure development affects energy resilience, focusing on two potential transmission pathways: industrial structure optimization and innovation capacity, with the results shown in

Table 6. Mechanism analysis results.

Variable	(1)	(2)	(3)	(4)
	ISO	ERES	IC	ERES
GDID	0.0193 **	0.0112 *	0.0980 ***	0.0110 *
	(2.22)	(1.95)	(2.76)	(1.92)
ISO		0.0675 ***
		(2.82)
IC				0.0153 **
				(2.15)
Controls	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
N	330	330	330	330
R2	0.658	0.697	0.675	0.692

*, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

. First, theoretically, the development of green digital infrastructure may enhance energy resilience by promoting industrial structure transition toward service-oriented and low-carbon directions, reducing energy consumption intensity, and improving energy utilization efficiency. To verify this mechanism, this study uses industrial structure optimization (ISO) as a key transmission variable and constructs a corresponding effect model. Column (1) of Table 6 shows that the impact coefficient of green digital infrastructure development (GDID) on industrial structure optimization is 0.0193, significant at the 5% level, indicating that the green data center pilot policy significantly promoted the increase in the proportion of the tertiary industry; column (2) shows that, after controlling for GDID, the impact coefficient of industrial structure optimization on energy resilience is 0.0675, significant at the 1% level, indicating that industrial structure tilting toward the service sector is conducive to enhancing energy resilience; meanwhile, the coefficient of GDID decreased from 0.0125 in the benchmark model to 0.0112, but is still significant at the 10% level, indicating that industrial structure optimization is an important transmission pathway through which green digital infrastructure development affects energy resilience. These findings provide strong support for Hypothesis 2, confirming that, in the Chinese context, industrial structure optimization serves as a significant pathway through which green digital infrastructure enhances energy resilience. The evidence clearly demonstrates that the green data center pilot policy effectively promotes a transition toward service-oriented industries, which, in turn, contributes to improved energy system stability and adaptability.

4.5.2. Innovation Capacity Enhancement Mechanism

Second, the development of green digital infrastructure may enhance energy resilience by promoting technological innovation, driving energy technology progress and management level improvement, and increasing the adaptability and recovery capability of energy systems. To verify this mechanism, this study uses innovation capacity (IC) as another key transmission variable and constructs a corresponding effect model. Column (3) of Table 6 shows that the impact coefficient of green digital infrastructure development on innovation capacity is 0.0980, significant at the 1% level, indicating that the green data center pilot policy significantly enhanced regional innovation capacity; column (4) shows that, after controlling for GDID, the impact coefficient of innovation capacity on energy resilience is 0.0153, significant at the 5% level, indicating that enhanced innovation capacity is conducive to strengthening energy resilience; meanwhile, the coefficient of GDID decreased from 0.0125 in the benchmark model to 0.0110, but is still significant at the 10% level, indicating that enhanced innovation capacity constitutes another important transmission pathway through which green digital infrastructure development affects energy resilience. These results strongly support Hypothesis 3, demonstrating that, in the Chinese context, innovation capacity enhancement is another key mechanism linking green digital infrastructure to energy resilience improvement. The statistically significant relationships between green digital infrastructure, innovation outcomes, and energy resilience metrics confirm our theoretical proposition about the innovation-driven pathway to more resilient energy systems.

5. Discussion

Our findings regarding the positive impact of green digital infrastructure on energy resilience both complement and extend several strands of the existing literature. The 2.83% improvement in energy resilience observed in our study represents a meaningful enhancement to energy system stability and adaptability, particularly given the typically gradual evolution of energy infrastructure. This improvement demonstrates that digital transformation can produce quantifiable benefits for energy resilience beyond theoretical assumptions.

Our research contributes methodologically to the growing literature employing quasi-experimental designs to evaluate environmental policies. Mbanyele et al. (2024) exploited a green lending mandate as a quasi-natural experiment to examine its effects on the labor investment inefficiency of firms with high carbon risk. They found that heightened climate regulatory risk through mandatory green lending requirements motivates firms with higher carbon risk to adjust their labor investments to levels supported by economic fundamentals, particularly by curbing overinvestment in labor [38]. Their findings complement our work by showing how green policies create behavioral changes in economic actors, though our focus is on infrastructure rather than financial mechanisms.

Similarly, Huang et al. (2025) investigated the influence of green finance initiatives on carbon-related marginal abatement costs using two competing hypotheses: regulatory versus technical effects. Using a synthetic DID approach, they found robust evidence that green finance initiatives significantly increase the marginal abatement cost of CO₂ emissions by about CNY 310 per ton in affected pilots compared to their counterparts [39]. While our study focuses on benefits rather than costs, their methodological approach using pilot programs as a quasi-natural design mirrors our strategy, though they employed synthetic DID to address potential non-parallel trends and small treatment group challenges.

In the context of green industrial policies, Liu et al. (2025) examined the link between green industrial policy and corporate green innovation using the staggered adoption of the green factory identification (GFI) in China as a plausibly exogenous shock. Their staggered difference-in-difference analysis demonstrated a significant positive association between the GFI and green innovation, with mechanisms including alleviating financing constraints and fostering external supervision [40]. These mechanisms parallel our findings on innovation capacity enhancement, though applied at the firm level rather than regional level.

The heterogeneous effects that we observed in green digital infrastructure’s impact are consistent with Liu et al.’s (2024) findings that green bond issuance enhances environmental responsibility engagements, with effects more pronounced among firms in low-polluting industries, those without environmental subsidies, and those with higher managerial abilities [41]. Both studies demonstrate that green interventions do not produce uniform outcomes but are contingent on contextual factors—regional characteristics in our case and firm characteristics in theirs.

Our research on energy resilience extends previous work by Li and Zhang (2023), who documented regional disparities in energy resilience across Chinese provinces but did not empirically connect these differences to digital infrastructure development [24]. Similarly, Sharmin and Dhakal (2022) developed a composite energy resilience index for Bangladesh with 15 indicators [25], whereas our comprehensive four-dimensional framework with 18 indicators provides a more nuanced measurement specifically designed to capture the multifaceted nature of energy resilience in diverse regional contexts.

While these studies collectively offer valuable insights into various aspects of green policy impacts, our work uniquely addresses the physical and digital infrastructure aspects of the green transition. Where most studies have focused on financial instruments, corporate behavior, or specific environmental metrics, our research bridges the gap between digital infrastructure development and comprehensive energy resilience—a connection that has received limited empirical attention despite its growing policy relevance in an increasingly digitalized and climate-vulnerable world.

6. Conclusions and Implications

6.1. Summary of Key Findings

Based on panel data from 30 provincial-level regions in China from 2011 to 2021, this study empirically examines the impact of green digital infrastructure development on energy resilience using the difference-in-differences method. The results show that the development of green digital infrastructure significantly enhanced regional energy resilience levels, with an average improvement of approximately 2.83%. This conclusion remains robust after controlling for other policy interferences, changing energy resilience measurement methods, and adopting different estimation models. Heterogeneity analysis further finds that the enhancement effect of green digital infrastructure development on energy resilience is more significant in regions with higher levels of economic development and better digital foundations. Mechanism analysis reveals two important transmission pathways: industrial structure optimization and innovation capacity enhancement.

6.2. Theoretical Implications

Our findings contribute to theoretical understanding in several ways. First, they extend complex adaptive systems theory by demonstrating how digital infrastructure enhances system adaptability and self-organization within energy networks. The observed positive relationship between green digital infrastructure and energy resilience supports the theoretical proposition that improved information processing capabilities strengthen a system’s ability to maintain stability under external disturbances. Second, our identification of industrial structure optimization as a mechanism enriches industrial ecology theory by showing how digital technologies facilitate cross-sector integration and resource optimization, creating more resilient industrial ecosystems. Third, our findings on innovation capacity enhancement support and extend innovation diffusion theory by illustrating how digital infrastructure accelerates both the development and deployment of energy innovations. Finally, our heterogeneity analysis contributes to the growing theoretical discourse on digital divide effects, showing that technological benefits are not uniformly distributed but rather contingent on existing development levels.

6.3. Practical Implications

Our research offers several practical implications for policymakers and stakeholders. First, green digital infrastructure should be systematically incorporated into national energy security strategies, with increased investment prioritized for regions with lower energy resilience. Policymakers should establish comprehensive evaluation standards and regulatory mechanisms for green digital infrastructure, guiding the development of energy-saving, environmentally friendly data centers and digital systems.

Second, regional differences should be carefully considered when promoting green digital infrastructure. For economically advanced regions with strong digital foundations, emphasis should be placed on innovative demonstrations and exploring new models of energy information integration. For less developed regions, priority should be given to basic digital capacity building and improving accessibility to digital technologies.

Third, policy instruments should be designed to leverage the identified mechanisms of industrial structure optimization and innovation enhancement. This includes supporting the digital transformation of traditional industries, fostering green technology innovation through industry–university–research cooperation, and building comprehensive innovation ecosystems that combine industry chains, innovation chains, and capital chains.

6.4. Limitations and Future Research Directions

Despite its contributions, this study has several limitations that point to future research opportunities. First, our analysis focuses on province-level data, which may mask important variations at more granular levels, particularly missing micro-level insights from individual firms that could provide a deeper understanding of implementation processes. Second, our study period ends in 2021, limiting our ability to observe longer-term effects. Third, while we identify industrial structure optimization and innovation capacity as key mechanisms, other potential pathways warrant exploration, including consumer behavior changes, energy market reforms, and governance improvements. Fourth, our focus on China limits the generalizability of findings, and our methodology of aggregating city-level policy implementation to provincial analysis presents certain analytical constraints. Finally, future research should investigate potential negative consequences or rebound effects of green digital infrastructure development, such as increased energy consumption from growing digital services or cybersecurity vulnerabilities in increasingly connected energy systems. A more comprehensive understanding of both benefits and risks would contribute to more balanced policy approaches in this rapidly evolving field.