Comparative Analysis of Spatial Distribution and Mechanism Differences Between Public Electric Vehicle Charging Stations and Traditional Gas Stations: A Case Study from Wenzhou, China

Abstract

With the impact of fossil energy on the climate environment and the development of energy technologies, new energy vehicles, represented by electric cars, have begun to receive increasing attention and emphasis. The rapid proliferation of public charging infrastructure for NEVs has concurrently influenced traditional petrol station networks, creating measurable disparities in their spatial distributions that warrant systematic investigation. This research examines Wenzhou City, China, as a representative case area, employing multi-source Point of Interest (POI) data and spatial analysis models to analyse differential characteristics in spatial layout accessibility, service equity, and underlying driving mechanisms between public electric vehicle charging stations (EV) and traditional gas stations (GS). The findings reveal that public electric vehicle charging stations exhibit a pronounced “single-centre concentration with weak multi-centre linkage” spatial configuration, heavily reliant on dual-core drivers of population density and economic activity. This results in marked service accessibility declines in peripheral areas, resembling a cliff-like drop, and a relatively low spatial equity index. In contrast, traditional gas stations demonstrate a “core-axis linkage” diffusion pattern with strong coupling to urban road networks, showing gradient attenuation in service coverage efficiency along transportation arteries, fewer suburban service gaps, and more gradual accessibility reductions. Location entropy analysis further indicates that charging station deployment shows significant capital-oriented tendencies, with certain areas exhibiting paradoxical “excess facilities” phenomena, while gas station distribution aligns more closely with road network topology and transportation demand dynamics. Furthermore, the layout characteristics of public charging stations feature a more complex and diverse range of land use types, while traditional gas stations have a strong dependence on industrial land. This research elucidates the spatial distribution patterns of emerging and legacy energy infrastructure in the survey regions, providing critical empirical evidence for optimising energy infrastructure allocation and facilitating coordinated transportation system transitions. The findings also offer practical insights for the construction of energy supply facilities in urban development frameworks, holding substantial reference value for achieving sustainable urban spatial governance.

1. Introduction

With the pervasive integration of automobiles in modern industrial societies, carbon dioxide emissions from conventional fossil fuel-powered vehicles have significantly contributed to atmospheric greenhouse gas concentrations, exacerbating global climate change [1]. Against this backdrop, numerous nations have established ambitious targets to achieve net-zero greenhouse gas emissions by 2050, with China being a prominent participant in this global endeavour. A key strategy for realising this objective involves accelerating the adoption of new energy vehicles (NEVs). Emerging during the mid-to-late 20th century, spurred by advancements in battery technology and heightened environmental consciousness following pollution crises and oil shortages [2], electric vehicles (EVs) have garnered substantial attention as a sustainable transportation solution [3]. These zero-emission vehicles (ZEVs) offer a viable pathway to mitigate climate change while reducing direct reliance on petroleum-based fuels [4].

Recent years have witnessed a marked surge in EV adoption. According to data from China’s Ministry of Public Security and the EVCIPA Open Service Platform, the national NEV stock reached 36.89 million units by June 2025, representing 10.27% of all registered vehicles. The first half of 2025 saw a record-breaking 5.622 million new NEV registrations, marking a 27.86% year-on-year increase. Notably, NEVs accounted for an unprecedented 44.97% of all new vehicle registrations during this period. Pure battery electric vehicles (BEVs) constituted the majority at 25.539 million units (69.23% of the NEV total), which represents an overwhelming majority (Data sources: EVCIPA Open Service Platform, https://evcipa.com/index, accessed on 1 July 2025; The Ministry of Public Security of the People’s Republic of ChinaChina, https://www.122.gov.cn/m/index/, accessed on 1 July 2025).

From the perspective of consumer acceptance, the driving range of electric vehicles (EVs) directly influences purchasing decisions. Consequently, extending operational range remains a critical imperative for accelerating EV adoption rates. While advancements in energy conversion efficiency and battery storage technology constitute essential technical developments, the strategic deployment of charging infrastructure (particularly public charging stations) represents an equally pivotal factor in facilitating EV proliferation [5]. Empirical evidence suggests that the accessibility of charging networks serves as a fundamental prerequisite for mainstream EV acceptance [6,7]. Studies have consistently demonstrated that expanding public charging station coverage effectively mitigates consumer range anxiety, thereby significantly enhancing purchase propensity [8]. This infrastructure development creates a virtuous cycle: as charging accessibility improves, potential buyers perceive EVs as more practical transportation solutions, driving increased market penetration which in turn justifies further infrastructure investment.

Current research on public charging infrastructure predominantly concentrates on the spatial configuration of these facilities. Numerous studies have investigated optimal siting strategies to maximise coverage while establishing appropriate inter-station distances, developing multifocal optimisation models [9,10]. These investigations prioritise charging network layout enhancement, with particular consideration given to electric vehicle (EV) operational range limitations. The decision-making framework based on the GIS-FAHP-MABAC location selection method for facilities’ location and optimisation is a current new direction. This method model attempts to strike a balance between quantitative indicators and subjective weights, reflecting a multi-source and composite perspective [11,12]. Spatial equity analysis from the public charging infrastructure perspective provides critical insights for assessing urban service delivery effectiveness [13]. Commonly employed spatial equity metrics include accessibility indices, Gini coefficients, Lorenz curves [14], and local spatial autocorrelation measures [15]. Among them, accessibility is typically used to assess the fairness and convenience of facility accessibility in a space, especially to describe a transportation trend [16] and spatio-temporal relationship [17], and to some extent, it can reflect the tendency of facility layout [18]. The two-step floating catchment area (2SFCA) method and its derivatives represent established techniques for quantifying spatial accessibility, finding widespread application in public service facility research including urban green spaces [19,20] and healthcare infrastructure [21,22]. Global and local Moran’s I index analyses constitute another essential toolset for examining spatial clustering patterns in urban public facilities [23], demonstrating particular efficacy in identifying facility aggregation types.

In the realm of urban development and planning, the spatial configuration of conventional fueling infrastructure has evolved into a distinctive urban feature [24]. The siting logic for petrol stations incorporates multiple determinants including transportation network connectivity, traffic volume patterns, land acquisition costs, and service radius parameters. These factors offer valuable referential insights for developing public charging infrastructure spatial strategies. However, contemporary research predominantly treats electric vehicle (EV) charging stations and traditional fueling facilities as discrete entities, lacking systematic comparative analysis between these two critical infrastructure types. While limited comparative studies exist, these primarily focus on facility usage behaviour [25,26] or examine mechanisms for hybridising/retrofitting existing gas stations with charging capabilities [27,28]. The spatial layer attributes and underlying distributional mechanisms of these infrastructure systems have received comparatively less scholarly attention. This study addresses this research gap by adopting a unified analytical framework to compare the spatial configuration characteristics of traditional gas stations and public charging facilities. Taking the Wenzhou metropolitan region (Zhejiang Province, China) as an empirical case study, we employ geographic information systems (GIS) and spatial analysis methodologies to: (1) Quantify differences in spatial distribution patterns. (2) Reveal divergent siting mechanisms through multivariable regression modelling. (3) Assess spatial equity implications for urban infrastructure planning. The findings provide critical insights for optimising short-term and long-term public charging infrastructure deployment strategies, offering practical guidance for urban planners seeking to balance energy transition objectives with equitable service provision.

2. Generalisation of Survey Regions

Wenzhou, located at the southernmost tip of Zhejiang Province in China (27°03′ N–28°36′ N, 119°37′ E–121°18′ E) (Data sources: https://www.wenzhou.gov.cn/art/2025/7/23/art_1229240253_59267972.html, accessed on 1 August 2025), occupies a trapezoidal topography sloping from southwest to northeast, characterised by predominantly mountainous and hilly terrain intersected by a serpentine coastline that winds along its eastern edge (Figure 1). The region’s geographical profile combines dramatic elevation changes with maritime influences, creating a distinctive landscape where rugged uplands gradually descend toward the East China Sea while coastal plains and estuaries punctuate the shoreline. As of Q1 2025, the city’s permanent population stood at 9.852 million according to the Wenzhou Automobile Distribution Industry Association (Data sources: https://news.66wz.com/system/2025/04/26/105682253.shtml, accessed on 1 August 2025), with 451,000 registered new energy vehicles (NEVs)—equating to 46 NEVs per 1000 inhabitants, more than double the national average of 22 ownership per thousand people. During the same period, NEV registrations reached 31,924 units, capturing 57.4% of the automotive market share. Despite its strong economic position, Wenzhou remains neither a provincial capital nor a planned municipality, characterised by underdeveloped rail transit and heavy reliance on road transportation for urban mobility. This coastal manufacturing hub’s well-established road network, combined with its exceptionally high NEV adoption rates and pronounced dependence on surface transportation infrastructure, renders it an ideal case study for comparing spatial distribution patterns between traditional fueling stations and emerging charging facilities. The region’s complex topography and road-centric transportation ecosystem provide a unique analytical framework for examining infrastructure equity and siting mechanisms in manufacturing-oriented urban contexts [29,30].

According to local government planning documents (https://wzfgw.wenzhou.gov.cn/art/2023/8/30/art_1229737564_58917908.html, accessed on 1 July 2025), the Wenzhou metropolitan area is projected to have 538 public charging stations with 1486 charging points in operation by 2025. These infrastructure targets form part of a broader strategy to incentivise enterprise-led expansion of new energy vehicle (NEV) charging networks, aiming to enhance facility coverage across urban regions. Statistical data reveals Wenzhou’s NEV adoption rate reached 62.2% in Q1 2025, ranking tenth nationally—a testament to the city’s advanced energy transition progress. This dual context of ambitious charging infrastructure development and exceptional vehicle electrification rates positions Wenzhou as a leading indicator for China’s transportation decarbonisation efforts. The region’s proactive policy framework and demonstrated market responsiveness provide compelling justification for its selection as a research site, offering valuable insights into the interplay between infrastructure planning and technological adoption in manufacturing-oriented urban economies. The forward-looking significance of this case study is further reinforced by Wenzhou’s status as a non-provincial capital city achieving provincial-level performance metrics, making it an ideal model for examining scalable solutions applicable to secondary cities undergoing rapid energy system transformation.

3. Research Methods and Data Processing

3.1. Methods Framework and Approach

This study employs a comparative analytical framework to evaluate spatial equity and accessibility between traditional gas stations and public charging infrastructure, comprising two interrelated components. Firstly, spatial distribution characteristics are examined through density mapping and service coverage analysis, revealing disparities in facility siting patterns and service capacity. Secondly, quantitative assessment utilises the two-step floating catchment area (2SFCA) model to calculate accessibility indices, complemented by Moran’s I statistics for spatial autocorrelation analysis. These methodologies enable systematic evaluation of infrastructure equity by quantifying spatial accessibility patterns and identifying clustering tendencies. The integrated approach, depicted in Figure 2, facilitates holistic comparison of service distribution mechanisms while accounting for urban morphological constraints. By combining qualitative spatial pattern recognition with quantitative accessibility metrics, the research framework provides a robust platform for assessing infrastructure equity in transitioning urban energy systems.

3.2. Data Source

The dataset for this research was primarily sourced from Points of Interest (POI) application programming interfaces (APIs) provided by Amap (https://lbs.amap.com/, accessed on 30 April 2025) and Baidu Maps (https://lbsyun.baidu.com/, accessed on 30 April 2025), supplemented by web scraping methodologies to extract geospatial data for public charging stations and traditional gas stations. All data was collected as of April 2025, with attribute fields including facility names, categorical identifiers, addresses, and WGS84 coordinate systems. Following data cleansing and standardisation procedures, the final dataset comprised 1373 public charging stations and 275 traditional gas stations. Auxiliary datasets encompassing road networks and administrative boundaries were also incorporated to support spatial context analysis. POI data, recognised as a valuable form of geospatial big data, offers high positional accuracy, extensive attribute granularity, and temporal relevance, making it particularly suitable for examining infrastructure spatial distribution patterns. These datasets form the foundational resource for subsequent spatial equity and accessibility analyses in accordance with established geospatial methodologies [31].

This study employs a multi-source data integration approach to evaluate infrastructure service equity. Geospatial datasets were primarily sourced from Points of Interest (POI) application programming interfaces (APIs) provided by Amap (https://lbs.amap.com/, accessed on 30 April 2025) and Baidu Maps (https://lbsyun.baidu.com/, accessed on 30 April 2025), supplemented by web scraping techniques to collect facility-specific attributes including names, categories, addresses, and WGS84 coordinates. Following data cleansing and standardisation procedures, the final dataset comprised 1373 public charging stations and 275 traditional gas stations as of April 2025. Demographic data was sourced from China’s Seventh National Population Census (https://www.stats.gov.cn/, accessed on 30 April 2025), with spatial analysis conducted at the sub-district level (188 streets, townships, and towns within Wenzhou’s administrative boundaries) to align with census zoning units and China’s administrative management framework. Auxiliary datasets including road networks and administrative boundaries were obtained from the National Geographic Information Public Service Platform (https://www.tianditu.gov.cn/, accessed on 30 April 2025). Furthermore, internal planning documents and urban functional zone distributions from Wenzhou’s municipal authorities were incorporated to contextualise findings within existing policy frameworks, enhancing the practical relevance of recommendations for infrastructure deployment and regional development strategies. This integrated dataset provides a robust foundation for assessing spatial equity through comprehensive attribute granularity and administrative relevance. The data source comparison is shown in

Table 1. Data source comparison.

Data Type	Data Purpose	Data Source
POI data	Provide facility space positioning and other related information	Gaode Map Data Service Interface (https://lbs.amap.com/, accessed on 30 April 2025) Baidu Map Data Service Interface (https://lbsyun.baidu.com/, accessed on 30 April 2025)
Topographic elevation data, urban built-up area land type data	Help explain the current situation and layout characteristics of the research area	Geospatial Data Cloud Platform (https://www.gscloud.cn/, accessed on 30 April 2025) Zenodo Data Sharing Platform (https://zenodo.org/records/16794007, accessed on 22 August 2025)
Road data	Analysis of current road traffic conditions	Gaode Map Data Service Interface (https://lbs.amap.com/, accessed on 30 April 2025)
Administrative division data	Division of the research area	China National Geospatial Information Public Service Platform (https://www.tianditu.gov.cn/, accessed on 30 April 2025)
Population data	Population and facility matching relationship	China “Seventh National Population Census Bulletin” (https://www.stats.gov.cn/sj/pcsj/rkpc/7rp/indexch.htm, accessed on 30 April 2025)

.

3.3. Research Methods

3.3.1. Standard Deviation Ellipse Method

The standard deviation ellipse (SDE) method quantifies the spatial dispersion and directional trends of geographic features by constructing an ellipse derived from the mean centre of point distributions. This approach calculates the standard deviation of x-coordinates and y-coordinates relative to the mean centre to define the ellipse’s major and minor axes, thereby capturing the primary orientation and spread of spatial patterns. In comparative analyses, the relative positioning of ellipse centres and their axial directionalities provide critical insights into spatial relationships between multiple elements, enabling the identification of clustering tendencies, alignment with geographical features, or associations with specific socioeconomic factors. The methodology’s statistical rigour and visual interpretability make it particularly valuable for assessing infrastructure distribution equity, as demonstrated in Equation (1), where the ellipse parameters are mathematically defined to reflect both central tendency and spatial variance.

$$C=\left(\begin{array}{cc}\mathrm{var}\left(x\right)& \mathrm{cov}\left(x,y\right)\\ \mathrm{cov}\left(y,x\right)& \mathrm{var}\left(y\right)\end{array}\right)={\displaystyle \frac{1}{n}}\left(\begin{array}{cc}{\displaystyle \sum _{i=1}^n{\tilde{x}}_i^2}& {\displaystyle \sum _{i=1}^n{\tilde{x}}_i{\tilde{y}}_i}\\ {\displaystyle \sum _{i=1}^n{\tilde{x}}_i{\tilde{y}}_i}& {\displaystyle \sum _{i=1}^n{\tilde{y}}_i^2}\end{array}\right),$$

(1)

where x and y represent the coordinates of element i, $\left(\tilde{x},\tilde{y}\right)$ represents the average centre of the element, and n is the total number of elements.

3.3.2. Kernel Density Analysis

Kernel density analysis is used to calculate the density of point features surrounding each output raster pixel. It reflects the relative concentration degree of point features in the geographical unit distribution [30]. Data points closer to the central point feature will be assigned a higher weight, while those farther away will be assigned a lower weight. The estimated density of each point is the weighted average density of any point within that range [32]. The calculation method is shown in Equation (2). Through kernel density analysis, the aggregation status of POI points in space can be determined. This analysis method is widely used for the identification of hotspots in point patterns [33].

$$\widehat{f}\left(x\right)={\displaystyle \frac{1}{nh}}{\displaystyle \sum _{i=1}^{n}K\left(\frac{x-X_i}{h}\right)},$$

(2)

In the formula, $\widehat{f}\left(x\right)$ represents the kernel density estimation value at point x; n is the total number of calculation points; h is the bandwidth (smoothing parameter), and K(x) is used to control the range of influence of the kernel function.

3.3.3. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis serves as a critical method for investigating the aggregation or dispersion patterns of spatial data. In this study, Moran’s I index was employed to quantify the spatial autocorrelation of facility distributions within the survey regions, encompassing both global and local Moran’s I metrics. The global Moran’s I statistic provides an overall assessment of spatial clustering trends, enabling determination of whether facilities exhibit significant aggregation, dispersion, or random distribution across the landscape [34]. For this analysis, streets and towns were designated as the primary analytical units, with the number of gas stations and charging stations quantified within each unit. A spatial adjacency matrix was constructed to define neighbourhood relationships between units, and calculations were performed using ArcGIS software. Moran’s I values range between −1 and 1, where positive values (0 to 1) indicate clustering of similar attributes, negative values (−1 to 0) suggest dispersion, and values near zero imply spatial randomness. The specific calculation methodology aligns with established formulae, such as that presented in Equation (3), which incorporates spatial weights and attribute deviations to generate the autocorrelation metric. This approach facilitates rigorous evaluation of facility distribution patterns and their spatial dependencies at both regional and local scales.

$$I={\displaystyle \frac{n{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}}}{S_0{\displaystyle {\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2}}},$$

(3)

In the formula, n represents the number of analysis units; x_i is the observed value of the i-th unit; $\overline{x}$ is the overall mean; w_ij is the spatial weight (adjacency relationship) between unit i and unit j; S₀ is the total sum of the spatial weights.

The Local Moran’s I statistic, often referred to as LISA (Local Indicators of Spatial Association), is a spatial analytical tool designed to detect and visualise localised clustering patterns within geographic datasets [35]. Unlike its global counterpart, which provides an aggregate measure of spatial autocorrelation across the entire survey regions, LISA operates at the individual spatial unit level, enabling the identification of statistically significant hotspots, coldspots, and spatial outliers. The calculation, as outlined in Equation (4).

$$I_i=\left(x_i-\overline{x}\right){\displaystyle {\sum }_{j=1}^n{w}_{ij}\left(x_j-\overline{x}\right)},$$

(4)

In the formula, n represents the number of analysis units; x_i is the observed value of the i-th unit; $\overline{x}$ is the overall mean; w_ij is the spatial weight (adjacency relationship) between unit i and unit j. Incorporates spatial weights matrices to assess the similarity between a given unit and its neighbours, thereby quantifying the degree of localised association. This method categorises spatial relationships into four distinct types: High-High (HH), where units with high values are surrounded by similarly high-valued neighbours; Low-Low (LL), indicating low-value units adjacent to other low-value units; Low-High (LH) and High-Low (HL), which represent spatial discordance where units with contrasting values are adjacent. By mapping these clusters, researchers can discern diffusion dynamics, such as the spread or concentration of facilities, and evaluate how regional infrastructure sharing trends manifest at sub-regional scales. This granular insight complements global autocorrelation analysis, offering a nuanced understanding of spatial heterogeneity and informing targeted policy interventions. Based on this, the diffusion trend of the sharing of regional facilities is analysed and studied, as shown in

Table 2. Characteristics and meanings of different clustering types.

Clustering Types	Characteristics	Meanings
HH (High-High)	The xi value is very high, and the adjacent xj value is also very high	High-value aggregation areas
LL (Low-Low)	The xi value is very low, and the adjacent xj value is also very low	Low-value aggregation areas
LH (Low-High)	The xi value is very low, while the adjacent xj value is also very high	Low-value “islands” of spatial anomalies
HL (High-Low)	The xi value is very high, while the adjacent xj value is also very low	High-value “islands” of spatial anomalies

.

3.3.4. Location Quotient Analysis

The Location Quotient (LQ) serves as a critical metric for evaluating the alignment between per capita facility provision within administrative regions and the city-wide average, thereby quantifying the spatial equilibrium of public service distribution relative to population needs. Mathematically expressed in Equation (5), LQ calculates the ratio of a region’s facility-to-population density to the corresponding city-level benchmark. A value of LQ = 1 signifies regional parity with the city-wide average in terms of facility-to-population ratios, indicating balanced provision. When LQ < 1, it denotes relative deficiency, where per capita facility availability in the region falls below the urban average, suggesting potential undersupply. Conversely, LQ > 1 reflects a surplus, with facilities exceeding the proportional needs of the local population. This analytical framework is particularly valuable for assessing how effectively public infrastructure layouts correspond to demographic distributions, enabling the identification of spatial mismatches or clustering phenomena. By highlighting areas of over- or under-provision, regional entropy analysis provides actionable insights for equitable resource allocation and policy adjustment, as referenced in prior studies [36].

$$L{Q}_i={\displaystyle \frac{\left(f_i/p_i\right)}{\left(F/P\right)}},$$

(5)

where f_i represents the number of facilities in the i-th area; p_i represents the population of the i-th area; F represents the total number of facilities within the survey regions; P represents the total population within the survey regions.

3.3.5. Gini Coefficient Analysis

The Gini coefficient quantifies distributional equity by measuring the concentration of relative demand and supply dynamics within service units, offering a standardised metric to evaluate spatial justice in public service provision. When applied to transportation infrastructure assessment, this index focuses on analysing the numerical disparity between population needs and facility availability across geographic units [37,38]. Calculated through Equation (6), the coefficient ranges from 0 to 1, where values approaching 0 indicate near-perfect equity in resource distribution, reflecting optimal alignment between service capacity and demographic requirements. Conversely, higher coefficients reveal greater imbalance, with clustered resources or unmet demand in specific areas. This methodology enables policymakers to assess how effectively total facility service capacity harmonises with population distribution patterns within the survey regions, identifying systemic inequities and guiding targeted interventions to enhance spatial accessibility. By translating complex supply-demand relationships into a single interpretable value, the Gini coefficient provides a robust framework for evaluating fairness in public service allocation and informing evidence-based urban planning strategies.

$$G=1-{\displaystyle \sum _{i=1}^n\left(X_i-X_{i-1}\right)\left(Y_i+Y_{i-1}\right)},$$

(6)

In the formula, X_i represents the cumulative population proportion after sorting of the population; Y_i represents the corresponding cumulative facility service proportion.

3.3.6. Gaussian-Based 2-Step Floating Catchment Area (Ga2SFCA) Method

The Two-Step Floating Catchment Area (2SFCA) method, initially proposed by Radke et al. (2000) [39] and subsequently refined by Luo et al. (2003) [40], represents a spatial accessibility analysis technique that evaluates service availability by integrating supply and demand dynamics. Rooted in the Two-Step Mobile Search Method framework, this approach employs a Gaussian function as the distance-decay function to model the gradual reduction in service utilisation probability with increasing distance, thereby offering a smoother and more realistic representation of spatial accessibility compared to traditional stepwise models. The methodology operates through two sequential stages: first step, for each supply point (e.g., charging stations or gas stations), it identifies all demand points (e.g., residential locations) within a predefined search radius d₀, calculating a supply-demand ratio R_j that reflects the capacity of each supply unit relative to its surrounding population. Second step, for each demand point, it aggregates the weighted R_j values from all supply points within the same d₀ threshold, yielding an accessibility score A_i that quantifies the cumulative opportunity for residents to access services. By combining geographic proximity with capacity-adjusted resource distribution, the Ga2SFCA method provides a nuanced assessment of spatial equity, enabling policymakers to identify underserved areas and optimise infrastructure planning based on empirically derived accessibility metrics [41,42].

$$A_i={\displaystyle {\sum }_{j\in \left\{d_{ij}\le d_o\right\}}{R}_j}={\displaystyle {\sum }_{j\in \left\{d_{ij}\le d_o\right\}}\left(S_j/{\displaystyle {\sum }_{k\in \left\{d_{kj}\le d_o\right\}}{P}_k}\right)},$$

(7)

In the formula: i represents the demand point; j represents the supply point; d_ij represents the distance between demand point i and supply point j; R_j represents the ratio of the service capacity of supply point j to the population it serves within its search threshold range do (supply-demand ratio); P_k represents the demand population at demand point k; S_j represents the service capacity of supply point j; A_i represents the accessibility of demand point i calculated by the two-step floating catchment area method.

3.3.7. Chi-Square Goodness of Fit Test

By covering and screening the land use in the urban built-up area, the facility locations that meet the construction conditions within the urban area can be further determined, in order to better study the relationship between facilities and specific land use types. The classification of land use types refers to the research of Li et al. (2025) [43].

Considering the complexity of the land use situation within the urban area and its generally significant influence on the layout of facilities [44], in order to test the coupling status of the two types of facilities in different land use types, the chi-square goodness of fit test was introduced. Its calculation formula is as shown in Equation (8). The basic idea of the chi-square goodness of fit test is to compare the actual number of facilities in different land use types with the expected number calculated based on the proportion of land area, thereby determining whether the facility distribution has a significant coupling with certain land use types.

$${\chi }^2={\displaystyle \sum _{i=1}^k\frac{{\left(O_i-E_i\right)}^2}{E_i}},$$

(8)

In the formula, O_i represents the actual observed value, E_i represents the expected value, and k represents the number of land use types. If the chi-square value is significantly greater than the critical value (under the given degrees of freedom and significance level), it indicates that there is a significant difference between the spatial distribution of the facilities and the “random distribution according to the area ratio” assumption.

The calculation formula for the expected number of facilities E_i of the i-th type of land use is as shown in Equation (9).

$$E_i={\displaystyle \frac{A_i}{A_{total}}}\times N,$$

(9)

In the formula, A_i represents the area of this type of land, A_total is the total area of the built-up area of the city, and N is the total number of such facilities.

In order to conduct a more detailed study on the intensity of this coupling situation, this research can adopt Cramer’s V as the effect size indicator to describe the degree of layout coupling of facilities in different land use types. The specific calculation is shown in Equation (10).

$$V=\sqrt{{\displaystyle \frac{{\chi }^2}{N\times \left(k-1\right)}}}$$

(10)

In the formula, is the chi-square statistic, N is the total sample size (the total number of facilities), and k is the number of land categories. The specific numerical correspondence is as follows: (1) V < 0.1: Almost no effect (extremely weak coupling); (2) 0.1 ≤ V < 0.3: Small effect (weak coupling); (3) 0.3 ≤ V < 0.50: Moderate effect (moderate coupling); (4) V ≥ 0.5: Large effect (strong coupling).

4. Data Analysis

4.1. Analysis of Spatial Distribution Situation

Based on distribution point data, the ratio of traditional gas stations to public charging points stands at approximately 1:4.9, revealing a notable spatial overlap with areas characterised by higher density urban road networks. To further investigate the spatial distribution patterns of public charging infrastructure relative to conventional gas stations, and to examine their spatial interrelationship with urban road systems, a Standard Deviational Ellipse (SDE) analysis was employed to map the geographical dispersion of all three elements—charging stations, gas stations, and road networks. The elliptical visualisation presented in Figure 3 illustrates the directional trends and concentration patterns of these facilities, demonstrating how charging infrastructure tends to cluster in regions with well-developed road infrastructure while maintaining a distinct proportional relationship with traditional gas points. This spatial analysis reveals both the complementary distribution dynamics between emerging electric vehicle infrastructure and established fuelling networks, as well as their collective alignment with urban transportation corridors, providing valuable insights for infrastructure planning and urban development strategies.

The spatial distribution directions of roads, charging stations and gas stations are highly consistent, presenting a feature of core aggregation. The following characteristics were discovered:

(1)

The standard deviation ellipse for road networks (depicted in red) exhibits the broadest geographical coverage, with its major axis oriented along a pronounced northeast-southwest trajectory. This directional alignment closely corresponds to the primary urban development pattern of Wenzhou City, reflecting the historical expansion of its transportation infrastructure.

(2)

The traditional gas station ellipse (depicted in green) demonstrates a significantly expanded service footprint, characterised by a larger coverage area that extends into urban peripheries and suburban zones. This spatial distribution reflects the establishment of robust service outreach capacity developed over decades of operational history, enabling gas stations to effectively cater to wider geographical demands while maintaining core urban coverage.

(3)

The public electric vehicle charging station ellipse (depicted in blue) displays the most pronounced spatial clustering, with a markedly restricted spatial footprint and a centre point positioned closer to the urban core. This pattern indicates a high degree of spatial aggregation, with infrastructure deployment strategically focused on addressing demand from high-density urban population centres rather than pursuing broad territorial coverage.

The discrete point data is transformed into a continuous smooth density surface as shown in Figure 4. The following characteristics can be obtained:

(1)

Public electric vehicle charging stations exhibit a pronounced concentration in Wenzhou’s central urban area, forming a distinct peak in the city core. This region, characterised by high population density and substantial commuter vehicle demand, drives significant requirements for new energy vehicle charging infrastructure, resulting in denser station distribution [45]. Conversely, inland areas demonstrate lower density trends, though several relatively independent secondary hotspots have emerged along the eastern coastal zones. These locations align with Wenzhou’s newer urban districts and satellite cities exhibiting higher urbanisation levels.

(2)

The spatial distribution of public electric vehicle charging stations follows a “concentrated around the urban core with diminishing density towards peripheral areas” pattern, a configuration observed in other Chinese urban centres [46]. However, given the broader metropolitan scope of this study’s research area, the revealed structure is characterised as a “strong single-core—weak multi-core composite spatial configuration” reflecting hierarchical clustering dynamics.

(3)

Traditional gas stations demonstrate a more dispersed spatial distribution. While still exhibiting pronounced peaks in the core urban area, their extensibility has significantly increased, with radial expansion from the centre showing clear alignment with transportation axes. A belt-like structure has formed along the coastal zone between the Shenyang—Haikou and Ningbo—Dongguan Expressways. Additional smaller concentration zones exist, predominantly located at expressway service areas and inland logistics hubs.

(4)

The distribution of traditional gas stations serves dual functions: meeting high gas demand in the urban core while strategically positioning along intercity transport corridors to service transient vehicles and industrial facilities. This pattern is strongly influenced by traffic network topology and logistics requirements, forming a “core-axial belt diffusion structure” comprising two components: an urban expressway-aligned core extension axis and a coastal expressway-oriented belt extension axis.

4.2. Analysis of Facility Accessibility Conditions

This study employed the Gaussian-based 2-Step Floating Catchment Area (Ga2SFCA) Method to evaluate spatial accessibility of charging stations and gas stations, utilising facility capacity as a weighting factor to calculate the supply-demand ratio for each facility within defined service radii, with contributions to demand points subsequently aggregated. However, due to challenges in acquiring precise specific capacities (i.e., service capabilities) of the two facilities, coupled with disparities in their energy replenishment methodologies, this study implemented a more refined differentiation framework grounded in the search radius parameter. Regarding search radius selection, a 1 km threshold was applied for charging stations and 3 km for gas stations, with this parameter partially informed by relevant Chinese planning documents while incorporating contextual adjustments. The broader survey regions encompassed multiple cities featuring varied topography including mountainous and coastal terrain, presenting a complex mosaic of urban-rural interfaces and diverse geographical conditions. These characteristics rendered direct application of urban-centric planning norms inappropriate, necessitating spatial scale calibration to better reflect actual transportation behaviour patterns. Due to the relatively small magnitudes and high degree of clustering exhibited by the data points, they do not align with the characteristic properties of a Gaussian distribution. Employing the standard deviation-based approach would create numerous regions with missing data points. Consequently, under these circumstances, a quantile-based classification method was utilised for partitioning. As illustrated in Figure 5, the analysis revealed distinct spatial patterns in service provision, with accessibility influenced by both infrastructure distribution and topographical constraints, highlighting the importance of context-specific parameterisation in accessibility modelling for heterogeneous urban-rural landscapes. Through analysis, the following characteristics can be identified:

(1)

The accessibility of both charging stations and gas stations exhibits a general pattern of “higher in the central urban area and lower in peripheral regions,” though significant differences emerge in their spatial distributions. Charging station accessibility is markedly concentrated in the urban core, forming a concentric ring structure that rapidly diminishes with distance from the city centre, demonstrating a relatively planar attenuation pattern. In contrast, while gas stations also follow a “central-high, peripheral-low” trend, their spatial decay is more gradual, with peripheral towns retaining non-zero accessibility levels and the overall pattern showing a gentler attenuation gradient. Notably, a high-accessibility band has developed along the western corridor traversed by the “ Lishui—Wenzhou Expressway” reflecting the influence of transportation infrastructure on service provision. This disparity highlights how charging infrastructure prioritises immediate urban demand concentration, whereas gas stations exhibit broader service outreach capabilities shaped by historical development and logistics network alignment.

(2)

Both charging stations and petrol stations exhibit high accessibility near urban core areas, yet a notable “disparity” in accessibility emerges in remote regions, reflecting a typical phenomenon where the distribution of facilities aligns with population density [47]. Furthermore, under China’s land use regulations, petrol stations face stringent siting requirements, necessitating relatively large land plots and maintaining safe distances from other structures. As cities expand, rising land costs and the need for substantial storage infrastructure complicate the establishment of new petrol stations within urban boundaries. In contrast, charging stations are strategically positioned in areas where users tend to remain for extended periods—such as residential complexes, commercial hubs, and hospitality venues—with fewer constraints on plot size, enabling flexible deployment in fragmented urban spaces. Consequently, charging stations demonstrate superior accessibility within city centres compared to their petrol-based counterparts, a trend amplified by evolving urban land use dynamics and the adaptive siting flexibility inherent to electric vehicle infrastructure.

This study analysed the relationship between road traffic networks and facility accessibility by extracting the expressway network within the Wenzhou metropolitan area and conducting density and coverage assessments. As illustrated in Figure 6, the radial components of the expressway system largely align with gas station locations, though sites outside high-speed density coverage predominantly cluster in adjacent zones, exhibiting a clear propensity for “proximity to major thoroughfares.” This underscores a pronounced structural coupling between road infrastructure and gas station spatial distribution. In contrast, charging stations demonstrate extensive areas beyond high-speed road network coverage, with facilities dispersed across the territory—particularly in Wenzhou city western and southern mountainous and hilly regions—and multiple clusters situated at significant distances from primary high-speed corridors. The spatial alignment between charging infrastructure and the high-speed transportation network is markedly weaker, a characteristic that may diminish their accessibility relative to road network proximity but also highlights a location selection mechanism more strongly influenced by destination-specific attributes rather than mere connectivity to major transport arteries.

4.3. Analysis of Facility Fairness

To describe the differences in benefits received by residents between the two types of energy facilities, this study, based on three dimensions—population matching degree, spatial aggregation, and overall imbalance-calculated the population-weighted Gini coefficient (G), location quotient (LQ), and local Moran’s I (I_i), and conducted a fairness analysis based on the results.

4.3.1. Overall Imbalance Degree: Gini Coefficient Comparison

Through computational analysis, the Gini coefficient for charging stations was determined to be 0.8771, while that for gas stations stood at 0.8498, both surpassing the equilibrium threshold of 0.4, though charging infrastructure exhibits a more pronounced disparity. This finding aligns with research by Cai et al. on Shanghai, which reported a Gini coefficient of 0.86 for charging station service accessibility [48], reinforcing the observation of marked spatial imbalance. Such results suggest that public charging stations may be excessively concentrated in specific zones, potentially indicating oversaturation in certain areas and underscoring the need for more equitable distribution strategies to address accessibility gaps.

4.3.2. Space Supply-Demand Matching Degree: Cold and Hot Spot Contrast

To investigate the spatial alignment between the two facility types and demand patterns, this study first conducted global spatial autocorrelation analysis using the Global Moran’s I statistic. For gas stations, the Global Moran’s I was calculated as 0.134 (Z = 5.58, P = 0), indicating a moderate positive spatial autocorrelation. This suggests that while gas stations exhibit some clustering tendencies, their distribution also incorporates elements of random dispersion. In contrast, public charging stations demonstrated a substantially higher Global Moran’s I of 0.378 (Z = 15.40, P = 0), reflecting a strong spatial aggregation pattern. This pronounced clustering, particularly within residential and commercial zones [49], aligns with findings from previous studies and underscores that charging infrastructure is more spatially concentrated compared to gas stations. The peak distribution of public charging stations occurs within the area bounded by Oujiang Road, Tangjiaqiao Road, Ouhai Avenue, and Cuimei Avenue—a region widely recognised as the core urban centre of Wenzhou City.

Based on the Local Moran’s Index, cold-hot spot analysis provides insights into the spatial clustering and dispersion patterns of charging stations and petrol stations, while also illuminating their relationships with neighbouring areas. Leveraging this methodological framework, the study investigates the spatial matching dynamics between facility distribution and regional demand characteristics. In essence, high-value clustering signifies that a given area and its adjacent zones exhibit elevated facility density, indicating concentrated service provision, whereas low-value clustering reflects a relative scarcity of infrastructure within those regions. This analytical approach enables a detailed examination of how both facility types are spatially organised relative to population and land use patterns, offering critical implications for understanding accessibility disparities and optimising resource allocation across urban landscapes. analysing Figure 7, the following characteristics can be summarised:

(1)

The Local Moran’s HH cluster regions for public charging stations are predominantly situated in densely populated urban commercial hubs, highway service zones, or newly developed residential areas. These zones, along with their adjacent territories, exhibit a high concentration of charging infrastructure, forming critical hotspots for charging supply. This pattern corroborates the aggregation phenomenon driven by the clustering tendencies of charging stations, particularly in areas with strong user demand and land use compatibility.

(2)

LL cluster regions are primarily observed in the township areas of Wencheng County and Taishun County. Characterised by sparse populations, topographically complex landscapes, and limited infrastructure investment, these regions display a notably sparse distribution of charging facilities, reflecting challenges in service penetration and accessibility in peripheral zones.

(3)

HL outlier regions, exemplified by Baizhangji Town in Wencheng County, demonstrate a paradoxical spatial pattern: while the town itself contains a relatively concentrated charging infrastructure, its surrounding areas exhibit significantly fewer facilities. Notably, only five public charging stations were identified across the three northern county towns bordering this region, highlighting a stark spatial discontinuity in service provision.

(4)

LH outlier regions are typically located near urban core areas. Though these zones have a relatively low number of charging facilities internally, they benefit from “spillover services” originating from adjacent high-density areas. For instance, Sanyang Street in Ouhai District, with only two public charging stations within its boundaries, leverages proximity to Wutian Street and Puzhou Street—where charging infrastructure exceeds 20 units—to meet local demand through regional service sharing.

Meanwhile, as shown in Figure 8, there is no significant clustering characteristic at the urban centre area of the gas stations. The specific features are as follows:

(1)

HH clusters are predominantly situated along major transport arteries or logistics hubs, with the most striking manifestation being the hollow circular configuration encircling the urban core—a characteristic feature of early gas station distribution patterns. This spatial arrangement underscores the superior service spillover capacity of gas stations relative to charging infrastructure, as evidenced by their ability to serve peripheral areas through strategic positioning.

(2)

LL clusters exhibit a spatial alignment with charging station distribution trends, with both facility types demonstrating a propensity to avoid mountainous topographies in site selection [50]. However, while gas stations prioritise traffic-related factors such as proximity to highways, charging stations appear more influenced by residential development patterns, reflecting divergent location choice mechanisms.

(3)

HL outlier regions are exemplified by Shatou Town in northern Wenzhou, where Provincial Road No.223 and the Zhuji—Yongjia Expressway traverse its administrative boundaries. Gas stations layout within this zone follows a distinct linear pattern along these corridors, contrasting sharply with the absence of facilities in five of the seven neighbouring towns—a phenomenon highlighting spatial discontinuity in service provision.

(4)

LH outlier regions, such as Pandai Street in Rui’an City, lack gas station infrastructure entirely. Nevertheless, the area benefits from spillover services originating from adjacent transportation hubs connected by the Wenzhou Ring Expressway, Shenyang—Haikou Expressway, and National Highway 104, demonstrating how peripheral zones can leverage proximity to major transport networks for service access despite local deficiencies.

Through analysis of global and local Moran indices, distinct spatial supply-demand matching patterns between charging stations and petrol stations become evident. Charging infrastructure demonstrates a more pronounced spatial clustering effect, with significant hotspot formation indicating uneven resource distribution that favours densely populated zones. This suggests a tendency toward resource aggregation in areas with high demand, potentially leading to localised oversaturation. Conversely, petrol stations exhibit a more historically established and mature spatial configuration, showing stronger alignment with transportation networks. While their clustering is less intense compared to charging stations, petrol stations demonstrate superior service spillover capabilities, effectively supplementing coverage in underserved peripheral areas through strategic positioning along major transport corridors.

4.3.3. Per Capita Facility Matching Degree: Location Quotient Analysis

The results of the Location Quotient (LQ) analysis are helpful for a deeper understanding of the differences in population service between charging stations and gas stations, as well as the fairness of facility configuration. The division method is an equal-interval division method with an interval of 0.3. According to the definition, when the LQ value is around 1, the matching degree is considered reasonable. As shown in Figure 9a, it can be observed that:

(1)

High-value areas (LQ ≫ 1) are predominantly concentrated in southwestern regions such as Yongjia County, alongside coastal zones including Longwan, Rui’an, and Pingyang. These areas exhibit a significantly higher per capita charging station density compared to the city average, reflecting pronounced resource concentration in regions with specific developmental or demographic characteristics.

(2)

Low-value areas (LQ ≪ 1) are dispersed across mountainous territories in northern Wenzhou or within urban cores of county-level cities like Cangnan County. Notably, multiple zones register LQ = 0, indicating complete absence of charging infrastructure—a phenomenon particularly evident in economically peripheral or topographically challenging locales.

(3)

Charging stations demonstrate elevated location quotient in newly established development zones and industrial clusters. For instance, Longwan District’s coastal area, despite a population of only 4000 residents, hosts 22 public charging stations—a stark illustration of how infrastructure deployment prioritises strategic industrial or logistical hubs over residential population metrics.

Overall, in remote suburban towns, charging stations exhibit low compatibility with local population distributions, whereas urban core areas such as Lucheng and Ouhai districts demonstrate strong compatibility, with most streets registering location quotient (LQ) values close to 1. Analysis of spatial cold and hot spots reveals a positive correlation between charging station density and population in core urban zones: high-density centres absorb facility expansion through rapid population growth, with charging demand predominantly concentrated in areas of intense human activity [51]. Conversely, in low-density peripheral regions, even modest infrastructure deployment can disproportionately inflate LQ metrics due to smaller resident populations. Notably, in certain northern and southern areas exhibiting elevated LQ values, population factors exert significant influence—regions with fewer than 5000 residents may achieve LQ values far exceeding the city average through the installation of just 1–2 charging stations. This phenomenon parallels observations in tourist destinations and expressway service areas, where strategic facility placement creates localised demand spikes unrelated to permanent population size [52].

From Figure 9b, it can be observed that the total number of gas stations is only 285, and their distribution is more limited, with a wider area that is not covered. The specific characteristics are as follows:

(1)

High LQ areas are predominantly concentrated at the southwestern junction of Yongjia and Rui’an, reflecting per capita gas station facilities exceeding the municipal average. A striking example is found in Qiaoxia Town (Yongjia County), where 10 gas stations operate within its administrative boundaries—a figure indicating excessive concentration. Notably, these stations are strategically positioned along expressway corridors and coexist with substantial industrial infrastructure. This configuration aligns more closely with transportation-driven facility deployment rather than conventional population-based demand models.

(2)

Medium-equilibrium LQ zones (approximately 0.9–1.2) exhibit a more fragmented spatial distribution compared to charging stations. This pattern likely arises from compounding factors including safety regulations governing gas station construction and land use economics [53]. Unlike charging stations, which demonstrate a clear “population-facility positive correlation,” gas station distribution adheres to a “transportation-facility coupling” paradigm, where infrastructure development is primarily shaped by vehicular mobility networks rather than residential density metrics.

4.4. Analysis of Coupling Between Facilities and Land Use Types

During the research process examining land use types associated with the two facility categories, we extracted the land use classification status within the urban built-up area of the study region using Li’s (2025) basic urban land use classification map for Chinese cities [43]. By integrating this data with POI (Point of Interest) records for both facility types within the same geographic scope, we conducted a coupled analysis of the relationship between facility quantity and land use type, as illustrated in Figure 10. A chi-square goodness-of-fit test was performed by comparing actual facility counts across different land use categories with expected values calculated proportionally to land area distribution. Through this comparative analysis of observed and expected values, we evaluated whether facility distribution exhibited significant coupling with specific land use types, with results presented in

Table 3. Statistics of observed and expected values of urban land use types and facility quantities within the built-up area of Wenzhou City.

Land Use Category	Urban Land (Area) 1/km2	Observed Value		Expected Value
		Number of Stations (EV)	Number of Stations (GS)	Number of Stations (EV)	Number of Stations (GS)
Residential	591.47	313	62	406.8	99.7
Business office	53.42	74	11	36.7	9
Commercial service	127.6	55	1	87.8	21.5
Industrial	69.84	114	57	48	11.8
Transpontition stations	63.22	31	5	43.5	10.7
Airport facilities	1.84	2	0	1.3	0.3
Administrativen	42.25	21	16	29.1	7.1
Educational	35.87	35	3	24.7	6
Medical	22.29	12	0	15.3	3.8
Sport and cultural	11.04	22	1	7.6	1.9
Park and Greenspace	310.01	235	68	213.2	52.3
Sum	1328.85	914	224	914	224

¹ Data source: Zenodo Data Sharing Platform (https://zenodo.org/records/16794007, accessed on 22 August 2025).

. It is important to note that the calculation of expected values maintains the fundamental statistical requirement that their total sum must equal the total sum of actual observed values.

Based on the chi-square goodness of fit test results and Figure 10, the following observations emerge:

(1)

For public electric vehicle charging stations (EV): χ² = 203.07, Cramer’s V = 0.15 (classified as weak coupling). This indicates that while the spatial distribution of charging stations across land use types significantly deviates from area-proportional expectations, the magnitude of this divergence is relatively modest. This aligns with the developmental stage of charging infrastructure, which remains in an expansion phase characterised by dispersed layouts rather than heavy reliance on specific land use categories. Notably, actual facility counts exceed expected values in commercial/office, educational/research, sports/cultural, and park/greenspace land uses, while falling short in residential and transportation-related zones. This suggests a preference for multifunctional composite areas or high-activity urban zones where residential and commercial functions coexist. It is also worth highlighting that residential land may contain unaccounted “private charging piles,” potentially indicating higher actual coverage in mixed commercial-residential areas than observed.

(2)

For gas stations (GS): χ² = 232.82, Cramer’s V = 0.32 (classified as moderate coupling). This demonstrates stronger concentration in specific land use types, with deviations from area proportions being more pronounced compared to charging stations. This reflects the maturity and site-selection dependency of traditional energy infrastructure. Gas stations predominantly cluster in industrial and administrative/office land uses, while being significantly underrepresented in commercial, medical, and educational zones. This spatial pattern mirrors the historical functional orientation of gas stations, which primarily served industrial logistics and conventional transportation demands. From an urban land use perspective, these findings underscore the relatively homogeneous functional attributes of industrial and administrative zones in accommodating traditional energy facilities.

5. Discussion

5.1. Spatial Layout Differences and Mechanisms

Energy infrastructure, as a material entity, is invariably situated within specific geographical contexts [54]. This study demonstrates that the spatial differentiation between charging stations and petrol stations fundamentally arises from the interplay of functional orientation and land use constraints. Despite their shared categorisation as energy facilities, significant disparities exist in their siting logic and operational priorities. The “strong single-core—weak multi-core” configuration of charging stations—manifested through a density peak concentrated in the urban core—reflects their demand-dependent nature. Electric vehicles’ inherently shorter driving ranges compared to internal combustion engine vehicles lead users to prefer operating within areas featuring higher charging infrastructure density, potentially inducing a “Malthusian effect” that fosters clustered service islands rather than axial belts or networks. Furthermore, refuelling behavioural differences reinforce this pattern: while fuel vehicles can refuel en route without additional stopping, electric vehicle charging requires longer dwell times, prompting users to prioritise charging near daily destinations or residential locations—activities inherently linked to parking availability [55,56]. Consequently, public electric vehicle charging stations exhibit relatively low location flexibility. In essence, electric vehicle users typically do not undertake long-distance journeys solely for charging during routine travel. Synthesising these factors including geographical environmental conditions and resident travel behavioural variations, the study employs distinct service radius parameters for petrol stations and charging stations, aiming to more precisely capture the spatial response characteristics and supply-demand dynamics inherent to both facility types in practical usage scenarios.

On the other hand, charging infrastructure, having evolved from conventional urban systems, retains inherent spatial compatibility by integrating established functionalities [57], allowing it to swiftly “embed” within urban fabric such as parking facilities and commercial complexes while simultaneously fulfilling parking and energy replenishment requirements. In contrast, petroleum refuelling stations face stringent safety distance regulations during construction, posing challenges for rapid urban integration. As urban land use patterns grow increasingly complex and land values escalate, these facilities have gradually migrated toward peripheral zones such as transportation hubs or high-density freight corridors, thereby establishing a “transportation-infrastructure” coupling paradigm that prioritises operational efficiency over immediate spatial accessibility. This structural divergence reflects broader shifts in energy transition dynamics, where legacy infrastructure adapts through spatial reconfiguration while emerging systems leverage existing urban frameworks for accelerated deployment. When comparing the differences in the two types of energy facilities in urban land use: charging facilities tend to be evenly distributed in multi-functional composite areas, while fueling facilities continue to follow the centralised distribution pattern centred around industries and administrative districts.

5.2. Structural Dilemma of Spatial Equity

From the perspective of time evolution, the development of Chinese charging infrastructure network has transitioned from an early policy-driven model to a market-driven paradigm. During initial implementation stages, government guidance strongly shaped infrastructure deployment, while later phases saw significant private capital inflows driving expansion through market mechanisms [58]. This study’s multi-dimensional analysis of charging stations reveals a persistent pattern of resource overconcentration in high-density urban cores and newly developed districts. Charging facilities are frequently deployed in advance along urban transportation corridors or planned new urban zones, with inadequate attention to service equity in older urban neighbourhoods and marginalised communities. While theoretically increased facility numbers should enhance service fairness, this research finds that capital-driven scale expansion has paradoxically worsened spatial imbalances. Government policy support for the Chinese new energy industry has indeed stimulated charging stations construction and market growth, attracting substantial capital that accelerated public charging infrastructure expansion. However, this capital-centric approach has introduced challenges such as speculative planning practices and inefficient spatial allocation. Case data from this study illustrate these dynamics: excluding unidentified operators, private enterprises constitute over 65% of charging station operators, in stark contrast to the gasoline station sector where state-owned enterprises dominate nearly 95% of operations. In 2015, the Chinese government unveiled the “Guidelines for the Development of Electric Vehicle Charging Infrastructure (2015–2020),” which emphasised “prioritizing the installation of charging stations” and pursuing “appropriately advanced” construction while focusing more on quantitative targets. However, the June 2025 “Notice on Promoting the Scientific Planning and Construction of High-Power Charging Facilities,” issued jointly by the National Development and Reform Commission’s Office and other agencies, shifted focus to “avoiding resource waste and disorderly development,” advocating instead for government-led coordination to promote intensive resource utilisation and factor optimisation. This policy evolution underscores China’s strategic recalibration in new energy vehicle charging infrastructure development, reflecting a transition from quantity-driven expansion to quality-oriented, resource-efficient planning. Research indicates that geographical and socio-economic disparities significantly shape the spatial distribution of electric vehicle charging infrastructure [59]. However, given China’s strong policy-driven approach to urban development, this phenomenon may reflect strategic preparatory positioning of infrastructure to support new energy vehicle (NEV) adoption in newly constructed zones or urban renewal areas. Such site selection strategies prioritise territorial coverage expansion within designated demand zones while demonstrating limited attention to equitable service accessibility for local populations—a critical aspect of spatial justice frequently neglected in capital-led infrastructure planning. This dynamic underscores broader tensions between centralised policy objectives and grassroots community needs, particularly in established residential areas where charging access remains constrained despite overall network expansion [60]. While subsidies and planning incentives have driven substantial short-term construction growth, these measures may not have sufficiently incorporated granular considerations of real-world utilisation scenarios, risking misalignment between infrastructure deployment and actual user demand patterns.

5.3. Perspective of Comparative Migration

Comparative studies of service infrastructure typically focus on intra-category comparisons across regions or temporal periods, though some analyses extend to heterogenous facilities sharing functional proximity [61,62]. This latter approach, frequently observed in healthcare infrastructure research examining tiered service systems, explores spatial interdependencies between facilities with divergent operational capacities or user demographics [63,64]. Within contemporary new energy infrastructure scholarship, such interdisciplinary comparative frameworks remain underutilised. Most existing analyses position petrol stations as passive objects requiring modernisation within evolving energy transition paradigms, with discourse predominantly centring on their technical conversion into charging infrastructure rather than systematic examination of fundamental spatial-functional disparities. This reductionist perspective overlooks critical divergences in siting logic, service radius determinants, and land use integration mechanisms between legacy fuelling systems and emerging electric vehicle support networks. By contrast, a comparative lens reveals how petrol stations’ “transportation-facility coupling” model—rooted in vehicular mobility patterns and safety regulations—contrasts sharply with charging infrastructure’s “urban fabric embedding” strategy, which leverages pre-existing urban amenities for demand-driven deployment. Such analytical oversight risks perpetuating planning paradigms that neglect the distinct operational ecosystems governing these two energy distribution systems.

Although both charging stations and petrol stations possess well-established research foundations, this unidimensional analytical framework fails to adequately elucidate their reciprocal influence mechanisms and spatial coupling dynamics. Despite the rapid expansion of electric vehicle market penetration, internal combustion engine vehicles and their supporting fueling infrastructure will retain significant operational relevance in the foreseeable future. Consequently, conceptualising petrol stations exclusively through the lens of “replacement” or “upgrading” proves insufficient for comprehensively grasping the synergistic and interdependent relationships between these dual energy distribution systems within urban contexts. A juxtaposed analytical perspective—examining their spatial coexistence and functional interplay—offers a more nuanced understanding of infrastructure layout dynamics. This approach not only clarifies competitive and complementary interactions but also provides actionable insights for optimising new facility deployments and retrofitting existing infrastructure, thereby enhancing the resilience and efficiency of urban energy transition processes. Such integrated analysis is particularly critical given the persistent coexistence of diverse propulsion technologies in metropolitan environments, where spatial planning must balance immediate operational requirements with long-term decarbonisation objectives.

5.4. Reference Value and Suggestions

This research framework employs quantitative evaluation of facility distribution through indicators such as spatial clustering, accessibility, and fairness, with its methodology designed to be context-agnostic and transferable across regions. Leveraging available data, the analytical approaches and tools developed herein can be readily adapted to other cities and countries globally. In terms of academic contributions, this study adopts a comparative transfer perspective that transcends conventional single-facility analysis by juxtaposing charging stations (representing new energy infrastructure) with gasoline stations (traditional energy infrastructure). This dual-lens approach elevates the research from mere descriptive analysis of facility status and trends to a more nuanced analytical framework that examines how historical patterns of energy infrastructure development influence contemporary spatial logic. By systematically comparing these two facility types, the research reveals how legacy energy systems shape the distribution of emerging sustainable infrastructure, both through historical evolutionary pathways and current demand dynamics. This horizontal comparative lens—contrasting new energy facilities with their traditional counterparts—provides a higher-order explanatory framework for understanding the complex mechanisms governing energy infrastructure layout, offering insights that extend beyond sector-specific discussions to inform broader urban planning and policy-making contexts. Based on the characteristics of this research, for the construction of new energy facilities in the future and the transformation and upgrading of traditional energy facilities, the following suggestions are hereby put forward:

(1)

Strengthen the government’s role in planning and guidance, and promote the construction of related planning systems. Strengthening the binding force of planning at both national and local levels is essential, particularly through integrating new energy vehicle charging infrastructure into urban renewal frameworks, urban-rural coordinated development strategies, and transportation network planning. China’s January 2025 release of the “Design Standards for Electric Vehicle Charging Stations” marks a critical institutional step, establishing clear construction benchmarks and operational principles. For effective implementation, local authorities should enforce spatial equilibrium through coverage rate mandates and regulatory benchmarks that guide operator deployment decisions, thereby mitigating regional disparities caused by profit-driven capital allocation. This structured approach ensures charging infrastructure development aligns with broader urban planning objectives while balancing market efficiency with equitable service distribution.

(2)

Improve and coordinate the market mechanism, and promote the synergy between capital and policies. Promoting the adoption of the Public–Private Partnership (PPP) model is essential to incentivize social capital participation in building charging infrastructure at community and county levels. In terms of policy support and financial subsidies, greater emphasis should be placed on dynamic principles that guide development based on actual demand and coverage requirements, thereby avoiding the pitfalls of excessive quantity accumulation or over-construction. This approach ensures infrastructure deployment aligns with real-world usage patterns while maintaining fiscal responsibility and operational efficiency.

(3)

Promote the transformation and integration of new energy facilities with traditional energy facilities. In China, traditional gas stations are predominantly operated by state-owned enterprises, featuring extensive coverage and a relatively mature layout across the road network, serving as critical infrastructure for energy supply. Moving forward, the focus should be on exploring transformation models for “oil, gas, electricity, and hydrogen” integrated energy service stations, while advancing the gradual evolution of gas stations into diversified energy supply hubs. This shift would not only modernise energy distribution infrastructure but also help alleviate the “range anxiety” experienced by electric vehicle drivers during long-distance travel.

(4)

Emphasise the deep integration with transportation demands and social equity. In the foreseeable future, as electric vehicle adoption continues to grow, infrastructure development must achieve more precise alignment with transportation demand patterns. This requires enhancing spatial integration with public transit hubs, residential neighbourhoods, and major commuting corridors to ensure equitable access to charging services across diverse social groups. Concurrently, social equity should be institutionalised as a core evaluation metric for infrastructure planning, ensuring that deployment strategies explicitly address accessibility disparities and prioritise underserved populations. Such measures are critical to creating a sustainable charging network that supports both technological adoption and inclusive urban mobility.

6. Conclusions

This study is based on multi-source geographic spatial data and employs various methods such as kernel density analysis, Gaussian two-step moving search algorithm (G2SFCA), spatial autocorrelation (Moran’s I), and location quotient (LQ) to select the Wenzhou city in China as the survey regions. It conducts a systematic juxtaposition and comparison of the spatial layout characteristics, accessibility differences, and fairness levels of public charging stations and traditional gas stations. The main conclusions obtained are as follows:

(1)

Public electric vehicle charging stations demonstrate a “strong single-core and weak multi-core” clustering structure, predominantly situated in core urban zones with high human activity density and exhibiting a robust correlation with population spatial distribution patterns. Traditional gas stations, by contrast, manifest a “core—axial band-like diffusion” characteristic, forming close coupling with transportation arteries and logistics nodes while possessing superior radiation capacities and broader service extensiveness. The results of the chi-square goodness of fit test indicate that both types of facilities show a significant coupling with the land use types, but the intensity varies. Charging stations exhibit a more dispersed and flexible layout, while gas stations are more concentrated in specific land uses such as industrial and administrative areas.

(2)

At the accessibility level, both facility types present a “centre-high, periphery-low” spatial pattern, though charging stations exhibit a more pronounced decline in peripheral regions with relatively fragile spatial capacity response. Gas stations, tightly integrated with road network development, achieve higher accessibility scores in certain mountainous and suburban areas along major transportation corridors.

(3)

Fairness measurement reveals that the spatial-population alignment of charging station services is generally inferior to that of gas stations, evidenced by higher Gini coefficients and steeper location quotient (LQ) value distributions. This suggests dominant influences of capital-driven siting decisions over population demand in specific regions, leading to resource over-concentration or even wastage.

(4)

From a research perspective, there is a pressing need to shift from isolated analyses of new energy charging infrastructure toward comparative examinations of heterogeneous energy facilities. Energy transition and infrastructure development are incremental processes; the “substitution” model of facility renewal operates more as evolutionary growth built upon existing foundations. While siting mechanisms and market conditions differ fundamentally, spatial interdependencies between legacy and emerging energy systems warrant deeper exploration.

The study acknowledges certain limitations that warrant consideration. For instance, there are indeed certain deficiencies in the “systematic comparison” aspect. The standardisation of relevant indicators has not been well presented, and the comparability issues caused by the differences in the units of different indicators have not been discussed in greater depth. In the specific research, the power levels of charging piles (fast charging/slow charging) were not distinguished, which may lead to overestimation or underestimation of the actual service capacity. In the future, based on more refined data, more detailed work should be carried out to try to construct a hierarchical Ga2SFCA model or a grouped comparison model to improve the practicality and alignment of the results with the real situation. The data of the electric vehicle travel chain (such as navigation trajectories), mobile phone signal data, traffic flow, etc., which are more dynamic, were not included. It is difficult to precisely simulate dynamic demands. The simulation of travel demand and behaviour based on big data for the study of the usage status of facilities is a direction that can be referred to in the future. In addition, in future research, by introducing relevant quantitative indicators and time-space data analysis, the interaction between policies and market mechanisms can be dynamically evaluated, and a more comprehensive analysis of the game and collaboration mechanisms between “capital-driven” and “policy regulation” can be conducted. At the same time, considering the development of the artificial intelligence field in recent years, if its “illusion rate” can be effectively reduced, it can be considered to build a charging station network decision-making system based on it [65,66]. These gaps present valuable avenues for future research to refine methodological approaches. Overall, this paper undertakes a comparative analysis of public charging infrastructure and traditional fueling stations, offering critical reflections on the adverse trends observed in contemporary new energy facility development. By elucidating spatial disparities and interdependencies between legacy and emerging energy systems, the study contributes to steering future infrastructure planning toward more systematic, precise, and equitable configurations. Such insights hold significant implications for urban development processes, facilitating informed decisions that align with energy structure greening objectives and sustainable urban transition goals.