Drifting toward Alliance Innovation: Patent Collaboration Relationships and Development in China’s Hydrogen Energy Industry from a Network Perspective

Abstract

The hydrogen energy industry, as one of the most important directions for future energy transformation, can promote the sustainable development of the global economy and of society. China has raised the development of hydrogen energy to a strategic position. Based on the patent data in the past two decades, this study investigates the collaborative innovation relationships in China’s hydrogen energy field using complex network theory. Firstly, patent data filed between 2003 and 2023 are analyzed and compared in terms of time, geography, and institutional and technological dimensions. Subsequently, a patent collaborative innovation network is constructed to explore the fundamental characteristics and evolutionary patterns over five stages. Furthermore, centrality measures and community detection algorithms are utilized to identify core entities and innovation alliances within the network, which reveal that China’s hydrogen energy industry is drifting toward alliance innovation. The study results show the following: (1) the network has grown rapidly in size and scope over the last two decades and evolved from the initial stage to the multi-center stage, before forming innovation alliances; (2) core innovative entities are important supports and bridges for China’s hydrogen energy industry, and control most resources and maintain the robustness of the whole network; (3) innovation alliances reveal the closeness of the collaborative relationships between innovative entities and the potential landscape of China’s hydrogen energy industry; and (4) most of the innovation alliances cooperate only on a narrow range of technologies, which may hinder the overall sustainable growth of the hydrogen energy industry. Thereafter, some suggestions are put forward from the perspective of an industrial chain and innovation chain, which may provide a theoretical reference for collaborative innovation and the future development and planning in the field of hydrogen energy in China.

1. Introduction

Hydrogen energy is a type of secondary energy, with abundant sources, that is green, low carbon, and can be widely used in areas such as transportation, power and heat generation, industrial production, aerospace and aviation, construction, and many other fields [1]. As the global energy transition continues to advance, hydrogen energy has become an important way to combat climate change and build a decarbonized society, and it is a hotspot for global industrial development [2,3,4]. To date, a great number of developed nations worldwide have prioritized the advancement of the hydrogen energy industry, making it a crucial strategic option for expediting energy transformation and upgrading systems while fostering new avenues for economic growth. The essential technologies of the global hydrogen energy industry chain are progressing, with fuel cell shipments experiencing rapid growth and costs continuing to decline. Moreover, there has been a significant acceleration in the construction of hydrogen energy infrastructure, leading to the gradual formation of regional networks for hydrogen energy supply [5,6,7,8,9].

As one of the world’s most important economies, China has incorporated hydrogen energy into its key future industries and has formulated medium- and long-term development plans to lead the high-quality development of the hydrogen energy industry [3,4,6,8,9]. Currently, China’s hydrogen production ranks first in the world; the market size of China’s hydrogen energy industry has exceeded CNY 500 billion, and is expected to be close to half of the global market by 2027. However, compared with the countries leading the development of the hydrogen energy industry, China’s hydrogen energy industry is still in the early stage of growth, facing many challenges [3,10]. Firstly, there is a serious homogenization of regional policy planning, and insufficient coordination at the national level. Second, technological innovation is insufficient, and certain key technologies in the industry chain lag behind foreign countries. Finally, economy is a key factor restricting the development of the hydrogen energy industry, with the high cost of the raw materials needed for hydrogen production and significant expenses associated with storage and transportation.

The acceleration of technological breakthroughs can be achieved through a collaborative approach between fundamental and applied research. Although enterprises are recognized as the primary drivers of innovation in the field of hydrogen energy, they often face challenges in theoretical research. Consequently, besides enhancing inter-enterprise cooperation, enterprises should also strengthen partnerships and technical collaborations with universities and research institutions, to facilitate the clustering development of innovative entities [11,12].

Hydrogen energy patents are key technical indicators for measuring the level of hydrogen energy development in a country and region. According to the “Top 100 Global Hydrogen Energy Industry Invention Patent Rankings 2022”, published by IPRdaily and IncoPat (http://www.iprdaily.cn/article_33358.html, accessed on 18 January 2024), China has demonstrated outstanding performance in hydrogen energy patents, occupying first place worldwide, with a 32% share in patent applications and a 40% share in the total number of patents granted. Chinese enterprises such as Sinopec, Huaneng Group, and State Grid are at the top of the list, demonstrating the innovative strength among Chinese entities regarding hydrogen energy patents.

Patent cooperation plays an important role in promoting the research and development of technology and regional collaborative innovation. Different enterprises, research institutes, universities, and others use patents to cooperate, exchange ideas, and innovate. This is conducive to the development and application of hydrogen energy technology, improving the competitiveness and influence of the hydrogen energy industry, as well as enhancing a country’s economic development capability. It is crucial to investigate how to promote hydrogen energy technological innovation and cost reduction while establishing a collaborative innovation system for hydrogen energy industry synergy and regional cooperation [13,14,15]. These efforts will ultimately enhance the competition and innovation ability of the whole production chain of the hydrogen energy industry in China. Therefore, it is essential to study the current status, cooperative relationships, and the direction of hydrogen energy-based collaborative innovation in China.

The purpose of this paper is to investigate the collaborative innovation relationships in the hydrogen energy sector of China. Firstly, institutions were selected to be major research objects based on the statistics of China’s hydrogen energy patent data from 2003 to 2023. The patent data were compared and analyzed in four dimensions, namely time, geography, institution, and technology. Then, a patent collaborative innovation network was constructed, with innovation entities as nodes and cooperation relationships as edges. The topological characteristics and structural evolution features of the five development stages were studied. The core innovation entities in China’s hydrogen energy industry were identified using classical centrality indicators. Thereafter, community clusters (i.e., innovation alliances) within the collaboration network were detected using a community detection algorithm, and the collaborative relationships and technical directions of innovation entities within the communities were analyzed. Finally, a discussion and summary was conducted from the perspective of the hydrogen supply chain. The study intends to reveal the development process of China’s hydrogen energy industry, draw a map of innovation and cooperation, and provide a theoretical basis for reference in the formulation of relevant policies.

The remainder of this paper is organized as follows. The research related to hydrogen energy and patent cooperation networks is reviewed in Section 2. Section 3 presents the main research methods of this study. Section 4 introduces the patent data sources and their statistical characteristics. Section 5 investigates and analyzes the collaborative innovation relationship of Chinese hydrogen energy research from the perspective of a complex network. Section 6 conducts a discussion and summary from the perspective of the hydrogen supply chain. Finally, some conclusions and proposals are given in Section 7.

2. Literature Review

Collaborative innovation is an innovation pattern that focuses on knowledge value-added and involves large-span integration among enterprises, governments, knowledge-producing institutions and intermediaries in order to achieve major scientific and technological innovation [16]. Regional collaborative innovation refers to the overall coordinated development of a region through the effective integration and collaboration of innovation resources and elements among innovation entities [17]. From the perspective of industrial collaborative innovation, scholars have conducted a number of studies in various areas. Wang et al. [18] revealed the mechanism of the collaborative innovation process by investigating the complex relationship between key factors affecting enterprises’ innovation performance in supply chain networks, based on knowledge management and innovation capability theories. Hong et al. [19] proposed a conceptual model for co-innovation in green supply chain industries, to understand the relationship between green supply chain co-innovation and innovation performance in organizations better. Many scholars explored collaborative innovation ability and partner selection when studying the mechanism and path of collaborative innovation among enterprises. Santamaria et al. [20] investigated the impact of the research partner’s geographic proximity and diversity on the outcome of innovation projects, which emphasized the importance of understanding the determinants of innovation failure in order to manage the innovation process. Csedo et al. [21] suggested an inter-organizational knowledge network platform, aimed at expediting the progress of sustainable development based on hydrogen.

From the perspective of hydrogen energy technology analysis and cooperative innovation, Yu et al. [22] constructed a Chinese hydrogen fuel cell vehicle innovation network using patent information from 25 Chinese provinces, filed between 2001 and 2020. Based on the dimensions of knowledge production, knowledge consumption, and network brokers, the network positions of these sample provinces in three periods divided by four main national policies were classified. Choi et al. [23] employed unsupervised learning to analyze patent data and identify technological themes in hydrogen energy across major countries. Their findings revealed that South Korea and Japan prioritize fuel cell technology, while France and the United States focus on hydrogen production. Tlili et al. [24] assessed the transport and natural gas sectors in the US, Europe, Japan, and China. They analyzed the feasibility of replacing carbon-based technologies in the market with new low-carbon hydrogen systems, and the time frame required for hydrogen to reach the competitiveness needed. Park et al. [25] predicted the demand for hydrogen cell electric vehicles in 2040, including the yearly and daily hydrogen requirements per charging station.

Several scholars have conducted research on the scale, density, and structure of the patent cooperation network from a complex network perspective. For example, the collaborative network within China’s wind industry was analyzed by Liu et al. [26], utilizing patent data obtained from the State Intellectual Property Office (SIPO). They argued that collaborative innovation is crucial for promoting the sustainable and healthy development of the wind energy industry. Alvarez-Meaza et al. [27] utilized bibliometrics and patent data analysis to examine scientific research and technological developments in the domain of fuel cell electric vehicles (FCEV) from 1999 to 2019. Liu et al. [28] constructed the patent cooperation network within China’s smart grid sector. They analyzed the patent application data from the SIPO, regarding patent cooperation related to the smart grid. The study focused on the network characteristics, structure, and patent cooperation trends, and offered relevant suggestions. Wang [29] explored the patent network of wireless communication collaborative innovation in the logistics field, covering the period from 2000 to 2020, and identified the main innovation entities and changes in the network structure. Liu et al. [30] constructed a patent collaboration network for China’s nuclear power field. They analyzed the structure of the network using the social network analysis (SNA) method, and explored the characteristics of the network from the perspective of multidimensional proximity.

Some scholars also studied the key institutions and their roles in patent cooperation between specific institutions. They predicted potential competing partners by analyzing the centrality of nodes and link predictions in patent cooperation networks. Xu et al. [31] constructed an AI collaborative innovation network and an AI patent cooperation network, based on nearly ten years of patent cooperation data in the field of AI in the Yangtze River Delta region of China. They explored the basic features and spatio-temporal evolution of these two networks, and predicted the future partnership between cities and innovation entities.

To summarize, although scholars have conducted significant and meaningful research, there are still some limitations in content, scope, and methods. Firstly, innovation research in the field of hydrogen energy is mostly limited to the general situation and prospect [5,6], the international comparison of policies and strategies [2,8,9], and the identification of technical direction and the prediction of technical feasibility [23,24]. There is a lack of identification of core innovative entities, of community clusters of innovation entities, and a lack of analysis of alliance-style collaborative relationships. Secondly, there are many research studies on collaborative innovation relationships from a network perspective, but very few of them are specific to China’s hydrogen energy field. In addition, existing works are limited to network characterization and structural evolution using SNA methods [26,27,28,29,30]. Thirdly, hydrogen energy collaborative innovation studies are often limited to a hotspot area, such as hydrogen fuel cells or hydrogen fuel cell vehicles [3,11,22,25,27], and lack an overview of collaborative innovation in the entire hydrogen energy industry. Therefore, in this work, we explore the innovation collaborative relationships in the whole Chinese field of hydrogen energy, from a network perspective, utilizing centrality indicators and community detection algorithms to identify core innovation entities and innovation alliances, respectively, and to investigate the collaborative relationships of innovation alliances.

3. Research Methodology

Most innovation activities involve interactions among different innovation entities such as enterprises, universities, and research institutes. These interactions and collaborative relationships of innovation entities can be abstracted and modeled by complex networks. In this paper, we use Python’s NetworkX and igraph packages to compute the topological characteristics of network and centrality indicators. We also employ classical algorithms for community detection to identify community clusters (i.e., innovation alliances) within the network.

3.1. Centrality Indicator

In complex network theory, node centrality is a crucial attribute for estimating the importance and influence of a node. Freeman [32] proposed the following four types of centrality measures of networks: degree centrality (DC), weighted degree centrality (WDC), betweenness centrality (BC), and closeness centrality (CC). A higher level of node centrality suggests stronger connections and a stronger impact on other nodes within the network, as well as a position of greater significance in the network.

Degree centrality and weighted degree centrality refer to the count of nodes directly adjacent to a given node and are calculated as shown in (1), where $a_{ij}$ denotes the elements of the adjacency matrix. If node i is adjacent to node j, $a_{ij}$ = 1; otherwise $a_{ij}$ = 0. ${\omega }_{ij}$ represents the weights of the edges in weighted networks and ${\omega }_{ij}$ = 1 in unweighted networks.

$$\mathrm{DC}\left(i\right)=\sum _{j=1}^n{a}_{ij}\ast {\omega }_{ij}$$

(1)

Betweenness centrality is a network centrality measure based on shortest paths, quantifying the degree of control that a node possesses over a resource, calculated as shown in (2). Here, nodes j and k represent any two nodes in the network, excluding node i, $s{p}_{jk}$ denotes the count of shortest paths between node j and node k, and $s{p}_{jk}^i$ represents the number of these paths that pass through node i.

$$\mathrm{BC}\left(i\right)=\sum _{i\ne j\ne k}^{ }\frac{s{p}_{jk}^i}{s{p}_{jk}}$$

(2)

Closeness centrality is the sum of the shortest path lengths between a given node and all other nodes in the network, calculated as shown in (3), where $d_{ij}$ denotes the shortest distance between nodes i and j. If there is no reachable path between these two nodes, then $d_{ij}$ = ∞.

$$\mathrm{CC}\left(i\right)=\sum _{j=1}^n\frac{1}{d_{ij}}$$

(3)

3.2. Community Detection Algorithm

Community detection refers to the process of identifying well-connected parts of a network and dividing them into smaller modules called communities. These communities are characterized by dense interconnections within themselves and sparse connections with other communities. Many algorithms have been proposed for community detection. What follows is a brief overview of four community detection algorithms used in this study, namely the Girvan–Newman (GN) algorithm, the label propagation algorithm (LPA), the k-clique algorithm, and the adjacent node similarity optimization combination connectivity algorithm (ASOCCA).

The GN algorithm is a community detection method that relies on graph partitioning principles [33] and uses the edge betweenness to denote the effect of each edge on the network connectivity. The edge betweenness refers to the count of shortest paths through this specific edge within the network. The edge betweenness of inter-community connectivity edges is relatively large, while the edge betweenness of intra-community edges is relatively small. By identifying and removing edges with higher edge betweenness, the communities in the network can be partitioned naturally.

LPA is a community detection algorithm that uses label propagation as its underlying principle [34]. The concept of label propagation is very concise—nodes that are closer to each other on the graph are likely to be similar. Each node updates its tag information to that of the most tagged adjacent node. As the labels spread, densely connected clusters of nodes rapidly reach a unique label. At convergence, nodes with the same label are considered to belong to the same community.

The k-clique algorithm [35] works on the principle that if a graph, G, has a complete subgraph with k nodes, then this subgraph can be referred to as a k-clique. When there are k-1 common nodes between two k-cliques, they are considered “neighboring” cliques. A series of such neighboring cliques form a maximal set, which is referred to as a community.

ASOCCA is a community detection algorithm proposed in our previous work [36]. The main idea is to find the most similar neighboring nodes for each node, based on local clustering coefficients. These neighboring node pairs are combined to generate multiple sets of connectivity components, which form the basic list of community structures. Finally, the optimal community structure is determined by calculating and comparing the modularity Q-value of each community structure. The ASOCCA algorithm consists of the following five steps: (1) compute the clustering coefficients for each node in the network and generate a list of clustering coefficients, called CC_list; (2) traverse all adjacent nodes of each node, using CC_list to determine the local similarity of these adjacent nodes and generate a list of similarities called Sim_list; (3) generate a list of neighbor node pairs with the highest similarity based on Sim_list, referred to as Max_neighbors_list; (4) combine these node pairs based on Max_neighbors_list, generate all the possible sets, and extract the first 100 groups of each combination, filtering out any duplicate sets of connected components to complete the basic community structure list; and (5) traverse each set of community structures, merge small communities using the merging strategy and threshold, $\lambda$, calculate and compare Q-values, then return the final community structure.

3.3. Evaluation Indicators of Community Detection

Modularity Q is a classic community detection evaluation criterion [37], calculated by Formula (4). Modularity is used to evaluate the organization of networks or graphs. It quantifies how well a network can be divided into distinct modules, clusters, or groups. Networks with high modularity exhibit strong connections between nodes within each module and weak connections between nodes in different modules. Modularity is commonly employed in optimization algorithms for identifying community structure in networks. Therefore, this metric is also widely used as an objective function to measure the accuracy of various community detection algorithms.

$$Q=\frac{1}{2m}\sum _{ij}^{ }\left(a_{ij}-\frac{k_i{k}_j}{2m}\right){\delta }_{ij}$$

(4)

where i, j denote the indices of nodes in the network, and $A=\left\{a_{ij}\right\}$ represents the adjacency matrix of the network. If no connections exist between node i and node j, then $a_{ij}$ = 0; otherwise, $a_{ij}$ = 1. ${\delta }_{ij}$ represents whether node i and node j belong to the same community or not. If they do, then ${\delta }_{ij}$ = 1; otherwise, ${\delta }_{ij}$ = 0. The closer the Q value is to one, the better is the performance of the community detection algorithm.

4. Data Sources and Visual Statistical Analysis

The patent data used in this study were sourced from the IncoPat Global Patent Database (https://www.incopat.com, accessed on 5 January 2023). When searching for patents on hydrogen energy technology, a combination of the International Patent Classification Number (IPC) method and the keyword method was used, to avoid omissions.

According to the latest “Patent Classification System for Green and Low-Carbon Technologies” issued by the SIPO of China and 22 research reports, the keyword search formula and IPC search formula were determined. The main keywords and IPCs are listed in

Table 1. Main keywords and international patent classification numbers related to hydrogen energy.

Keyword

International Patent Classification Number

Hydrogen Energy, Hydrogen Source, Hydrogen Production, Hydrogen Transportation, Hydrogen Storage, Hydrogen Filling, Hydrogen Fuel, Low-Carbon Hydrogen, Zero-Carbon Hydrogen, Green Hydrogen, Blue Hydrogen, Purple Hydrogen, Powdered Hydrogen, Hydrogen Fueling Guns, Alkaline Electrolysis of Water, Proton Exchange Membranes, Anion Exchange Membranes, Solid Oxides, Hydrogen Recirculation System, Hydrogen Circulating Pumps, Hydrogen Supply, Hydrogen Systems, Hydrogen Logistics, Liquid Hydrogen, Gaseous Hydrogen, Solid Hydrogen, Hydrogen Finished Gas, Industrial Hydrogen By-Products, Propane Dehydrogenation By-Products, Chlor-Alkali By-Products, High Pressure Gaseous Hydrogen, Hydrogen Bottles, Hydrogen Tanks, Hydrogen Compressors, Hydrogen Cars, Hydrogen Electric, Hydrogen Heavy Duty Trucks, Hydrogen Doped Gas Pipeline, Pure Hydrogen Pipeline, Ortho- and Neutral-Hydrogen Conversion Catalysts, Bipolar Plates, Platinum Catalysts, Membrane Electrodes, Propane Dehydrogenation, Methanol Cracking, Pressure Swing Adsorption

F25J1/02, F25B17/12, C01B3/00, C10G1/06, C10G45/00, C10G47/00, C10G65/00, C10G67/00, C10G69/00, C07C7/163, C01B3/00, B01J27/24, C12P3/00, C25B1/00, C25B9/00, C25B15/00, C01B3/00, F17C, B22F, B82Y, C01B6/00, C22C, B01J20/00, C01B31/00, C01B32/00, C08G18/00, C08G59/00, C08G77/00, C08K3/00, C08K5/00, C08L63/00, C08L83/00, C08L101/00, H01M4/86, H01M8/00, H01M12/00

.

The patent granting time was limited to between January 1986 and January 2023, and the patent types selected were invention patents and utility models. The search results yielded a total of 76,760 registered effective patents for hydrogen energy technologies. Among them, there were 29,658 patents from Chinese applicants or from those who listed China as the patent country (dataset A), spanning the period from January 2003 to January 2023. The number of patents filed by China inter-institutional collaborations was 4095 (dataset B). Both datasets A and B were used as experimental data.

4.1. Statistical Analysis of Hydrogen Energy Patent Data

Preliminary statistical analyses were conducted for 29,658 patents in dataset A. Figure 1 illustrates the number of patent applications, granted patents, and cooperative patents for Chinese hydrogen energy from 2003 to 2023. The number of patent applications and grants has steadily increased since 2003, with a peak of 4690 applications in 2020 and a peak of 6220 grants in 2021. The number of patents for institutional cooperation has consistently increased over the years, surpassing 100 in 2012 and doubling to a record high of 705 in 2021 compared with 2020, demonstrating the strengthening of innovation cooperation in the hydrogen energy field. The date of data collection being in early 2023 and the delayed effect in the process of patent disclosure resulted in a decline in both 2022 and 2023.

Figure 2 shows the top 10 provincial administrative regions for the number of Chinese hydrogen energy patents. It can be observed that Beijing (5417), Jiangsu (3397), and Guangdong (2551) rank as the top three regions with the highest innovation activity. These ten provinces and cities account for 78.52% of all Chinese patents in the field of hydrogen energy.

The annual trend of patent applications in ten provinces and cities is illustrated in Figure 3. Initially, Beijing took the lead, but, from 2018 onwards, Jiangsu and Guangdong clearly caught up. This indicates the intense technological competition in China’s hydrogen energy industry market. The regional pattern of China’s hydrogen energy industry is undergoing change, as provincial and municipal governments focus more on layout planning and guidance, while enterprises continuously unleash their innovative vitality under market orientation. Notably, the central and western provinces, such as Sichuan, Hubei, and Shaanxi, are actively establishing their own competitive advantages for hydrogen energy industry development.

The distribution of patent innovation entities is illustrated in Figure 4. Enterprises constitute the majority of patent applicants, indicating their robust research and development (R&D) capabilities and emphasis on commercial applications within the hydrogen energy sector. Universities and research institutes exhibit the second highest level of R&D strength, displaying a greater inclination toward fundamental theoretical research. Furthermore, enterprises are more active in transforming technological innovation achievements into patents, thereby establishing their competitiveness within the industry. Figure 5 shows the number of cooperative patents between innovative entities from 2004 to 2023, which shows that inter-enterprise cooperation (E–E) far exceeds other types of cooperation. It can be concluded that enterprises occupy an extremely key position in the innovation of, and cooperation on, hydrogen energy technology.

The IPC represents the technological classification of patents. By analyzing IPC codes, we can identify technology hotspots and determine future development trends. According to IPC codes, the top 10 technology compositions of China’s hydrogen energy patents are shown in Figure 6, and their classification explanation is given in

Table 2. Top 10 IPC subclasses and their meanings for China’s hydrogen energy patents.

IPC Subclass	Meaning
H01M	Method or device for directly converting chemical energy into electric energy, such as a battery pack.
B01J	Chemical or physical methods, such as catalysis or colloid chemistry.
C01B	Non-metallic elements and their compounds.
C25B	Electrolytic or electrophoretic processes for producing compounds or nonmetals; the equipment used therein.
C10G	Hydrocarbon oil cracking; the preparation of liquid hydrocarbon mixtures through, e.g., destructive hydrogenation reactions, oligomerization reactions, and polymerization reactions; recovering hydrocarbon oil from oil shale, oil mines, or oil gas; refining hydrocarbon-based mixtures.
F17C	A container for holding or storing compressed, liquefied, or solidified gas; fixed-capacity gas storage tank.
C07C	Acyclic or carbocyclic compounds.
B01D	Separation methods.
C02F	Treatment of water, wastewater, sewage, or sludge.
F25J	Liquefaction, solidification, or separation of gas or gas mixtures through pressurization and cooling treatments.

. It can be observed that the hotspots of technological innovation in China’s hydrogen energy research field are concentrated in areas such as hydrogen fuel cells (H01M), chemical raw materials for hydrogen production (B01J, C10G, and C07C), electrolysis for hydrogen production (C01B and C25B), hydrogen storage and transportation tanks (F17C), the separation of hydrogen (B01D), the liquefaction and solidification of hydrogen (F25J), and methods for waste liquid and wastewater treatment (C02F).

5. Patent Collaborative Innovation Network

To analyze the collaborative innovation relationship in China’s hydrogen energy sector, the patent collaborative innovation network is constructed according to obtained dataset B. In this section, we first discuss the fundamental characteristics and evolution patterns of the constructed network. Subsequently, several important innovation entities are identified in the network. Then, based on the performance comparison of four community detection algorithms in the network, the ASOCCA algorithm is selected and executed to obtain the optimal community clusters (i.e., innovation alliances). Thereafter, the partnerships and future development trends of the innovation alliances are discussed. Finally, regional cooperation within this network is further explored and analyzed.

5.1. Construction and Analysis of the Patent Collaborative Innovation Network

The patent collaborative innovation network is divided into five sub-networks, based on time intervals. In the network, each node represents an innovative entity listed on the patent application, such as enterprises, universities, or research institutes. The size of a node corresponds to its number of neighbors, while a node with a larger size indicates that this entity has more collaboration partners. In addition, the color assigned to a node signifies the number of collaborations it engages in. We use a gradient color of red, yellow, and blue to show the difference in the number of collaborations from low to high. If multiple patent entities are involved in one patent application, it is considered a partnership and forms connecting edges between nodes. These connecting edges symbolize cooperation among patent applicants, and their thickness reflects the frequency (i.e., the intensity) of collaboration. The color of each edge is formed by blending and transitioning the colors of its two end nodes. To illustrate the collaborative relationships visually between innovation entities more intuitively, this paper employs Gephi to generate five sub-networks, along with an overall network, as shown in Figure 7. These networks are undirected weighted networks, because the collaborative innovation relationship is bilateral and cooperative.

In the first sub-network (from 2003 to 2005), there existed a limited scale with sporadic connections, which indicates that China’s hydrogen energy industry had not yet garnered significant attention. In the second stage (from 2006 to 2010), the sub-network remained in its early stages, with individual core nodes, and these nodes were loosely interconnected. In the third stage (from 2011 to 2015), the sub-network experienced an expansion in scale, with certain nodes occupying central positions within the network. Several innovative entities formed a fully connected component and demonstrated a phenomenon of group formation through multi-party collaboration. In the fourth stage (from 2016 to 2020), both the scale of the sub-network and the number of nodes visibly increased. The interconnections between nodes became stronger, while fixed collaboration among innovation entities further intensified. In the most recent stage (from 2021 to 2023), there was an additional increase in the number of nodes within the sub-network. Moreover, certain innovative entities exhibited a notable rise in their frequency of fixed collaboration, which indicates enhanced trust and tacit understanding among multiple innovative entities, fostering an atmosphere conducive to conglomerate innovation. Consequently, scientific and technological collaborations within the field of hydrogen energy flourished.

Table 3. Main topology properties of the patent collaborative innovation network, including the number of nodes (N), the number of edges (E), average degree (〈DC〉), average weighted degree of centrality (〈WDC〉), network diameter (D), average path length (L), network density ($\rho$), and average clustering coefficient (〈C〉).

Period	N	E	$〈\mathbf{DC}〉$	$〈\mathbf{WDC}〉$	D	L	$\mathit{\rho }$	$〈\mathit{C}〉$
2003–2005	11	8	1.455	6.909	2	1.273	0.145	0.583
2006–2010	57	55	1.93	14.807	4	2.055	0.034	0.517
2011–2015	231	205	1.775	8.987	5	2.369	0.008	0.653
2016–2020	559	487	1.742	6.286	10	4.016	0.003	0.608
2021–2023	701	642	1.832	12.762	10	3.808	0.003	0.695
2003–2023	1105	1097	1.986	13.987	14	5.420	0.002	0.638

shows the basic characteristics of the patent collaborative innovation network during each period from 2003 to 2023. The number of nodes witnessed a remarkable increase from 11 to 1105 over the past two decades; a hundredfold surge. This signifies an escalating scale of innovation entities within the hydrogen energy domain and an evident expansion rate. Concurrently, the number of edges experienced a corresponding approximate hundredfold rise, indicating a significant augmentation in opportunities for collaboration among innovation entities and fostering closer interconnections. The network density, on the other hand, gradually decreased, suggesting that the number of new partnerships formed between innovation entities lagged behind the influx of new nodes into the network. Consequently, ample room for cooperation remains untapped among innovation entities. The average path length of the network continues to grow, indicating that the paths for establishing cooperative relationships among innovation entities have been extended over time. The ability of collaborative exchange and knowledge information transfer between innovation entities has increased, and there are still some potential opportunities for cooperation to be explored. The increase in the average clustering coefficient indicates a high level of aggregation and closer ties among innovation entities. Overall, with the expanding scale of the patent collaborative innovation network, the innovation collaboration capacity has been increasing. There are more opportunities and space for collaboration among innovation entities.

5.2. Analysis of Important Nodes in the Patent Collaborative Innovation Network

In this subsection, we aim to identify several important innovation entities in the patent collaborative innovation network. As discussed in Section 3, several centrality metrics have been proposed to evaluate the importance of nodes in the network. Based on these indicators,

Table 4. Top 10 innovation entities in China’s hydrogen energy field.

Ranking	DC	WDC	BC	CC
1	China Petrochemical Co., Ltd.	China Petrochemical Co., Ltd.	Dalian University of Technology	Zhejiang Geely Holding Group Co., Ltd
2	State Grid Co., Ltd.	China Huaneng Group Clean Energy Technology Research Institute Co., Ltd.	China Petrochemical Co., Ltd.	Beijing Yuanda Xinda Technology Co., Ltd.
3	Tsinghua University	Sichuan Huaneng Hydrogen Energy Technology Co., Ltd.	State Grid Co., Ltd.	China Academy of Sciences Institute of Process Engineering
4	State Grid Corporation	Sichuan Huaneng Fujiang Hydropower Co., Ltd.	Tsinghua University	Nanjing University of Science and Technology
5	Dalian University of Technology	Huaneng Mingtai Electric Power Co., Ltd.	Chinese Academy of Sciences Dalian Institute of Chemical Physics	China Academy of Launch Vehicle Technology
6	Global Energy Internet Research Institute Co., Ltd.	Sichuan Huaneng Kangding Hydropower Co., Ltd.	Global Energy Internet Research Institute Co., Ltd.	China Huadian Science and Industry Group Co., Ltd.
7	China Huaneng Group Clean Energy Technology Research Institute Co., Ltd.	Sinopec Refining and Chemical Engineering (Group) Co., Ltd.	State Grid Zhejiang Electric Power Co., Ltd. and Electric Power Science Research Institute	China International Shipping Container (Group) Co., Ltd.
8	Chinese Academy of Sciences Dalian Institute of Chemical Physics	Shenhua Group Co., Ltd.	Wuhan Institute of Technology	Beijing Institute of Technology
9	South China University of Technology	Tsinghua University	South China University of Technology	Beijing China Hydrogen Green Energy Technology Co., Ltd.
10	Sinopec Corp.	State Grid Co., Ltd.	Shanxi Luan Mining Industry (Group) Co., Ltd.	Dongguan Institute of Technology

shows the top-ranked important nodes in the patent collaborative innovation network.

China Petrochemical Co., Ltd. holds a prominent position in the hydrogen energy field, ranking first in two out of four indicators and second in one. Its main patent contributions lie in hydrogen production from chemical raw materials. The corporation is a central hub node in the network, collaborating with numerous domestic innovation entities and establishing multiple partnerships. It dominated several cooperative companies and had great influence and control over collaborative innovation and R&D in the field of hydrogen energy.

Other entities that rank in the top 10 in several indicators include China Huaneng Group Clean Energy Technology Research Institute Co. Ltd., State Grid Co., Ltd., Tsinghua University, and so on. These entities have played a vital role in promoting technological innovation because they have control over many resources. China Huaneng Group frequently collaborates with many companies in the Sichuan Huaneng group, which is reflected in their WDC ranking. Their collaborative innovation focuses on hydrogen production via the electrolysis of water. The innovation of hydrogen energy technology at the Chinese Academy of Sciences Dalian Institute of Chemical Physics focuses on electrolytic water and chemical raw materials for hydrogen production, as well as hydrogen fuel power battery systems and transportation applications. Tsinghua University’s technological innovation mainly involves hydrogen fuel power battery systems and hydrogen production via the electrolysis of water, while South China University of Technology mainly specializes in hydrogen production using chemical raw materials. The patent advantage of Zhejiang Geely Holding Group Co., Ltd. lies in hydrogen fuel power battery systems and transportation applications, and it holds the highest CC indicator, indicating its key position in resource transfers and fruitful knowledge exchange. Notably, the companies of the Sichuan Huaneng group have a high WDC, but for the other three indicators they are not in the top 10, indicating that, although the cooperation frequency is high, their collaboration partners are relatively fixed and limited.

To summarize, these innovative entities are important to China’s hydrogen energy industry, as they help to maintain the function of the network and promote the sustainable growth of the hydrogen energy industry in China.

5.3. Community Detection of Patent Collaborative Innovation Network

For the patent collaborative innovation network, community detection is the process of identifying multiple innovation entities with similar attributes and close relationships in the network to form a cluster, thereby dividing the entire network into multiple clusters. We employed four classic community detection algorithms (the GN algorithm, the LPA, the k-clique algorithm, and ASOCCA) to detect community clusters (i.e., innovation alliances) within the network. The modularity Q [37] serves as a performance evaluation metric for community detection.

Figure 8 illustrates the Q value and the number of communities using the four algorithms. Higher values of Q indicate better community structures. A general network with a Q value between 0.3 and 0.7 indicates a good community structure [38]. The performances of our previously proposed algorithm (ASOCCA) and of the GN algorithm are better, as they achieve the best and the second-best Q values of 0.911 and 0.905, respectively. The LPA and the k-clique algorithm exhibit slightly inferior performances, as their Q values are between 0.85 and 0.90.

Based on this experimental comparison, we employed the results of community detection with the ASOCCA algorithm to identify the alliances of cooperation among patent innovation entities, which provided a theoretical basis for judging and analyzing the cooperation relationships and cooperation trends between innovation entities.

5.3.1. Global Sensitivity and Uncertainty Analysis

The results of community detection depend on both the data and the algorithm used. Variations in the results can be attributed to changes in algorithm parameters or input data. In this subsection, we used complete patent data as our dataset, so only variations in the community detection algorithm will affect the result. The ASOCCA algorithm has two input parameters that impact its performance (i.e., the Q value). The first input parameter is the minimum number of nodes required for a community (referred to as the threshold), while the second one is the merging strategy (referred to as the m-strategy) used during community formation, which is one of the following strategies: clustering coefficient–clustering coefficient, degree–degree, clustering coefficient–degree, or degree–clustering coefficient [36]. To investigate how changes in these two input factors influence the result, we employed global sensitivity and uncertainty analysis (GSUA) [39]. For this purpose, we utilized an open-source software called the Sensitivity Analysis For Everyone (SAFE) toolbox [40], which incorporates a set of sampling, sensitivity analyzing, and visualization methods.

First, we performed a one-at-a-time (OAT) sensitivity analysis using the elementary effects test (EET) method [41]. The parameter ranges were conducted using OAT sampling, with the Latin hypercube sampling (LHS) strategy and a trajectory design type. We set the number of elementary effects, r, to 200, resulting in 600 algorithm executions (r × (M + 1), where M is the number of input factors). Each execution returned a Q value for the current input factor configuration. Subsequently, we employed the EET method to analyze the sensitivity of two input parameters, which is depicted in Figure 9a. The global sensitivity, represented by the mean of elementary effects (EEs), was computed as 0.2362 for “threshold” and 0.1109 for “M-strategy”. This indicates that algorithm performance (i.e., Q value) was more influenced by the input parameter “threshold”.

Second, we used Monte Carlo simulations and the PAWN (derived from the authors’ names) method [42] to perform a global sensitivity and uncertainty analysis. We ran Monte Carlo simulations of the community detections against a prescribed number of input factor combinations, randomly drawn from the feasible parameter space using an all-at-a-time (AAT) sampling process. We set the sampling strategy to Latin hypercube sampling (LHS) and the number of samples to 500. Each algorithm execution provided a Q value, called Q-sim, and we measured the distance from the theoretical upper bound of the Q value, which is one, for each Q-sim. We defined the distance (1 – Q-sim) as a performance metric called BIAS. We set the number of conditioning intervals to access the conditional cumulative distribution functions (CDFs) to five, and the number of bootstrap resamples to derive confidence intervals to 500. Finally, we used the PAWN method to compute the Kolmogorov–Smirnov (KS) statistic between the empirical unconditional CDF and the conditional CDFs for different conditioning intervals. As shown in Figure 9b, the sensitivity (mean KS) of the “threshold” and “m-strategy” were 0.3654 and 0.2776, respectively. This result was consistent with that of the EET method, indicating that the algorithm performance is more sensitive to the variation of “threshold”. Based on this analysis and additional experiments, we derived the optimal configuration of algorithm parameters.

5.3.2. Communities of the Patent Collaborative Innovation Network

We utilized the ASOCCA algorithm to identify the communities within the patent collaborative innovation network. To describe the results visually, we used Gephi to draw the community structure of the patent collaborative innovation network after community detection, as shown in Figure 10. Only node IDs are displayed, while the entity names (node labels) have been omitted. There are 119 communities in the network, with each community being visually represented by nodes sharing the same color.

5.4. Collaborative Analysis of Innovation Alliances

In this subsection, we analyze several types of innovation alliances (i.e., community clusters), including the largest alliance, core alliance, and special alliance for full connectivity. To explore the technical hotspots of collaboration, we can analyze the key technology distribution of some core entities in these innovation alliances. As shown in Figure 11, collaborative technologies are focused on the development of hydrogen fuel cells (H01M), chemical raw materials for hydrogen production (B01J, C10G, and C07C), electrolysis for hydrogen production (C01B and C25B), and hydrogen storage and transportation tanks (F17C).

In the field of hydrogen energy innovation, many innovative entities tend to focus on the same technologies and often seek cooperation or assistance on a particular technology. Thus the technology dependency relationships between the collaborating entities are formed. Technology dependencies in the collaboration of innovative entities are crucial for accurately assessing the systemic importance of technologies and representing the subordination between innovative entities. As a result, Figure 12 shows a technology dependency map for the major innovation entities of alliances.

5.4.1. Largest Alliance

The largest alliance, A, includes State Grid Co., Ltd. (201), State Grid Corporation (801), Tsinghua University (654), and Global Energy Internet Research Institute Co., Ltd. (141) as its core nodes. Additionally, Alliance A contains a total of 70 innovation entities, such as Xi’an Jiaotong University. State Grid Co., Ltd. ranks fifth in the Top 100 Global Hydrogen Energy Industry Invention Patent Rankings 2022 (hereinafter referred to as the Hydrogen Top 100). The direction of technological innovation of State Grid encompasses various aspects of the hydrogen energy industry, including hydrogen fuel cells, storage power stations, generation systems, hydrogen refueling, and production. The reason for this is that State Grid and many regional subsidiaries are themselves large in function, with extensive business developments and collaborations with numerous innovation entities. Among these is Tsinghua University, whose main collaborative innovation directions include hydrogen fuel cells and hydrogen production from electrolytic water (H01M, C01B, and C25B), which is exactly the same as Xi’an Jiaotong University. The main areas of collaborative innovation for Global Energy Internet Research Institute Co. Ltd. include hydrogen energy storage systems, fuel cells, hydrogen production, and power conversion control.

5.4.2. Core Alliances

Alliance B is centered on China Petrochemical Co., Ltd. (1025), and consists of 41 innovation entities. China Petrochemical Co., Ltd., a subsidiary of the Petrochemical group, ranks first in the Hydrogen Top 100. Its collaborative innovations are mainly focused on hydrogen production from chemical raw materials (B01J, C10G, and C07C) in the hydrogen energy industry chain. According to the statistics of the Petroleum and Chemical Industry Planning Institute, China is currently the largest hydrogen producer in the world. Hydrogen production from chemical raw materials is a highly mature method, and Alliance B, led by China Petro-chemical Co. Ltd., has the most advantages in this area. Its main partners are Sinopec Refinery Engineering (Group) Co., Ltd., and a series of subsidiaries of China Petrochemical Co., Ltd., such as China Petroleum & Chemical Corporation Petrochemical Research Institute, China Petrochemical Co., Ltd., and Fushun Petrochemical Research Institute.

Alliance C is centered on the Chinese Academy of Sciences Dalian Institute of Chemical Physics (636), and contains 18 innovation entities, such as the University of Petroleum (East China), Shaanxi University of Science and Technology, China Petroleum and Natural Gas Co., Ltd., and Shanghai Automotive Industry Corporation (SAIC). As shown in Section 5.2, the Chinese Academy of Sciences Dalian Institute of Chemical Physics possesses comprehensive hydrogen energy technologies (H01M, B01J, C01B, C25B, and C07C), with each technological direction ranking at the forefront domestically, and significantly influences the development of hydrogen energy technology. Within Alliance C, there is strong cooperative cohesion, leading many universities and enterprises to engage in joint innovation.

Alliance D, with Dalian University of Technology (83) as the core, includes 18 innovation entities, such as Dalian University, Jiangsu University, Nanjing University of Technology, Shenyang University of Technology, Beijing University of Architecture, and He’nan Hydrogen Source Technology Co., Ltd. Dalian University of Technology primarily focuses on hydrogen fuel cell-related technology research, and it has led and acted as the ‘glue’ for the collaboration and joint innovation between enterprises and universities.

The relationship between Alliance C and Alliance D is close, and the important connection between them is the two core nodes of the Chinese Academy of Sciences Dalian Institute of Chemical Physics, and Dalian University of Technology. These two innovative entities have the geographical advantage of innate cooperation and frequently collaborate on hydrogen fuel cell research and development. At present, there are not many cooperation institutions and joint projects between Alliances C and D. In the future, these two alliances could expand cooperation opportunities through the above two core nodes and try to cooperate in the field of hydrogen production from chemical raw materials.

Alliance E has the South China University of Technology (192) as its core node, and comprises 19 innovation entities, including China University of Petroleum (Beijing) and China Offshore Oil Group Co., Ltd (CNOOC). It is characterized by geographical alliances and primarily focuses on collaboration in hydrogen production (B01J, C10G, C07C, C01B, and C25B). This synergy enables Guangdong’s innovation entities to exploit their geographical proximity to the sea fully, promoting collaboration among prominent entities such as CNOOC, the PetroChina group’s companies, Guangzhou Automobile Group, and Foshan Automobile Gas Company.

5.4.3. Special Alliance for Full Connectivity

Alliance F stands out as a unique community cluster in the collaborative innovation network, containing 18 innovation entities such as China Huaneng Group Corporation, China Huaneng Group Clean Energy Technology Research Institute Co. Ltd, Huaneng Group Technology Innovation Centre, and a series of subsidiaries of Sichuan Huaneng. These multiple innovation entities form an all-connected component, i.e., the community nodes are interconnected with each other entirely. This is common in collaborative innovation networks; in this case, this component emerged in the 2011–2015 sub-network and further increased in subsequent phases. The final overall network comprises approximately six or seven fully connected or nearly fully connected components, among which Alliance F is the most typical. As can be seen from the structure, there is no core node in F; instead, each node tightly clusters with others. However, there is a concern that Alliance F’s technological innovation direction solely focuses on water electrolysis for hydrogen production, neglecting other areas. The worry arises from the high frequency of cooperation between partners in Alliance F, leading to an overly restricted innovation direction, due to fixed partnerships. This limitation may hinder the long-term development of hydrogen technology.

5.5. Analysis of Regional Cooperation Innovation

After analyzing the innovative entities of the patent collaborative innovation network, it is essential to explore regional cooperation within this network further. There is spatial regional interdependence of hydrogen energy technologies among Chinese provinces, and spatial regional networks can reveal potential ecological and environmental mechanisms in the innovation ecology of the hydrogen energy industry [43,44]. We extracted the region information where innovative entities are located in the network and formed a region-based collaborative evolutionary process, corresponding to Figure 7. The spatial and temporal evolution of regional cooperation is shown in Figure 13, where the granularity of regional refinement is China’s 34 provinces and municipalities directly under the central government. It is noted that the cooperation records within the same region are not considered here.

In the early stage (a, b), Beijing, as the most core region, collaborated with multiple areas on hydrogen energy. In the middle stage (c, d), more and more core regions emerged, including Liaoning, Shanghai, Guangdong, and Jiangsu. Recently (e), Sichuan, Shaanxi, Shandong, and Tianjin have also been actively involved in cooperation. As can be seen from the above evolution, innovative entities in developed provinces and cities such as Beijing, Shanghai, Zhejiang, Jiangsu, and Guangdong have established frequent and efficient inter-provincial collaborations. Innovative entities in Sichuan, Shaanxi, Liaoning, and Shandong have experienced a gradual increase in the number of inter-provincial cooperation opportunities, to enhance their competitive advantages. From the overall perspective of regional cooperation (f), it can be concluded that Beijing and Sichuan, as well as Beijing and Liaoning, are the most frequent regional partners in collaboration. In addition, a cluster pattern of ecological development in the hydrogen energy industry has gradually formed at the regional level.

Table 5. Characteristics of industrial development in key regions.

Regions	Representative Provinces and Cities	Characteristics
Yangtze River Delta	Shanghai and Jiangsu	The region has the highest number of cities and a concentration of universities, possessing abundant experience in research and demonstration of fuel cell electric vehicles.
Pearl River Delta	Guangzhou, Foshan, and Shenzhen	Three major hydrogen fuel cell vehicle innovation zones have been established, with a leading hydrogen refueling network plan nationwide.
Bohai Rim	Beijing and Liaoning	It possesses the conditions for the whole hydrogen energy industry chain, with hydrogen energy application demonstration projects in two major fields: transport and steel.
Northwest, Southwest	Inner Mongolia, Sichuan, Chongqing, and Yunnan	They are important areas in China for the research and development of renewable energy hydrogen production and fuel cell electric propulsion. They undertake the task of achieving large-scale low-cost hydrogen production and promoting the integration of hydrogen production from renewable energy and hydrogen energy storage.

illustrates the current characteristics of key regions in hydrogen energy industry development.

Currently, a number of provinces’ and municipalities’ governments have formulated the path and mode for hydrogen energy development, issuing many policy documents [5,6,8,9,45,46] and a series of investment and incentive mechanisms. For instance, Beijing’s Hydrogen Industry Development Plan (2021–2025) aims to construct 37 hydrogen refueling stations and produce a total of 10,000 fuel cell vehicles. Similarly, Shanghai’s Implementation Plan for New Energy Vehicle Industry (2021–2025) plans to manufacture 10,000 fuel cell vehicles and build over 70 different types of hydrogen refueling stations. Guangdong Province Action Plan for Accelerating the Construction of Fuel Cell Vehicle Demonstration City Clusters (2022–2025) plans to invest in the manufacture of 10,000 fuel cell vehicles and establish 200 hydrogen refueling stations. Sichuan Province’s Hydrogen Industry Development Plan (2021–2025) intends to invest in hydrogen infrastructure support systems, create 6000 fuel cell vehicles (including heavy trucks, medium and light logistics vehicles, and buses), and establish 60 various types of hydrogen refueling stations. These policies are highly homogenized, and the supporting policy details are still lacking. In terms of policy incentives, financial incentives are mainly adopted. Supply-based incentive policies are mainly implemented at the provincial and municipal levels, providing subsidies, preferential electricity prices, tax benefits, and other policies for hydrogen energy-related enterprises. Single financial incentives are not enough to promote the industry, and the incentive policy system needs to be further improved.

In regions with a strong foundation for collaborative innovation, it is recommended that the hydrogen energy industry is integrated with expert resources, and that the establishment of technical standards is promoted through industry groups. The government should establish a high-quality R&D platform dedicated to the hydrogen energy industry, promote the high-level rollout of related products and patents, and formulate R&D lists and standards for hydrogen energy products. Additionally, the government should enhance the recruitment of experts in hydrogen energy. A diverse incentive system could be implemented to provide favorable support facilities for experts, including competitive salaries, relaxed household registration, housing, and education for their children.

6. Discussion

Under the dual-carbon policy, China’s hydrogen energy industry is attracting wide attention from state-owned enterprises, financial institutions, and private companies. In the future, the Chinese hydrogen energy industry will develop towards technological localization, clean hydrogen production, and diversified hydrogen utilization.

The transition to a hydrogen economy is an important issue for practitioners and society. The development of hydrogen energy involves multiple stakeholders, including industry, academia, government, and energy departments. The government sets strategic directions and coordinates overall efforts while the energy department commands cooperation. Industry and academia implement specific actions, generate output, and drive technological innovation. The financial department invests in industries and academia to provide funding support. As mentioned in Section 1, technological innovation is crucial for the hydrogen energy industry. China advocates for domestic technology research and development, requiring close collaboration between industry and academia, as well as policy support from the government and energy sectors. In Section 5.2, a list of the top 10 influential innovative entities in both industry and academia is provided (Table 4), along with an analysis of their respective expertise in different technological areas within the supply chain. Section 5.3 explores collaborative alliances among these innovative entities in both sectors (Figure 10), analyzing the technical directions of core entities and the codependency in each alliance (Figure 11 and Figure 12). Policymakers and practitioners can refer to the analysis and summary.

The hydrogen supply chain incorporates a series of stages from production and supply to customer usage, including energy selection, production, storage, transportation, and distribution, as well as application [47,48,49]. Multiple links along the entire hydrogen supply chain interact with each other, as shown in Figure 14. These individual links are integrated to form a national supply chain network. We present the distribution of key technologies in the hydrogen sector in Figure 6 and Table 4. These innovative entities and their associated key technologies have significant impacts on the hydrogen supply chain. Moreover, the network of the supply chain largely depends on specific regional circumstances. According to the analysis of the regional distribution of hydrogen technology innovation in Section 4.1, and the regional collaborative innovation in Section 5.5, the collaboration between regions is strengthening. This leads to a deeper division of work within the hydrogen supply chain, resulting in clearer industrial development positioning for each region, as listed in Table 5. The core modules, including hydrogen production, storage and transportation, and fuel cell vehicles, are forming industry clusters.

In summary, the deployment and development of hydrogen supply chains necessitates careful consideration of various aspects:

(1)

In the hydrogen production stage, China mainly relies on fossil materials and industrial by-products to produce gray hydrogen and blue hydrogen. The scale of green hydrogen produced through the electrolysis of water is still relatively small. It is necessary to increase the deployment of renewable energy for green hydrogen production.

(2)

In the storage and transportation stage, high-pressure gaseous and liquid hydrogen storage and transportation technologies are relatively mature. However, transporting hydrogen in gas cylinders, tank trucks, or tankers has limitations due to their small volume, high cost, and safety risks. Solid-state storage technology is currently not mature. Therefore, a better approach would be to deploy hydrogen pipeline infrastructure.

(3)

In the distribution stage, China has the largest number of refueling stations globally, with a regional concentration feature. Currently, many provinces and cities have actively planned to deploy hydrogen stations. When deploying these stations, attention should be paid to their types and locations.

(4)

In the application stage, some key materials of fuel cells, such as solid oxide and anion exchange membranes, still rely on imports. Japan and South Korea have relatively mature hydrogen fuel cell vehicle technology. The technological limitations, cost, and scale of Chinese fuel cell vehicles restrict their commercialization, with higher purchasing costs and a lack of complete commercial capability. In this stage of deployment, emphasis should be placed on technological innovation and breakthroughs while relying on government subsidies and policy support. Therefore, hydrogen fuel cells and vehicles are a key focus in investment planning by provincial governments.

(5)

Each region prioritizes the development of a regional labor division. As the industry further matures, regions will be connected through infrastructure such as hydrogen pipelines to form a nationwide network that spans from nearby to distant areas.

7. Conclusions and Proposals

In this paper, a patent collaborative innovation network based on hydrogen energy patents in China from 2003 to 2023 was constructed, and the topological characteristics and evolutionary process of the network were investigated in five stages. Then, the core innovative entities and their influence in the network were identified through node centrality measures, and the collaboration density and direction among these entities were analyzed. The innovation alliances in the network were identified using the ASOCCA algorithm that we proposed in previous work. Subsequently, the collaborative relationships between the core entities and other entities in the innovation alliances were analyzed and discussed. Finally, a discussion and summary from the perspective of the hydrogen supply chain was given. Based on the above research, this paper draws the following conclusions and proposals.

China has raised the development of hydrogen energy to a strategic position. Hydrogen energy-related technology patents have shown explosive growth since 2016, and the number of patents will continue to increase at a high rate in the coming period. Hydrogen energy industry technology collaborative innovation has experienced an evolution from initial, multi-center, and community clusters, and the development resilience has gradually increased, basically forming an interactive pattern of upwards-moving competition. The other conclusions made from the above research are as follows:

(1)

The regional innovation development of the hydrogen energy industry presents a “3 + X” distribution, with Beijing, Jiangsu, and Guangdong in the first echelon, followed by Shanghai, Zhejiang, Liaoning, and other provinces and cities. Among them, Sichuan, Shandong, Tianjin, Shaanxi, Henan, and Shanxi are catching up. Regarding regional collaboration, five key areas have emerged: the Yangtze River Delta (represented by Shanghai and Jiangsu), the Pearl River Delta (centered around Guangzhou and Shenzhen), the Bohai Rim (centered around Beijing and Liaoning), the Northwest (represented by Inner Mongolia and Shaanxi), and the Southwest (represented by Sichuan).

(2)

Enterprises with strong R&D strength in hydrogen energy that are focusing on commercial applications file the main body of patent applications, accounting for the largest proportion. China Petro-chemical Co., Ltd., State Grid Co., Ltd., China Huaneng Group, Global Energy Internet Research Institute Co., Ltd., and other entities have played a key role in promoting technological innovations. Universities and research institutes have successively set up hydrogen energy research centers and have increased their research investment. Tsinghua University, South China University of Technology, Dalian University of Technology, Xi’an Jiaotong University, and Zhejiang University are at the forefront of Chinese universities in hydrogen energy technology research. Chinese Academy of Sciences Dalian Institute of Chemical Physics leads the pack of research institutes.

(3)

The collaboration innovation in the field of hydrogen energy has evolved remarkably over time, with increasing intensity and developmental resilience. The collaboration innovation network has grown in scale, diameter, and clustering coefficients, indicating that it has become stronger and more stable. During the years of 2003–2005, the network was in its early stages, while 2006–2010 saw the beginning of the network prototype. From 2011 to 2015, the sub-network scale grew significantly, and, between 2016 and 2020, there was a significant increase in the number of collaborations between innovation entities. Between 2021 and 2023, fixed collaborations among innovative entities in the sub-network increased significantly, forming a group-type innovation atmosphere. Scientific and technological collaboration in the field of hydrogen energy is flourishing.

There are still some obvious shortcomings in industrial technology cooperation and innovation in the field of hydrogen energy in China. To improve the technological progress of the hydrogen energy industry, it is essential to optimize the industrial layout and enhance the level of the multi-principal and cross-regional division of labor and technology cooperation.

(1)

It is necessary to promote collaborative innovation among industry, universities, and research institutes further. The proportion of cooperation between enterprises and colleges/universities (research institutes) is still relatively low. From the results of community detection, it is evident that enterprises tend to cooperate with a series of subsidiaries within the group. They also tend to carry out patent cooperation activities with institutions that have the same or similar geographic locations. This tendency may be due to the consideration of cooperation costs and risks. However, it leads to the fixation on the object of cooperation and limits the direction of cooperation.

(2)

Most of the innovation alliances cooperate only in a narrow range of technologies and limited directions, which may hinder the overall sustainable growth of the hydrogen energy industry. From a policy perspective, it is crucial to introduce incentives to encourage entities within the alliance to carry out cooperative research and development in other technological directions. Additionally, entities between alliances should be encouraged to cooperate and seek the help of technologically mature enterprises in areas where they lack expertise. This will help to optimize the allocation of resources and fully leverage the technical advantages of each institution.

(3)

The upstream hydrogen production technology of the industry chain is relatively mature. However, research efforts need to be increased for hydrogen storage and transportation, midstream hydrogen fuel cells, and hydrogen power generation to reduce costs and promote industrialization. The technological development level in downstream application areas of the industrial chain is relatively low. Therefore, the government should guide powerful enterprise entities to increase the application and R&D of downstream products and help enterprises obtain high profits. As a positive demonstration, it will lead more innovative entities to enter the downstream track and promote the development of the whole chain of the hydrogen energy industry.

There are some key research directions in the field of hydrogen energy that need to be further explored in the future, as follows: the production, storage, and transportation of green hydrogen; solid-state hydrogen storage and transport; gaseous hydrogen transport pipelines; hydrogen refueling technology; and hydrogen fuel cells and vehicles. Renewable-energy-based hydrogen production is a long-term direction for achieving true zero carbon emissions. Large-scale investment in green hydrogen production and storage is imperative. Solid-state hydrogen that uses metal hydrides, chemical hydrides, or nanomaterials as carriers offers advantages such as a high storage density, good safety performance, and the purity of the stored hydrogen. This is a promising research direction, but it is still in the theoretical research stage, both domestically and internationally. Strengthening pipeline construction for transporting gaseous hydrogen between production plants, refueling stations, and end-users offers advantages such as a large transport capacity over long distances with low energy consumption losses. Hydrogen refueling mainly relies on hydrogen stations, which require strengthened construction and even full coverage. However, the construction costs for these are expensive. Fuel cell vehicles have technological characteristics such as a long range, fast refueling, a high power density, and low-temperature self-starting, making them the most promising research direction. The development of hydrogen stations and hydrogen fuel cell vehicles requires strong support and investment from local governments, which is also a focal point in the regional hydrogen energy industry.