Marine three-dimensional space high-potential area identification layout method based on suitability evaluation

Abstract

The invention discloses an ocean three-dimensional space high-potential area identification layout method based on suitability evaluation, and relates to the technical field of ocean space planning management, and the method comprises the steps: collecting sea area natural, social and ecological data and historical three-dimensional right confirmation project cases, carrying out the preprocessing, and constructing a sea area three-dimensional space basic database; analyzing a historical three-dimensional right confirmation case based on a graph neural network, and constructing a sea activity-space layer relation graph; and analyzing the compatibility of different sea-using activity combinations in historical projects by using the sea-using activity-spatial layer relation graph. According to the method, a multi-dimensional suitability evaluation system and a compatibility probability matrix are constructed, a high-potential area meeting natural conditions, ecological constraints and social requirements at the same time can be recognized, plane conflicts of traditional two-dimensional planning are avoided through layered weighted stacking analysis, vertical space resource allocation is optimized through a three-dimensional layered utilization rule, and the compatibility of the high-potential area is improved. Therefore, the same sea area can be compatible with various sea-using activities, and low-efficiency or conflicting layout is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of marine space planning and management, and in particular to a method for identifying and laying out marine three-dimensional high-potential areas based on suitability evaluation. Background Art

[0002] With the continuous growth of global ocean development and utilization, the scarcity of marine space resources has become increasingly prominent, and conflicts over sea use have continued to intensify. Nearshore waters have gathered almost all types of sea use activities, such as sea enclosure aquaculture, port shipping, tourism and leisure, resulting in high spatial utilization intensity, complex types, many potential conflicts, and great management challenges. For example, in the same vertical sea area, shipping or offshore wind power construction may be carried out on the water surface, the middle layer of the water body is suitable for the development of three-dimensional aquaculture, and the bottom layer may be used for laying submarine pipelines or mineral resource exploration. The mutual interference of multi-level sea use activities has become the main bottleneck restricting the efficient use of sea space. By reasonably dividing the vertical space of the sea area, the three-dimensional layered utilization model provides a mechanism for sharing resources and coordinated development for different types of sea use activities, which helps to improve the efficiency of sea space utilization and reduce the difficulty and management cost of sea area use demonstration and stakeholder coordination.

[0003] In the existing technology, overlapping areas are often identified as high-potential areas through overlay analysis, but the temporal and spatial characteristics of conflicts are not taken into consideration, which can easily lead to frequent contradictions when planning schemes are implemented. Therefore, how to construct a sea use activity-spatial layer relationship map based on graph neural networks, predict the compatibility probability of new sea use combinations, and recommend optimized layouts to ensure the accuracy of the identification and layout of high-potential areas in the marine three-dimensional space is the problem to be solved by the present invention. To this end, a method for identifying and laying out high-potential areas in the marine three-dimensional space based on suitability evaluation is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation, so as to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation includes the following steps: S1. Collect natural, social, and ecological data of marine areas and historical three-dimensional property rights confirmation project cases, perform cleaning, conversion, and standardization pre-processing, unify the coordinate system and accuracy standards, and build a three-dimensional spatial basic database of marine areas; S2. Analyze historical three-dimensional land rights confirmation cases based on graph neural networks, using sea use activities and spatial layers as nodes and relationships in historical cases as edges to construct a sea use activity-spatial layer relationship map and quantify compatibility probability; S3. Analyze the compatibility of different combinations of sea-use activities in historical projects using the sea-use activity-spatial layer relationship map and extract compatibility and conflict patterns; S4. Combine historical patterns and use graph neural network models to predict the compatibility probability of new combinations of maritime activities and identify potential conflicts. S5. A hierarchical analysis method is used to determine weights and conduct a hierarchical suitability evaluation of sea use activities. This generates a scoring map and, combined with the compatibility probability prediction results, categorizes high, medium, and low potential areas, prioritizing recommendations for three-dimensional layouts with low conflict probabilities. S6. Based on high-potential areas, recommend optimized layout plans to match the layout requirements of high-potential areas.

[0006] A further improvement of the technical solution of the present invention is that: S1 specifically includes: Collect three categories of data on the target sea area: natural, social, and ecological. Natural data includes water depth, bottom type, and marine disaster risk; social data covers transportation networks, port location conditions, and tourism supporting facilities; and ecological data covers marine ecological protection red lines, seawater quality, and biological resource distribution. At the same time, collect case data on historical three-dimensional land rights confirmation projects, including types of sea use activities, spatial distribution, conflict records, and solutions. Clean the collected raw data, including format unification, outlier correction, and missing value filling. At the same time, perform data conversion to unify data in different formats into a format that is easy to process, and perform standardization to unify the data's measurement units and classification standards to ensure that all spatial data are in the same coordinate system and that the accuracy meets project requirements. This eliminates errors and conflicts caused by inconsistent data formats and standards, and improves data quality and consistency. Based on the processed data, the architecture of the basic database of the three-dimensional space of the sea area is designed, and the core table structure and field definition are determined. The core table structure of the database includes the basic information table of the sea area, the natural data table, the social data table, the ecological data table and the historical case table. The processed data is imported into the database according to the designed architecture, and the data entry verification is performed to ensure that the data is stored accurately. At the same time, a data index and query mechanism is established to form a basic database of the three-dimensional space of the sea area.

[0007] A further improvement of the technical solution of the present invention is that: S2 specifically includes: Based on historical three-dimensional rights confirmation cases, the types of sea use activities and spatial layers (water surface, water body, seabed, and subsoil) are defined as two types of nodes in the graph. Node attributes include the type of sea use activity, spatial scope, temporal characteristics, and the physical and ecological attributes of the spatial layer. The interactive relationships in historical three-dimensional rights confirmation cases are used as edges, and edge attributes record the intensity of conflicts. The attribute encoding format of nodes and edges is unified, and topological verification is used to ensure the connectivity of the graph. Each sea use activity must be associated with at least one spatial layer, and spatial layer nodes must be indirectly connected through activity nodes. Isolated nodes or redundant edges are eliminated to form a structured initial graph. A graph neural network (graph convolutional network) is used to train the initial graph to learn the latent feature representations of nodes and edges. The input is the initial node features and graph structure information, and the output is a low-dimensional embedding vector of nodes and edges. The node embedding integrates the semantic features of the activity type and the spatial layer, and the edge embedding encodes the potential pattern of compatibility or conflict, resulting in a trained graph neural network model. Based on the trained graph neural network model, the compatibility of any activity-spatial layer combination in the graph is probabilistically quantified. By calculating the node embedding similarity and edge embedding weight values, a compatibility probability matrix is ​​generated, in which each element represents the compatibility score of a specific activity in a certain spatial layer. At the same time, a dynamic update mechanism for the graph is established to integrate the activities, spatial layers and relationship data in new property rights confirmation cases into the graph in real time, and the model parameters are retrained and updated to ensure that the compatibility quantification results reflect the latest sea use practice requirements and output an interpretable sea use activity-spatial layer relationship graph.

[0008] A further improvement of the technical solution of the present invention is that: S3 specifically includes: Based on the constructed sea use activity-spatial layer relationship graph, its topological structure and node / edge characteristics are analyzed. A breadth-first search-based graph traversal algorithm is used to identify high-frequency subgraph patterns. Starting from a high-frequency node, the algorithm traverses its neighborhood within three hops to extract recurring subgraph structures. This in turn identifies the common connection between sea use activities and the spatial layer. Degree centrality is used to calculate the degree of the node and quantify its importance. The top 10% of key nodes are retained to screen out key activity and spatial layer nodes. Graph embedding technology (Node2Vec) is used to map nodes and edges into a low-dimensional vector space, preserving their structure and attribute information. The embedded vectors are then grouped using the K-means clustering algorithm to identify combinations of sea use activities with similar compatible or conflicting patterns. Compatible patterns are characterized by close node embedding distances and high edge weights, while conflicting patterns correspond to separated embeddings or very low edge weights. The edge weight distribution is analyzed using a graph attention mechanism to extract conflict propagation patterns across spatial layers and form a pattern rule library. The universality of the model is verified through backtesting of historical cases, the matching degree between the extracted model and the real case is calculated, invalid models with overfitting or low coverage are eliminated, and the significance of the model is evaluated using statistical tests to ensure its non-randomness. The verified models are classified according to the conflict type to build a structured knowledge base.

[0009] A further improvement of the technical solution of the present invention is that: S4 specifically includes: The historical sea use activity-spatial layer relationship map is converted into a heterogeneous graph structure, with clear node and edge types. The conflict intensity attribute in the historical pattern is embedded as the edge weight. A graph neural network is used to model heterogeneous interactions. Multi-type node encoders are used to capture feature differences between the activity and spatial layers. The attention mechanism is used to aggregate neighborhood information to generate node embeddings that fuse structure and attributes. For newly submitted combinations of sea use activities, they are mapped into a pre-trained heterogeneous graph as new nodes, and their embedded representations are updated through the message passing mechanism of the graph neural network. Based on the connection relationship between the new sea use activity and the existing spatial layer, the interaction weights between the new sea use activity and its neighboring nodes are dynamically calculated. Combined with the feature distribution learned from the historical pattern, the compatibility probability of the combination is output. If the predicted probability is lower than the preset conflict threshold, it is marked as a potential conflict. If it is higher than the conflict threshold, it is determined to be compatible. For activity combinations predicted to be potential conflicts, we further analyze their conflict types and transmission paths. Through the edge weight distribution of the graph neural network, we identify the direct sources and indirect transmission chains of the conflicts, extract the conflict patterns as structured rules, and store them in a dynamic knowledge base.

[0010] A further improvement of the technical solution of the present invention is that the process of extracting the conflict pattern into structured rules is: Based on the edge weight distribution of graph neural networks, the direct source nodes and indirect impact paths of the conflict are located, high-weight conflict edges directly connected to the new active node are extracted, the direct conflicting parties are identified, and the gradient contribution of the edge weight is calculated through backpropagation. The indirect transmission chain of cross-node hops is tracked. At the same time, combined with a topological sorting algorithm, the nodes are sorted by dependency to ensure that the path is free of loops and covers key nodes. The complete transmission path diagram of the conflict is output, annotated with edge weights and node types. Based on the node types, edge weights, and historical labels in the transmission path, the conflict types are summarized, including spatial overlap conflicts, ecological chain conflicts, and temporal conflicts. The conflict intensity (weight mean) and impact range (transmission path length) are quantified. The identified transmission paths are converted into structured rules, and the rules are then encoded into a machine-readable format through logic programming. New rules are checked for redundancy and conflict with existing rules in the knowledge base. Similarity calculation based on graph embedding is used to merge duplicate rules, eliminate low-coverage rules, verify the matching degree between the new rule base and historical cases, ensure compatibility with historical data, retain the core rules covering the top 20%, add timestamps to the rules, and regularly eliminate outdated entries that have not been triggered in five years.

[0011] A further improvement of the technical solution of the present invention is that: S5 specifically includes: Based on the spatial characteristics of the sea area and the needs of sea-use activities, a hierarchical suitability evaluation index system was established, which includes three primary indicators: natural, social, and ecological. These indicators are further divided into secondary indicators, which are further divided into tertiary indicators. The analytic hierarchy process (AHP) was used to determine the weight of each indicator. The weight distribution was obtained by constructing a judgment matrix, conducting a consistency test, and calculating the eigenvector. A weighted overlay analysis was performed on the rasterized data based on weights to generate a suitability score map for marine activities. The score range was 1-5, with 1 being unsuitable and 5 being optimal. The continuous score was discretized into three levels using the natural breakpoint method: suitable (≥4 points), relatively suitable (3-4 points), and unsuitable (<3 points). The boundaries of marine functional zoning were then superimposed, and prohibited development areas were eliminated to form a preliminary distribution map of suitable areas. The suitability score map is multiplied by the compatibility probability predicted by the graph neural network to obtain a comprehensive potential value. The potential areas are divided into high potential areas, medium potential areas and low potential areas according to percentiles. High potential areas are recommended for three-dimensional layout, which meet both high suitability and high compatibility. Conflict risk points in medium potential areas that require manual review are automatically marked, and low potential areas are directly excluded.

[0012] A further improvement of the technical solution of the present invention is that the expression of the sea activity suitability scoring pattern is as follows: ; Where, For grid cells Total suitability score (range 1-5 points), 、 、 is the weight of natural, ecological and social indicators, For the natural index k in Standardized score (1-5 points), 、 is the standardized score of ecological and social indicators (1-5 points), n is the number of secondary and tertiary indicators under the natural indicator, m is the number of secondary and tertiary indicators under the ecological indicator, and q is the number of secondary and tertiary indicators under the social indicator; The expression of the comprehensive potential value is as follows: ; Where, For Grid The overall potential value (range 0-5), is the compatibility probability, is an indicator function, where the non-prohibited development area is 1, otherwise it is 0, and the high potential area is highly adaptable ( ) and high compatibility ( ), the medium potential zone is excellent in a single indicator of adaptability and compatibility, and the low potential zone is unqualified in any indicator.

[0013] A further improvement of the technical solution of the present invention is that: S6 specifically includes: Based on the spatial characteristics of high-potential areas and the demand for marine activities, key parameters were extracted, including natural and social constraints such as water depth range, ecological sensitivity, and adjacent infrastructure. These parameters were aligned with planning objectives. Spatial overlay analysis was used to exclude areas that conflicted with existing functional zoning to ensure that the recommended solutions met statutory boundaries. Combined with the compatibility probability matrix output by the graph neural network, grid cells that met both high suitability and high compatibility were selected to form an initial set of candidate areas. For candidate areas, the vertical space ownership scope of the water surface, water body, seabed, and subsoil is divided according to the three-dimensional layered utilization rules of the sea area. A layered weighted method is used to allocate sea activities, with activities with strong spatial exclusivity prioritized in the dominant layer, and activities with high compatibility sharing the remaining space. For cross-layer activity combinations, vertical safety distance recommendations are generated, and a three-dimensional space occupancy diagram is output to ensure that activities on each layer do not interfere with each other. Use the historical case library to backtest and verify the preliminary layout plan, calculate the matching degree between the plan and existing successful cases, eliminate options with deviations exceeding the preset deviation threshold, determine the final optimized layout plan, and synchronously update the optimized layout plan to the dynamic knowledge base, recording the layout parameters and decision-making basis.

[0014] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared to the prior art: The present invention provides a method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation. By constructing a multi-dimensional suitability evaluation system and a compatibility probability matrix, it can identify high-potential areas that simultaneously meet natural conditions, ecological constraints, and social needs. It also avoids the plane conflicts of traditional two-dimensional planning by using layered weighted overlay analysis. The three-dimensional layered utilization rules optimize the vertical space resource allocation, making the same sea area compatible with multiple sea use activities and reducing inefficient or conflicting layouts.

[0015] This paper provides a method for identifying and arranging high-potential marine areas in three-dimensional space based on suitability assessment. This method uses a compatibility prediction model driven by a graph neural network. By dynamically updating the relationship graph and conducting real-time conflict detection, it can pre-identify conflict paths in medium- and low-potential areas and quantify the potential risks of combining historical conflict patterns with new activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] Example 1, as Figure 1 、 Figure 2 As shown, the present invention provides a method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation, comprising the following steps: S1. Collect natural, social, and ecological data of sea areas and historical three-dimensional property rights confirmation project cases, carry out cleaning, conversion and standardization pre-processing, unify the coordinate system and accuracy standards, build a three-dimensional spatial basic database of sea areas, and collect three categories of natural, social and ecological data of the target sea areas. Among them, natural data include water depth, bottom type, marine disaster risk, etc., social data covers transportation network, port location conditions, tourism supporting facilities, etc., ecological data involves marine ecological protection red line, seawater quality, biological resource distribution, etc. At the same time, collect historical three-dimensional property rights confirmation project case materials, including types of sea use activities, spatial layer distribution, conflict records and solutions. Data sources include government public data, remote sensing images, industry reports and field surveys. After the collection is completed, conduct preliminary screening to eliminate duplicate, missing or obviously erroneous data to ensure the integrity and reliability of the data base, and clean the collected raw data, including format unification, outlier correction and missing value filling; At the same time, data conversion is carried out to unify data in different formats into a format that is easy to process, and standardization is carried out to unify the measurement units and classification standards of the data to ensure that all spatial data are in the same coordinate system and the accuracy meets the project requirements, eliminate errors and conflicts caused by inconsistent data formats and standards, and improve data quality and consistency. Based on the processed data, the architecture of the marine three-dimensional spatial basic database is designed, and the core table structure and field definition are determined. The core table structure of the database includes the marine basic information table, natural data table, social data table, ecological data table and historical case table. Among them, the marine basic information table includes ID, coordinate range and administrative division, the natural data table includes water depth, bottom quality and disaster level, the social data table includes transportation network and port location score, the ecological data table includes red line range and water quality level, and the historical case table includes sea use combination, conflict type and solution. The processed data is imported into the database according to the designed architecture, and data entry verification is carried out to ensure that the data is stored accurately. At the same time, a data index and query mechanism is established to form a marine three-dimensional spatial basic database. S2. Analyze historical three-dimensional land rights confirmation cases based on graph neural networks, taking sea use activities and spatial layers as nodes and relationships in historical cases as edges. Construct a sea use activity-spatial layer relationship map and quantify compatibility probabilities. Based on historical three-dimensional land rights confirmation cases, define sea use activity types and spatial layers (water surface, water body, seabed, and subsoil) as two types of nodes in the map. Node attributes include the type of sea use activity, spatial range, temporal characteristics, and physical and ecological attributes of the spatial layer. Interaction relationships in historical three-dimensional land rights confirmation cases are used as edges, and edge attributes record conflict intensity. The attribute encoding format of nodes and edges is unified, and topological verification is used to ensure graph connectivity. Each sea use activity must be associated with at least one spatial layer, and spatial layer nodes must be indirectly connected through activity nodes. Isolated nodes or redundant edges are removed to form a structured initial map. A graph neural network (graph convolutional network) is used to train the initial map to learn the potential feature representations of nodes and edges. The input is the initial node features and graph structure information, and the output is a low-dimensional embedding vector of nodes and edges. Node embeddings integrate the semantic features of activity types and spatial layers, and edge embeddings encode potential patterns of compatibility or conflict. During the training process, known compatibility labels in historical cases are used as supervisory signals, and the model parameters are optimized through the cross-entropy loss function, so that the embedding vector can accurately reflect the compatibility probability distribution between different activities and spatial layers, improve the generalization ability of relationship prediction, and obtain a trained graph neural network model. Based on the trained graph neural network model, the compatibility of any activity-spatial layer combination in the graph is probabilistically quantified. By calculating the node embedding similarity and edge embedding weight value, a compatibility probability matrix is ​​generated, in which each element represents the compatibility score of a specific activity in a certain spatial layer. At the same time, a dynamic update mechanism for the graph is established to integrate the activity, spatial layer and relationship data in the new right confirmation case into the graph in real time, retrain and update the model parameters to ensure that the compatibility quantification results reflect the latest sea use practice requirements and output an interpretable sea use activity-spatial layer relationship graph; The expression of node embedding similarity is as follows: ; Where, is the node embedding similarity, Sea use activity node and the embedding vector of the spatial layer node s (generated by GNN), is the vector norm, is a learnable weight matrix for nonlinear feature interaction, is the vector concatenation operation, is the balance coefficient, set to 0.5, The larger the value of , the stronger the semantic compatibility between the activity and the spatial layer; The expression of edge embedding weight value is as follows: ; Where, is the edge embedding weight value, For the edge The initial embedding of is the edge feature transformation matrix, q is the attention weight vector, is the activation function (slope = 0.2), which alleviates the vanishing gradient. is a learnable parameter vector (the transpose of the column vector, i.e., the row vector), which is used to map the concatenated high-order features into a scalar weight value. Edge feature Embedded with both end nodes 、 The concatenation result is a high-order representation after linear transformation, A value close to 1 indicates less historical conflict or stronger synergy; The expression of the compatibility probability matrix is ​​as follows: ; Where, is the compatibility probability matrix, representing sea use activities The compatibility probability at spatial layer s, is the control coefficient, which controls the sharpness of the probability distribution. For strong compatibility, High conflict; S3. Analyze the compatibility of different combinations of sea-use activities in historical projects using the sea-use activity-spatial layer relationship graph, extract compatibility and conflict patterns, analyze the topological structure and node / edge features based on the constructed sea-use activity-spatial layer relationship graph, and identify high-frequency subgraph patterns using a breadth-first search-based graph traversal algorithm. Starting from a high-frequency node, traverse its neighborhood within 3 hops to extract recurring subgraph structures, and then determine the common connection between sea-use activities and the spatial layer. Calculate the degree of the node using degree centrality, quantify the node importance, retain the top 10% of key nodes, and screen out key activities and spatial layer nodes. Use graph embedding technology (Node2Vec) to map nodes and edges into a low-dimensional vector space, preserving their structure and attribute information. Use the K-means clustering algorithm to group the embedded vectors and identify sea-use activity combinations with similar compatibility or conflict patterns. Compatibility patterns are characterized by close node embedding distances and high edge weights, while conflict patterns correspond to separated embeddings or extremely low edge weights. The edge weight distribution is analyzed using a graph attention mechanism, and the conflict transmission patterns across spatial layers are extracted to form a pattern rule library. The universality of the patterns is verified through backtesting historical cases, and the matching degree between the extracted patterns and real cases is calculated. Invalid patterns with overfitting or low coverage are eliminated, and statistical tests are used to evaluate the significance of the patterns to ensure their non-randomness. The verified patterns are classified by conflict type to construct a structured knowledge base. The expression for calculating degree centrality is as follows: ; Where, is the degree centrality of the node, v is the current node to be calculated, u is all other nodes in the graph (including sea activities and spatial layer nodes), V is the set of all nodes in the graph, and E is the set of all edges in the graph. is an indicator function that takes 1 when there is an edge between nodes v and u, and takes 0 otherwise. The larger the value of , the more extensive the connection between nodes in the graph; The expression for embedding vector grouping is as follows: ; ; ; Where J is the objective function of the K-means clustering algorithm, i.e., the intra-cluster sum of squared errors, K is the preset number of clusters, x is the node or edge embedding vector generated by Node2Vec, is the set of vectors in the rth cluster, is the centroid of the rth cluster. The smaller J is, the more similar the vectors within the cluster are. The embedding distance of the nodes within the compatible mode cluster is small, and the nodes within the conflict mode cluster are scattered. The expression of the matching degree between the pattern and the real case is as follows: ; Where, is the matching degree between the pattern and the real case, P is the extracted subgraph pattern, C is the historical case set, is the graph substructure corresponding to the c-th case, is a subgraph isomorphism relationship, that is, the pattern P is exists in The closer it is to 1, the more universal the pattern is; The expression of the statistical test is as follows: ; Where, is the chi-square statistic, is the actual observed frequency, is the expected frequency when randomly distributed, The larger it is, the more significant the non-randomness of the pattern is; S4. Combined with historical patterns, a graph neural network model is used to predict the compatibility probability of new sea use activity combinations and identify potential conflicts. The historical sea use activity-spatial layer relationship map is converted into a heterogeneous graph structure, node types and edge types are clarified, and the conflict intensity attributes in the historical patterns are embedded as edge weights. Graph neural networks are used to model heterogeneous interactions. Multi-type node encoders are used to capture the feature differences between the activity and spatial layers. The attention mechanism is used to aggregate neighborhood information and generate node embeddings that fuse structure and attributes. During model training, the compatible labels of historical activity combinations are used as supervisory signals, and the cross-entropy loss function is optimized to enable the network to learn the implicit feature distribution of compatible and conflicting patterns. For newly submitted sea use activity combinations, they are mapped to the pre-trained heterogeneous graph as new nodes, and their embedded representations are updated through the message passing mechanism of the graph neural network. Based on the connection relationship between the new sea use activity and the existing spatial layer, the interaction weight with the neighboring nodes is dynamically calculated. Combined with the feature distribution learned from the historical pattern, the compatibility probability of the combination is output. If the predicted probability is lower than the preset conflict threshold, it is marked as a potential conflict. If it is higher than the conflict threshold, it is determined to be compatible. For the activity combination predicted as a potential conflict, its conflict type and transmission path are further analyzed. Through the edge weight distribution of the graph neural network, the direct source and indirect transmission chain of the conflict are identified, and the conflict pattern is extracted as a structured rule and stored in the dynamic knowledge base. At the same time, the newly confirmed compatibility / conflict cases are fed back to the model training set, and the graph neural network parameters are updated using an incremental learning strategy to ensure that the model adapts to the emergence of new activities and continuously improves the prediction accuracy and pattern coverage. Furthermore, the process of extracting conflict patterns into structured rules is: Based on the edge weight distribution of graph neural networks, the direct source nodes and indirect impact paths of the conflict are located, and high-weight conflict edges (edge ​​weight > 0.7) directly connected to the new active node are extracted. The direct conflicting parties are identified, and the gradient contribution of the edge weight is calculated through backpropagation. The indirect transmission chain of cross-node hops is tracked. At the same time, combined with a topological sorting algorithm, the nodes are sorted by dependency to ensure that the paths are loop-free and cover key nodes. The complete transmission path diagram of the conflict is output, and the edge weights and node types are annotated. Based on the node types, edge weights and historical labels in the transmission path, the conflict types are summarized. The conflict types include spatial overlap conflicts, ecological chain conflicts and temporal conflicts. Spatial overlap conflicts refer to physical occupancy on the same layer (overlapping area percentage), ecological chain conflicts refer to cross-layer ecological impacts, and temporal conflicts refer to contradictions in activity cycles. The conflict intensity (weighted mean) and impact range (transmission path length) are quantified, and the identified transmission paths are converted into structured rules. The rules are then encoded into a machine-readable format through logic programming. New rules are tested for redundancy and conflict with existing rules in the knowledge base. A graph-embedding-based similarity calculation is used (cosine similarity > 0.9 is considered duplicate), duplicate rules are merged, low-coverage rules are eliminated, and the new rule base is verified for matching with historical cases (rollback occurs if coverage drops > 5%) to ensure compatibility with historical data. The core rules covering the top 20% are retained, timestamps are added to the rules, and outdated entries that have not been triggered in five years are regularly eliminated. S5. A hierarchical analysis method is used to determine weights and conduct a hierarchical suitability evaluation of sea use activities. This generates a scoring map and, combined with the compatibility probability prediction results, categorizes high, medium, and low potential areas, prioritizing recommendations for three-dimensional layouts with low conflict probabilities. S6. Based on high-potential areas, recommend optimized layout plans to match the layout requirements of high-potential areas.

[0020] Example 2, as Figure 1 、 Figure 2As shown, based on Example 1, the present invention provides a technical solution: preferably, S5 specifically includes: Based on the spatial characteristics of the sea area and the needs of sea-use activities, a hierarchical suitability evaluation index system was established, which includes three types of first-level indicators: natural, social, and ecological. These indicators are divided into second-level indicators, and the second-level indicators are divided into third-level indicators. The analytic hierarchy process (AHP) was used to determine the weight of each indicator. By constructing a judgment matrix, a consistency test was conducted, and the eigenvector was calculated to obtain the weight distribution. The natural conditions had a higher weight (0.4-0.6), followed by ecological constraints (0.3-0.4), and social factors had a lower weight (0.1-0.2). The weight results were used to standardize the data of each indicator and eliminate dimensional differences. Based on the weights, a weighted overlay analysis was performed on the rasterized data to generate a sea-use activity suitability score map. The scoring range is 1-5 points, where 1 point is unsuitable and 5 points is optimal. The natural breakpoint method is used to discretize the continuous score into three levels: suitable is ≥4 points, relatively suitable is 3-4 points, and unsuitable is <3 points. Then, the boundaries of marine functional zoning are superimposed, and the prohibited development areas are eliminated to form a preliminary distribution map of suitable areas. The suitability score map is multiplied by the compatibility probability predicted by the graph neural network to obtain a comprehensive potential value. The potential areas are divided into high potential areas, medium potential areas, and low potential areas according to percentiles. High potential areas are recommended for three-dimensional layout, which meet both high suitability and high compatibility. Conflict risk points in medium potential areas that require manual review are automatically marked, and low potential areas are directly excluded. The following is a classification of primary, secondary, and tertiary indicators, as well as the data and data types for each indicator:

[0021] In addition, the expression for scoring the sea activity suitability is as follows: ; Where, For grid cells Total suitability score (range 1-5 points), 、 、 is the weight of natural, ecological and social indicators, For the natural index k in Standardized score (1-5 points), 、 is the standardized score of ecological and social indicators (1-5 points), n is the number of secondary and tertiary indicators under the natural indicator, m is the number of secondary and tertiary indicators under the ecological indicator, and q is the number of secondary and tertiary indicators under the social indicator; The expression of comprehensive potential value is as follows: ; Where, For Grid The overall potential value (range 0-5), is the compatibility probability, is an indicator function, where the non-prohibited development area is 1, otherwise it is 0, and the high potential area is highly adaptable ( ) and high compatibility ( ), the medium potential zone is when a single indicator of adaptability and compatibility is excellent, and the low potential zone is when any indicator is unqualified; S6 specifically includes: Based on the spatial characteristics of high-potential areas and the needs of sea-use activities, key parameters are extracted, including natural and social constraints such as water depth range, ecological sensitivity and adjacent infrastructure. At the same time, the planning goals are connected. Through spatial overlay analysis, areas that conflict with existing functional zoning are excluded to ensure that the recommended scheme complies with the statutory boundaries. Combined with the compatibility probability matrix output by the graph neural network, grid units that meet both high suitability and high compatibility are screened out to form an initial set of candidate areas. At this stage, it is necessary to focus on avoiding rigid constraint areas such as ecological protection red lines and channel anchorages. For the candidate areas, the water surface is divided according to the three-dimensional layered utilization rules of the sea area. The vertical space ownership scope of water bodies, seabeds and subsoil is allocated using a hierarchical weighted method for allocating sea-use activities. Activities with strong spatial exclusivity are prioritized in the dominant layer, and activities with high compatibility share the remaining space. For cross-layer activity combinations, vertical safety distance recommendations are generated, and a three-dimensional space occupancy diagram is output to ensure that activities on each layer do not interfere with each other. The preliminary layout plan is backtested and verified using the historical case library, and the matching degree between the plan and existing successful cases is calculated. Options with deviations exceeding the preset deviation threshold are eliminated, and the final optimized layout plan is determined. The optimized layout plan is synchronously updated to the dynamic knowledge base, and the layout parameters and decision-making basis are recorded.

[0022] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation, characterized by: The following steps are involved: S1. Collect natural, social, and ecological data of the sea area and historical three-dimensional rights confirmation project cases, and pre-process them to build a three-dimensional spatial basic database of the sea area; S2. Analyze historical three-dimensional land rights confirmation cases based on graph neural networks and construct a sea use activity-spatial layer relationship map; S3. Analyze the compatibility of different combinations of sea-use activities in historical projects using the sea-use activity-spatial layer relationship map and extract compatibility and conflict patterns; S4. Combine historical patterns and use graph neural network models to predict the compatibility probability of new combinations of maritime activities and identify potential conflicts. S5. Determine weights based on the analytic hierarchy process, conduct a hierarchical suitability evaluation of sea use activities, generate scoring maps, and combine the compatible probability prediction results to divide high, medium, and low potential areas; S6. Based on high-potential areas, recommend optimized layout plans to match the layout requirements of high-potential areas.

2. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 1 is characterized by: Said S1 specifically includes: Collect data on the natural, social, and ecological aspects of the target sea area. Also collect case studies of historical three-dimensional land rights confirmation projects, including types of sea use activities, spatial distribution, conflict records, and resolutions. Clean the collected raw data, including format unification, outlier correction and missing value filling. At the same time, perform data conversion and standardization to unify the data's measurement units and classification standards. Based on the processed data, the architecture of the basic database of the three-dimensional space of the sea area is designed, and the core table structure and field definition are determined. The core table structure of the database includes the basic information table of the sea area, the natural data table, the social data table, the ecological data table and the historical case table. The processed data is imported into the database according to the designed architecture, and the data entry verification is performed. At the same time, a data index and query mechanism is established to form a basic database of the three-dimensional space of the sea area.

3. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 1 is characterized by: The S2 specifically includes: Based on historical three-dimensional land rights confirmation cases, the types of sea use activities and spatial layers are defined as two types of nodes in the graph. Node attributes include the type of sea use activity, spatial scope, temporal characteristics, and the physical and ecological attributes of the spatial layer. The interactive relationships in historical three-dimensional land rights confirmation cases are used as edges, and edge attributes record the intensity of conflicts. The attribute encoding format of nodes and edges is unified. Spatial layer nodes must be indirectly connected through activity nodes, and isolated nodes or redundant edges are eliminated to form a structured initial graph. A graph neural network is used to train the initial graph to learn the latent feature representations of nodes and edges. The input is the initial node features and graph structure information, and the output is a low-dimensional embedding vector of nodes and edges. The node embedding integrates the semantic features of the activity type and the spatial layer, and the edge embedding encodes the potential patterns of compatibility or conflict. The trained graph neural network model is obtained. Based on the trained graph neural network model, the compatibility of any activity-space layer combination in the graph is probabilistically quantified, and a compatibility probability matrix is ​​generated by calculating the node embedding similarity and edge embedding weight values. At the same time, a dynamic update mechanism for the graph is established to integrate the activities, space layers and relationship data in new property rights confirmation cases into the graph in real time, retrain and update the model parameters, and output an interpretable sea use activity-space layer relationship graph.

4. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 1 is characterized by: The S3 specifically includes: Based on the constructed sea use activity-spatial layer relationship graph, its topological structure and node / edge characteristics are analyzed. A breadth-first search-based graph traversal algorithm is used to identify high-frequency subgraph patterns. Starting from a high-frequency node, the algorithm traverses its neighborhood within three hops to extract recurring subgraph structures. This in turn identifies the common connection between sea use activities and the spatial layer. Degree centrality is used to calculate the degree of the node and quantify its importance. The top 10% of key nodes are retained to screen out key activity and spatial layer nodes. Graph embedding technology is used to map nodes and edges into a low-dimensional vector space, preserving their structure and attribute information. The embedded vectors are grouped using the K-means clustering algorithm to identify combinations of sea use activities with similar compatible or conflicting patterns. The edge weight distribution is analyzed using a graph attention mechanism to extract conflict propagation patterns across spatial layers and form a pattern rule library. The universality of the model is verified through backtesting of historical cases, the matching degree between the extracted model and the real case is calculated, invalid models with overfitting or low coverage are eliminated, and the significance of the model is evaluated using statistical tests. The verified models are classified according to the conflict type to construct a structured knowledge base.

5. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 1 is characterized by: The S4 specifically includes: The historical sea use activity-spatial layer relationship map is converted into a heterogeneous graph structure, with clear node and edge types. The conflict intensity attribute in the historical pattern is embedded as the edge weight. A graph neural network is used to model heterogeneous interactions. Multi-type node encoders are used to capture feature differences between the activity and spatial layers. The attention mechanism is used to aggregate neighborhood information to generate node embeddings that fuse structure and attributes. For newly submitted combinations of sea use activities, they are mapped into a pre-trained heterogeneous graph as new nodes, and their embedded representations are updated through the message passing mechanism of the graph neural network. Based on the connection relationship between the new sea use activity and the existing spatial layer, the interaction weights between the new sea use activity and its neighboring nodes are dynamically calculated. Combined with the feature distribution learned from the historical pattern, the compatibility probability of the combination is output. If the predicted probability is lower than the preset conflict threshold, it is marked as a potential conflict. If it is higher than the conflict threshold, it is determined to be compatible. For activity combinations predicted to be potential conflicts, we further analyze their conflict types and transmission paths. Through the edge weight distribution of the graph neural network, we identify the direct sources and indirect transmission chains of the conflicts, extract the conflict patterns as structured rules, and store them in a dynamic knowledge base.

6. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 5 is characterized by: The process of extracting conflict patterns into structured rules is as follows: Based on the edge weight distribution of graph neural networks, the direct source nodes and indirect impact paths of the conflict are located, high-weight conflict edges directly connected to the new active node are extracted, the direct conflicting parties are identified, and the gradient contribution of the edge weight is calculated through backpropagation. The indirect transmission chain of cross-node hops is tracked. At the same time, combined with a topological sorting algorithm, the nodes are sorted by dependency and the complete transmission path diagram of the conflict is output, annotated with edge weights and node types. Based on the node types, edge weights, and historical labels in the transmission paths, we classify conflict types, including spatial overlap, ecological chain, and temporal conflicts. We quantify the intensity and scope of the conflicts, and transform the identified transmission paths into structured rules. These rules are then encoded into a machine-readable format through logic programming. New rules are checked for redundancy and conflict with existing rules in the knowledge base. Similarity calculation based on graph embedding is used to merge duplicate rules, eliminate low-coverage rules, verify the matching degree between the new rule base and historical cases, retain the core rules that cover the top 20%, add timestamps to the rules, and regularly eliminate outdated entries that have not been triggered in five years.

7. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 1 is characterized by: The S5 specifically includes: Based on the spatial characteristics of the sea area and the needs of sea-use activities, a hierarchical suitability evaluation index system is established, which includes three primary indicators: natural, social, and ecological. These indicators are further divided into secondary indicators, which are further divided into tertiary indicators. The analytic hierarchy process is used to determine the weight of each indicator. The judgment matrix is ​​constructed, consistency is tested, and the characteristic vector is calculated to obtain the weight distribution. A weighted overlay analysis was performed on the rasterized data based on weights to generate a suitability score map for marine activities. The score range was 1-5, with 1 being unsuitable and 5 being optimal. The continuous score was discretized into three levels using the natural breakpoint method: suitable (≥4 points), relatively suitable (3-4 points), and unsuitable (<3 points). The boundaries of marine functional zoning were then superimposed, and prohibited development areas were eliminated to form a preliminary distribution map of suitable areas. The suitability score map is multiplied by the compatibility probability predicted by the graph neural network to obtain a comprehensive potential value. The potential areas are divided into high potential areas, medium potential areas and low potential areas according to percentiles. High potential areas are recommended for three-dimensional layout, which meet both high suitability and high compatibility. Conflict risk points in medium potential areas that require manual review are automatically marked, and low potential areas are directly excluded.

8. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 7 is characterized by: The expression of the sea use activity suitability score map is as follows: ； Where, For grid cells The total suitability score, 、 、 is the weight of natural, ecological and social indicators, For the natural index k in The standardized score of 、 is the standardized score of ecological and social indicators, n is the number of secondary and tertiary indicators under natural indicators, m is the number of secondary and tertiary indicators under ecological indicators, and q is the number of secondary and tertiary indicators under social indicators; The expression of the comprehensive potential value is as follows: ； Where, For Grid The comprehensive potential value of is the compatibility probability, is an indicator function, which is 1 in non-prohibited development areas and 0 otherwise.

9. The method for identifying and arranging high-potential areas in marine three-dimensional space based on suitability evaluation according to claim 8 is characterized by: The S6 specifically includes: Based on the spatial characteristics of high-potential areas and the demand for sea-use activities, key parameters were extracted, including water depth range, ecological sensitivity, and natural and social constraints of adjacent infrastructure. These parameters were aligned with planning objectives. Spatial overlay analysis was used to exclude areas that conflicted with existing functional zoning. Combined with the compatibility probability matrix output by the graph neural network, grid cells that met both high suitability and high compatibility were selected to form an initial set of candidate areas. For candidate areas, the vertical space ownership scope of the water surface, water body, seabed, and subsoil is divided according to the three-dimensional layered utilization rules of the sea area. A layered weighted method is used to allocate sea activities, with activities with strong spatial exclusivity prioritized in the dominant layer, and activities with high compatibility sharing the remaining space. For cross-layer activity combinations, vertical safety distance recommendations are generated, and a three-dimensional space occupancy diagram is output; Use the historical case library to backtest and verify the preliminary layout plan, calculate the matching degree between the plan and existing successful cases, eliminate options with deviations exceeding the preset deviation threshold, determine the final optimized layout plan, and synchronously update the optimized layout plan to the dynamic knowledge base, recording the layout parameters and decision-making basis.

Images