Land space planning decision-making method and system based on knowledge graph

Abstract

The invention discloses a territorial space planning decision-making method and system based on a knowledge graph, and relates to the technical field of territorial space planning, and the method comprises the steps: obtaining space vector data and unstructured planning text data of territorial space planning, analyzing the business logic of territorial space planning, and constructing a multi-dimensional ontology network; mapping the space vector data and the unstructured planning text data into a multi-dimensional ontology network to generate a territorial space planning knowledge graph, constructing a land use state reachable graph based on the territorial space planning knowledge graph, traversing the land use state reachable graph by using a backtracking search algorithm to obtain a land use index distribution scheme, and determining the land use index distribution scheme according to the land use index distribution scheme. The method comprises the following steps: establishing a hierarchical timing Petri network based on a land use index allocation scheme, performing state evolution based on the hierarchical timing Petri network, detecting resource occupation conflicts and sequential logic conflicts, eliminating the resource occupation conflicts and the sequential logic conflicts, and outputting a decision result.

Description

Technical Field

[0001] This invention relates to the field of land spatial planning technology, and in particular to a land spatial planning decision-making method and system based on knowledge graphs. Background Technology

[0002] With the development of geospatial information technology, land spatial planning has shifted from traditional map-based operations to digital management based on geographic information systems (GIS). Existing planning support technologies primarily rely on the overlay analysis and statistics of spatial vector data. However, land spatial planning is a typical multi-source, heterogeneous, and knowledge-intensive task. Besides structured geospatial data, planning decisions are also strongly constrained by a large amount of unstructured text data, including laws and regulations, the "three zones and three lines" control rules, industry standards, and approval process specifications. Existing GIS technologies and database systems struggle to achieve deep semantic-level fusion of spatial vector data and natural language rule text, resulting in a significant semantic gap between planning data and control rules. Planners often need to manually review numerous documents to verify the compliance of spatial layouts. This approach is not only inefficient but also prone to omissions or misjudgments when faced with massive amounts of land and complex constraints, failing to achieve truly intelligent machine-assisted decision-making.

[0003] Currently, Chinese invention patent application number 202511669402.5 discloses an intelligent optimization method and system for land spatial planning. This method calculates boundary elasticity identification and degradation potential indices by collecting multi-temporal remote sensing image data, constructs a linkage pressure index, and combines it with threshold judgment to achieve sensitivity classification and response strategy adjustment for ecological red line boundaries. However, this technology mainly relies on the monitoring of physicochemical indicators of raster data and static threshold triggering mechanisms. It struggles to achieve deep semantic-level fusion of unstructured planning control rules text and multi-source spatial data, and lacks the ability to model the temporal logic and dynamically simulate the planning implementation process. Consequently, it cannot achieve globally optimal allocation of land use indicators under complex rigid control constraints, nor can it automatically detect and resolve potential temporal logic conflicts and resource occupation conflicts during project implementation. Summary of the Invention

[0004] The technical problem solved by this invention is that existing technologies have difficulty in solving the problem of knowledge silos formed in land spatial planning due to the difficulty in deeply integrating spatial data and planning rule texts. In addition, when faced with multiple rigid control constraints and strong spatiotemporal coupling requirements, traditional static planning methods lack the ability to dynamically extrapolate the planning implementation process, resulting in the difficulty in accurately calculating the optimal land use index allocation scheme and the inability to automatically detect and resolve potential temporal logic conflicts and resource occupation conflicts during project implementation.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a knowledge graph-based land spatial planning decision-making method, comprising the following steps: Step S1: Obtain spatial vector data and unstructured planning text data of the land spatial planning, parse the business logic of the land spatial planning and construct a multi-dimensional ontology network, map the spatial vector data and unstructured planning text data into the multi-dimensional ontology network, and generate a land spatial planning knowledge graph. Step S2: Based on the land spatial planning knowledge graph, construct a land use status reachability map, and use a backtracking search algorithm to traverse the land use status reachability map to obtain a land use index allocation scheme; Step S3: Based on the land use index allocation scheme, establish a hierarchical time-assigned Petri net; Step S4: Based on the hierarchical time-assigned Petri net, perform state evolution, detect resource occupation conflicts and temporal logic conflicts, and output the decision result after eliminating resource occupation conflicts and temporal logic conflicts.

[0006] Preferably, step S1 includes the following sub-steps: Step S101: Analyze the business logic of land and space planning, extract business concept knowledge points, define class nodes, object attributes and data attributes, and construct a multi-dimensional ontology network; The business logic includes land space survey, land space planning, land use control land classification guidelines, land space planning data submission standards, and business process specifications for administrative approval and planning implementation management. The class nodes include spatial resource dimension, planning goal dimension, control rule dimension, implementation project dimension and evolution process dimension. The spatial resource dimension includes natural geographic entities and planned land parcels. The planning goal dimension includes population indicators, economic indicators and ecological indicators. The control rule dimension includes the three zones and three lines and legal and regulatory constraints. The implementation project dimension includes development and construction projects. The evolution process dimension includes land parcel status change events. The object attributes include inclusion and adjacency relationships describing spatial topology, constraint relationships describing rule restrictions, occupancy relationships describing project implementation, contribution relationships describing indicator achievement, and temporal evolution relationships describing state changes. The data attributes include the geometric coordinates and area values ​​of the land parcel, the semantic description string of the planning text, the values ​​of the control indicators, the time window parameters of the project, and the status identifier code of the approval process.

[0007] Preferably, step S1 further includes the following sub-steps: Step S102: parse the polygon vertex coordinate sequence and boundary information in the spatial vector data, generate spatial unit nodes, and extract the adjacency and inclusion relationships between spatial unit nodes based on coordinate operations to generate spatial topological edges connecting different spatial unit nodes. An entity recognition model is constructed, which includes a BERT embedding layer, a bidirectional long short-term memory network layer, and a conditional random field layer. The unstructured planning text data is input into the entity recognition model, and the entity labeling sequence is output. The entity labeling sequence includes spatial entity labels, control indicator labels, constraint operator labels, and threshold value labels. Based on dependency parsing, a syntactic parse tree is constructed for the entity annotation sequence. Semantic quadruples are extracted based on the dependency relationships between entities. The semantic quadruples include subject, index, operator, and value. Based on the semantic quadruple, class nodes in the multidimensional ontology network are retrieved, corresponding control rule nodes with constraint logic are generated, and the values ​​in the semantic quadruple are assigned to the data attributes of the control rule nodes, and an associated index edge is established from the spatial unit node to the control rule node. Step S103: Map the spatial unit nodes and control rule nodes to the multidimensional ontology network, traverse the temporal logic definition in the evolution process dimension, instantiate directed constraint edges between spatial unit nodes of land use change, and construct a knowledge graph of national land spatial planning.

[0008] Preferably, step S2 includes the following sub-steps: Step S201: Taking the current land use type of the spatial unit node as the initial state, and according to the planning evolution rules extracted from the control rule dimension, select variable spatial unit nodes with the potential for land use type change. The land use type change operation of the variable spatial unit node is defined as the first state transition edge. Based on the evolution of the variable spatial unit node from the initial state to the planned target land use type, a land use state reachability graph is constructed, wherein the graph nodes of the land use state reachability graph represent the land use type distribution state of the variable spatial unit node.

[0009] Preferably, step S2 further includes the following sub-steps: Step S202 involves extracting the ecological protection red line boundary from the control rules dimension and extracting the planning overall control index threshold from the planning target dimension. A backtracking search algorithm is used to traverse the land use status reachability graph from the initial state. For each graph node on the traversal path, each graph node is defined as a state node. The spatial location of the state node and its topological relationship with the ecological protection red line boundary are determined. Specifically, this includes: If the boundary of the ecological protection red line is touched, the state branch shall be removed; If the ecological protection red line boundary is not touched, the cumulative area of ​​all types of land use under the statistical state node is calculated to obtain the land use index occupancy. Based on the land use index occupancy, the time attribute of the current state node is determined. If the planning period is nearing its end and the land use quota does not meet the threshold of the overall planning control index, then this state branch will be removed. If the process is in the planning period and the land use quota exceeds the threshold of the overall planning control index, then the state branch is retained and the penalty cost for process violations is calculated. Step S203: Extract development and construction cost data from the implementation project dimension, obtain the cumulative construction cost based on the development and construction cost data, extract ecological value parameters from the planning target dimension, obtain the cumulative ecological value loss based on the ecological value parameters, construct a cost estimation function, and calculate the cost estimate using the cost estimation function. The mathematical expression for the cost estimate is: ; in, As the cost estimate, The land use status can be reached as shown in the current state in the diagram. To accumulate construction costs, To accumulate ecological value loss, and These are the preset normalized weighting coefficients. This serves as an index for the overall control indicators of the plan. The first dimension defined in the planning objective dimension The priority weight of the overall control indicators of the project plan, For the first The target values ​​of the overall control indicators of the project plan, In the current state Next The land use quota occupancy of this indicator For heuristic search weights, The consequences of procedural violations; The A* search algorithm is used to search for the path that minimizes the cost estimate. The second state transition edges in the path with the minimum cost estimate are extracted and arranged in time sequence. All the second state transition edges are combined into a state transition sequence, and the state transition sequence is used as the land use index allocation scheme.

[0010] Preferably, step S3 includes the following sub-steps: Step S301: Based on the event logic and causal relationships in the state transition sequence, construct a hierarchical time-assigned Petri net. The hierarchical time-assigned Petri net includes a macro-level development time sequence layer and a micro-level land parcel approval layer, specifically including: In the macro-development time sequence layer, corresponding transitions are generated for each development and construction activity in the state transition sequence, and time sequence libraries and directed arcs connecting the transitions are established based on the predecessor and successor relationships of the development and construction activities. In the micro-plot approval layer, a corresponding resource repository is generated for each spatial unit node; Establish input and output arcs that point to the corresponding resource library from the transitions.

[0011] Preferably, step S3 further includes the following sub-steps: Step S302: Define the Token that flows between the time series repository and the resource repository in the hierarchical time-assigned Petri net as an object carrying data attributes. The data attributes include the project unique identifier code extracted from the multidimensional ontology network, project type priority information, location dependency parameters, and plot geometric coordinates and area values. Construct a data interaction interface, traverse the implementation project dimension in the multidimensional ontology network, extract the corresponding land use project instance according to the state transition sequence, obtain data attributes, write the obtained data attributes into the Token, determine the initial land use state according to the state transition sequence, place the Token in the corresponding resource repository, and generate the Token distribution state. Step S303: Extract the planned start time and construction period of the project from the implementation project dimension of the multidimensional ontology network, and use the planned start time and construction period of the project as the ignition time interval of the corresponding transition. The ignition condition for a transition is defined as follows: all input time series locations and resource locations of the transition contain the corresponding Token, and the current inference time is within the ignition time interval.

[0012] Preferably, step S4 includes the following sub-steps: Step S401: Based on the hierarchical time-assigned Petri net, perform state evolution according to discrete time steps, and monitor the number of tokens in the resource pool and the arrival status of the transition ignition time interval. When it is detected that the resource library representing the same spatial unit node contains two or more mutually exclusive tokens within the same time step, it is determined to be a resource occupation conflict. When it is detected that one of the transitions has reached the ignition time interval, but the input timing library or resource library of the transition does not contain a token that meets the ignition conditions, causing the transition to fail to ignite, it is determined to be a timing logic conflict.

[0013] Preferably, step S4 further includes the following sub-steps: Step S402: For the identified resource occupation conflict, read the data attributes encapsulated inside the conflicting Token and parse the position dependency parameter; Based on the location dependency parameter, projects that have resource occupation conflicts are divided into rigid location projects and flexible location projects. If the project is a rigid site selection project, a time-based avoidance strategy should be implemented, specifically including: Keeping the resource library associated with the Token unchanged, calculate the expected release time, and postpone the ignition time interval of the Token to after the expected release time of the conflicting resources. If the number of postponements exceeds a preset threshold, mark the corresponding rigid site selection project as infeasible and terminate the simulation. If the project is a flexible site selection project, a spatial dimension avoidance strategy will be implemented, specifically including: Search the knowledge graph of the national land spatial planning for alternative spatial unit nodes that have the same planning purpose as the current spatial unit node and are not occupied. Read the data attributes of the alternative spatial unit nodes and calculate the construction suitability score. Redirect the Token to the resource storage corresponding to the alternative spatial unit node with the highest construction suitability score and update the Token distribution status. Calculate the suitability score for construction. The suitability score is negatively correlated with the cost estimate. Redirect the token to the target node with the highest score. Add the spatial unit node to the token's taboo list. Within a preset number of iterations, redirect the token to the spatial unit node in the taboo list again. For timing logic conflicts, retrieve the upstream transition corresponding to the missing token, calculate the remaining minimum operation time of the upstream transition, and if the current simulation time allows for delay, automatically postpone the ignition time interval of the current transition; if the current simulation time does not allow for delay, generate a critical path warning and compress the duration parameter of the upstream transition. Step S403: Update the current state of the hierarchical timed Petri net by the delayed ignition time interval or the redirected token distribution state, and return to execute step S401 until all resource occupation conflicts and timing logic conflicts are eliminated, and output the decision result.

[0014] A knowledge graph-based land spatial planning decision-making system includes a mapping module, an optimization module, a modeling module, and a deduction module; The mapping module is used to acquire spatial vector data and unstructured planning text data of land spatial planning, parse the business logic of land spatial planning and construct a multidimensional ontology network, map the spatial vector data and unstructured planning text data to the multidimensional ontology network, and generate a land spatial planning knowledge graph. The optimization module is used to construct a land use status reachability map based on the land spatial planning knowledge graph, and to traverse the land use status reachability map using a backtracking search algorithm to obtain a land use index allocation scheme. The modeling module is used to establish a hierarchical time-assigned Petri net based on the land use index allocation scheme; The inference module is used to perform discrete-time step state evolution based on the hierarchical time-assigned Petri net, detect resource occupation conflicts and temporal logic conflicts through the hierarchical time-assigned Petri net, adjust the project based on the project priority information extracted from the land spatial planning knowledge graph, and output the decision results.

[0015] The beneficial effects of this invention are as follows: By analyzing business logic to construct a multi-dimensional ontology network, heterogeneous data is mapped to generate a land spatial planning knowledge graph, achieving semantic-level deep integration of control rules and spatial data. A hierarchical time-assigned Petri net is established, and the tokens circulating in the net are defined as dynamic objects encapsulating location dependency parameters and project priority information. State evolution based on discrete time steps is performed, enabling the system to not only use backtracking search algorithms to traverse the land use state reachability graph to obtain the land use index allocation scheme with the lowest global cost, but also to accurately capture conflicts in dynamic simulations. Based on the token attributes, a time dimension avoidance strategy of delaying the ignition time is automatically executed for rigid site selection projects, and a spatial dimension avoidance strategy of redirecting to alternative nodes is executed for flexible site selection projects. Thus, under the premise of strictly adhering to the rigid constraints of the three zones and three lines and the overall planning control index threshold, the planning decision-making is transformed from static layout optimization to dynamic implementation simulation, ensuring the feasibility of the planning scheme and the conflict-free implementation of spatiotemporal logic. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the steps of a knowledge graph-based land spatial planning decision-making method according to an embodiment of the present invention; Figure 2 This is a basic flowchart of a knowledge graph-based land spatial planning decision-making system provided as an embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example 1, referring to Figure 1 This paper presents a knowledge graph-based method for land spatial planning decision-making, which includes the following steps: Step S1: Obtain spatial vector data and unstructured planning text data of the national land spatial planning, parse the business logic of the national land spatial planning and construct a multidimensional ontology network, map the spatial vector data and unstructured planning text data into the multidimensional ontology network, and generate a knowledge graph of national land spatial planning.

[0019] Step S2: Based on the knowledge graph of land spatial planning, construct a land use status reachability map, and use a backtracking search algorithm to traverse the land use status reachability map to obtain a land use index allocation scheme.

[0020] Step S3: Based on the land use index allocation scheme, establish a hierarchical time-assigned Petri net.

[0021] Step S4: Perform state evolution based on hierarchical time-assigned Petri nets, detect resource occupation conflicts and temporal logic conflicts, and output decision results after eliminating resource occupation conflicts and temporal logic conflicts.

[0022] This invention constructs a multi-dimensional ontology network by parsing business logic, maps heterogeneous data to generate a knowledge graph for land and space planning, and achieves semantic-level deep integration of control rules and spatial data. It establishes a hierarchical time-assigned Petri net, defining the tokens circulating in the net as dynamic objects encapsulating location dependency parameters and project priority information, and performing state evolution based on discrete time steps. This enables the system to not only use a backtracking search algorithm to traverse the land use state reachability graph to obtain the land use index allocation scheme with the lowest global cost, but also to accurately capture conflicts in dynamic simulations. Based on the token attributes, it automatically implements a time dimension avoidance strategy of delaying the ignition time for rigid site selection projects and a spatial dimension avoidance strategy of redirecting flexible site selection projects to alternative nodes. Thus, under the premise of strictly adhering to the rigid constraints of the three zones and three lines and the overall planning control index threshold, it realizes the leap from static layout optimization to dynamic implementation simulation in planning decisions, ensuring the feasibility of the planning scheme and the conflict-free implementation of spatiotemporal logic.

[0023] In a specific embodiment, step S1 includes the following sub-steps: Step S101: Analyze the business logic of land and space planning, extract business concept knowledge points, define class nodes, object attributes and data attributes, and construct a multi-dimensional ontology network.

[0024] The business logic includes land space surveys, land space planning, land use control land classification guidelines, land space planning data submission standards, and business process specifications for administrative approval and planning implementation management.

[0025] The types of nodes include spatial resource dimension, planning goal dimension, control rule dimension, implementation project dimension, and evolution process dimension. Spatial resource dimension includes natural geographical entities and planned land parcels. Planning goal dimension includes population indicators, economic indicators, and ecological indicators. Control rule dimension includes the three zones and three lines and legal and regulatory constraints. Implementation project dimension includes development and construction projects. Evolution process dimension includes events that change the status of land parcels.

[0026] Object attributes include inclusion and adjacency relationships describing spatial topology, constraint relationships describing rule restrictions, occupancy relationships describing project implementation, contribution relationships describing indicator achievement, and temporal evolution relationships describing state changes.

[0027] Data attributes include the geometric coordinates and area values ​​of the land parcel, the semantic description string of the planning text, the values ​​of the control indicators, the time window parameters of the project, and the status identifier code of the approval process.

[0028] Step S102: Parse the polygon vertex coordinate sequence and boundary information in the spatial vector data to generate spatial unit nodes, and extract the adjacency and containment relationships between spatial unit nodes based on coordinate operations to generate spatial topological edges connecting different spatial unit nodes.

[0029] Natural language processing (NLP) techniques are used to parse unstructured planning text data, extract semantic descriptions of plot ratio, land use, and development timing restrictions, and generate control rule nodes, specifically including: An entity recognition model is constructed, which includes a BERT embedding layer, a bidirectional long short-term memory network layer, and a conditional random field layer. Unstructured planning text data is input into the entity recognition model, and the output is an entity label sequence, which includes spatial entity labels, control indicator labels, constraint operator labels, and threshold value labels.

[0030] Based on dependency parsing, a syntactic parse tree is constructed for the entity annotation sequence. Semantic quadruples are extracted based on the dependency relationships between entities. The semantic quadruples include the subject, index, operator, and value.

[0031] Based on the semantic quadruple, class nodes in the multidimensional ontology network are retrieved, control rule nodes containing corresponding constraint logic are generated, and the values ​​in the semantic quadruple are assigned to the data attributes of the control rule nodes. An associated index edge is established from the spatial unit node to the control rule node.

[0032] The construction status value and population statistics in the current status statistics are analyzed, and the construction status value and population statistics are used as attribute values ​​of spatial unit nodes. A correlation index edge is established from the spatial unit node to the control rule node and attribute value.

[0033] Step S103: Map spatial unit nodes and control rule nodes to a multi-dimensional ontology network, traverse the temporal logic definition in the evolution process dimension, instantiate directed constraint edges expressing predecessor and successor logic between spatial unit nodes of land use change, and construct a knowledge graph of national land spatial planning.

[0034] To clarify the data structure foundation of the multidimensional ontology network and the instantiation mapping logic of the control rule nodes, this embodiment adopts the following specific construction and processing steps: Based on the network ontology language, a graph pattern of a multidimensional ontology network is defined. Specifically, a land parcel class is defined in the spatial resource dimension, a floor area ratio constraint class is defined in the control rule dimension, and object attributes connecting nodes of different classes are defined, including the constrained attribute describing the constraint logic. The domain of the constrained attribute is set to the land parcel class, and the value domain is set to the floor area ratio constraint class.

[0035] The semantic quadruple that responds to the output of the entity recognition model, where the semantic quadruple is formally defined as In the control rule dimension, perform the following instantiation operation: Instantiate a control rule node belonging to the control rule dimension. It assigns a unique node identifier, Rule_Ind_XX, and reads the semantic quadruple from the string. Elements, will Elements as control rule nodes The tag attributes, read Elements and Elements, will Elements and Element writing control rule node The data attributes are used to transform unstructured text into structured node data carriers.

[0036] Traverse all spatial unit nodes in the spatial resource dimension of the multidimensional ontology network. For each spatial unit node, read the land use attribute value and compare it with the semantic quadruple. If the elements are consistent, and the result is consistent, then the current spatial unit node and the control rule node are checked. Between them, instantiate and generate a directed associative index edge of type "constrained to".

[0037] Taking the processing of the planning text "The plot ratio of Class I industrial land shall not exceed 2.0" as an example: the system extracts semantic quadruples. Generate a control rule node with the identifier Rule_Ind_01, set the floor area ratio as the label, and... The data attributes are stored in 2.0. The traversal reveals that the land use nature of one of the spatial unit nodes is "Class I industrial land". Therefore, a "constrained by" edge is established from one of the spatial unit nodes to Rule_Ind_01, realizing the semantic-level transformation from natural language constraints to graph topology, so that subsequent steps S2 and S4 can directly access the constraint conditions through graph traversal algorithm.

[0038] To address the problem in existing technologies where spatial vector data and unstructured planning text data such as laws and regulations lack semantic connections, making it difficult for machines to understand planning business logic and spatiotemporal evolution patterns, this invention constructs a multidimensional ontology network comprising five dimensions: spatial resources, planning objectives, control rules, implementation projects, and evolutionary process. It utilizes an entity recognition model composed of a BERT embedding layer, a bidirectional long short-term memory network layer, and a conditional random field layer to deeply analyze the text, extracting semantic quadruples to instantiate control rule nodes. Then, through associative index edges, rule constraints are precisely mapped to spatial unit nodes. Furthermore, directed constraint edges expressing temporal logic are generated based on the evolutionary process dimension. This constructs a land spatial planning knowledge graph capable of uniformly expressing physical spatial entities, abstract control rules, and dynamic evolutionary logic, achieving semantic-level fusion of multi-source heterogeneous data and the computability of planning knowledge.

[0039] In a specific embodiment, step S2 includes the following sub-steps: Step S201: Taking the current land use type of the spatial unit node as the initial state, and according to the planning evolution rules extracted from the control rules dimension, select variable spatial unit nodes with the potential for land use type change.

[0040] The land use type change operation of the variable spatial unit node is defined as the first state transition edge. Based on the evolution of the variable spatial unit node from the initial state to the planned target land use type, a land use state reachability graph is constructed. The graph nodes of the land use state reachability graph represent the land use type distribution state of the variable spatial unit node.

[0041] Step S202: Extract the ecological protection red line boundary from the control rules dimension and extract the planning overall control index threshold from the planning target dimension. Using a backtracking search algorithm, traverse the land use status reachability graph from the initial state. For each graph node on the traversal path, define each graph node as a state node and determine the spatial location of the state node and its topological relationship with the ecological protection red line boundary. Specifically, this includes: If the boundary of the ecological protection red line is touched, the state branch shall be removed.

[0042] If the ecological protection red line boundary is not touched, the cumulative area of ​​all types of land use under the statistical state node is calculated to obtain the land use index occupancy. Based on the land use index occupancy, the time attribute of the current state node is determined.

[0043] If the planning period is nearing its end and the land use quota does not meet the overall planning control threshold, then this state branch will be removed.

[0044] If the process is in the planning period and the land use quota exceeds the threshold of the overall planning control index, then the state branch is retained and the penalty cost for process violations is calculated.

[0045] Step S203: Extract development and construction cost data from the implementation project dimension, extract ecological value parameters from the planning objective dimension, construct a cost estimation function, and calculate the cost estimate using the cost estimation function. The mathematical expression for the cost estimate is: ; in, As the cost estimate, The land use status can be reached as shown in the current state in the diagram. To accumulate construction costs, To accumulate ecological value loss, and These are preset normalized weighting coefficients used to balance the dimensional differences between capital investment and the loss of ecosystem service value. The total number of overall control indicators for the plan. This serves as an index for the overall control indicators of the plan. The first dimension defined in the planning objective dimension The priority weight of the overall control indicators of the project plan, For the first The target values ​​of the overall control indicators of the project plan, In the current state Next The land use quota occupancy of this indicator For heuristic search weights, The consequences of procedural violations.

[0046] ; To estimate the cost of deviation, it is defined as the weighted Euclidean distance between the current state indicator vector and the planned target indicator vector.

[0047] In this embodiment, the following is set , This embodiment simulates an ecologically-first land spatial planning scenario. Since the damage to ecological resources is often irreversible, the decision-making model assigns a higher weight to cumulative ecological value loss than to cumulative construction costs. This unbalanced weight allocation mechanism makes the search algorithm extremely sensitive to the occupation of high-ecological-value land parcels when calculating path costs. This forces the system to proactively avoid ecologically sensitive areas and prioritize the development of existing land with lower ecological value when generating land use quota allocation schemes, thereby achieving an ecologically-first, green-development-oriented planning orientation at the algorithm level.

[0048] set up There is a significant difference in scale between the overall planning control targets (usually measured in hectares) and construction costs (usually measured in tens of thousands or hundreds of millions of yuan). More importantly, the overall planning control targets are rigid constraints that must be adhered to; any slight deviation should be considered a serious violation. By setting high... The order-of-magnitude penalty weights act like a magnifying glass. Once the current land use quota deviates from the target value, This will instantly amplify such a tiny deviation into a huge estimated cost. This causes the priority of the corresponding path to drop sharply, ensuring that the algorithm can quickly identify and eliminate those inferior solutions that violate the overall control index, guiding the search direction to converge rapidly towards the compliant region that meets the rigid constraints, and greatly improving the optimization efficiency and the compliance rate of the solution.

[0049] When the land use quota exceeds the threshold of the overall planning control index, The value is a weighted product of the excess range and the duration determined by the time attribute of the current state. If the land use quota is less than the threshold of the overall planning control index, then... It is 0.

[0050] The actual costs include construction costs calculated from development and construction cost data and ecological value losses calculated based on ecological value parameters.

[0051] The A* search algorithm is used to search for the path that minimizes the cost estimate. The second state transition edges in the path with the minimum cost estimate are extracted and arranged in time sequence. All the second state transition edges are combined into a state transition sequence, and the state transition sequence is used as the land use index allocation scheme.

[0052] To address the challenge of achieving a dynamic balance between economic and ecological benefits and ensuring compliance in existing land spatial planning decisions while strictly adhering to the dual rigid constraints of ecological protection red line boundaries and overall planning control thresholds, this invention constructs a land use status reachability graph. It utilizes a backtracking search algorithm to traverse the graph and directly prune status branches based on their spatial location and the topological relationship with the red line. Simultaneously, it calculates the penalty costs for process violations caused by exceeding indicator limits, achieving differentiated management of ultimate rigid constraints and process-specific flexible risks. Based on this, a cost estimation function is constructed that integrates accumulated construction costs, accumulated ecological value loss, and estimated deviation costs. The A* search algorithm is then used to find the path with the minimum cost estimate, thereby automatically generating an optimal land use indicator allocation scheme that balances capital investment and ecological service value loss while conforming to temporal logic.

[0053] In a specific embodiment, step S3 includes the following sub-steps: Step S301: Based on the event logic and causal relationships in the state transition sequence, construct a hierarchical time-assigned Petri net. The hierarchical time-assigned Petri net includes a macro-level development time sequence layer and a micro-level land parcel approval layer, specifically including: At the macro-development time sequence layer, corresponding transitions are generated for each development and construction activity in the state transition sequence, and time sequence libraries and directed arcs connecting the transitions are established based on the predecessor and successor relationships of the development and construction activities.

[0054] At the micro-level land parcel approval layer, a corresponding resource repository is generated for each spatial unit node.

[0055] Establish input and output arcs that point from transitions to the corresponding resource repositories.

[0056] Step S302: Define the Token that flows between the time-series repository and the resource repository in the hierarchical time-assigned Petri net as an object carrying data attributes. The data attributes include project priority information, location dependency parameters, geometric coordinates and area values ​​of the plot extracted from the multidimensional ontology network.

[0057] Construct a data interaction interface, traverse the implementation project dimension in the multidimensional ontology network, extract the corresponding land use project instance based on the state transition sequence, and obtain the unique identifier and data attributes of the land use project instance.

[0058] The extracted data attributes are written into the Token, completing the instantiation transformation from a static land use project instance to a dynamic Token.

[0059] It should be noted that the data attributes in this embodiment are fields statically stored in the knowledge graph nodes. The system reads these static fields through an interface and completely copies them into the Petri net's token structure. Therefore, the token is not just a counting marker, but a dynamic container carrying data attributes. This allows the Petri net to directly access the geometric coordinates and priority information of items during the subsequent deduction process in step S4. The geometric coordinates are used to calculate distances, and the priority information is used to determine conflicts, achieving data homology between the knowledge graph semantics and the dynamic deduction.

[0060] The initial land use state is determined based on the state transition sequence, and the token is placed in the corresponding resource vault to generate the token distribution state.

[0061] In this embodiment, in order to solve the data heterogeneity problem between the static land spatial planning knowledge graph and the dynamic hierarchical time-assigned Petri net, the system constructs a data mapping interface.

[0062] Specifically, land use project instances in the multidimensional ontology network are statically stored graph nodes, while tokens circulating in the hierarchical time-assigned Petri net are defined as a structured data carrier. The specific mapping mechanism of the interface is as follows: The interface locates a specific land use project instance within the implementation project dimension of the multidimensional ontology network based on the project index in the state transition sequence. The interface reads the object attributes and data attributes associated with the land use project instance. The system generates a unique Token object and encapsulates the extracted data attributes as internal variables of the Token. For example, the Token's data structure can be represented as Token={ID: instance identifier, Priority: project priority information, Geometry: geometric coordinates}. The Token, encapsulated with attribute information, is placed in the resource repository representing the starting land parcel. In this way, when the Token flows between the time-series repository and the resource repository, the hierarchical timed Petri net can not only count the number of tokens but also read the project priority information carried within the Token in real time in subsequent step S4, thereby supporting priority-based resource occupation conflict handling.

[0063] Step S303: Extract the project plan start time and construction period from the implementation project dimension of the multidimensional ontology network, and use the project plan start time and construction period as the ignition time interval of the corresponding transition.

[0064] The ignition condition for a transition is defined as follows: all input time series locations and resource locations of the transition contain the corresponding Token, and the current inference time is within the ignition time interval.

[0065] To address the problem in existing technologies where data heterogeneity exists between static planning knowledge graphs and dynamic deduction models, preventing traditional Petri nets from utilizing the complex spatial semantic information of land parcels for refined decision-making, this invention innovatively establishes a hierarchical time-assigned Petri net comprising a macro-development time sequence layer and a micro-land parcel approval layer. By constructing a data interaction interface, the circulating tokens are redefined as dynamic containers carrying project priority information, location dependency parameters, and the geometric coordinates of the land parcels. This achieves a complete instantiation conversion from static land use project instances in the multidimensional ontology network to dynamic tokens, ensuring that the data source of the graph semantics and dynamic deduction is the same. This allows the model to directly access and utilize micro-spatial attributes for in-depth logical judgment when evolving based on the ignition time interval, thereby effectively supporting accurate conflict detection and spatiotemporal collaborative deduction in complex planning scenarios.

[0066] In a specific embodiment, step S4 includes the following sub-steps: Step S401: Based on the hierarchical time-assigned Petri net, the state evolution is carried out according to the discrete time step, and the number of tokens in the resource pool and the arrival status of the transition ignition time interval are monitored.

[0067] When the resource library representing the same spatial unit node is found to contain two or more mutually exclusive tokens within the same time step, it is determined to be a resource occupation conflict.

[0068] When it is detected that one of the transitions has reached the ignition time interval, but the transition's input time series library or resource library does not contain a token that meets the ignition conditions, causing the transition to fail to ignite, it is determined to be a time series logic conflict.

[0069] Step S402: For the identified resource occupation conflict, read the data attributes encapsulated inside the conflicting Token and parse the position dependency parameter.

[0070] Based on the location dependency parameter, projects that experience resource occupation conflicts are classified into rigid location projects and flexible location projects.

[0071] The system comes pre-loaded with a location constraint semantic dictionary, which includes a set of strong dependency keywords. and weak dependency keyword set The set of strongly dependent keywords includes on-site expansion, on-site renovation, within the red line, fixed location, and adjacent location; the set of weakly dependent keywords includes new construction, site selection, relocation, and adjustment. In step S302, when constructing the Token data attributes, the unstructured planning text data corresponding to the project is scanned, and the set of strongly dependent keywords appearing in the text is statistically analyzed. and weak dependency keyword set The word frequencies were calculated using a weighted method to determine the positional dependence parameter. Location dependency parameter The mathematical expression is: ; in, To prevent the minimum value where the denominator is zero, Take 0.000001.

[0072] like The location is highly dependent and must be implemented on a specific plot of land, thus the corresponding project is determined to be a rigid site selection project.

[0073] like It has the potential for scheduling between different spatial unit nodes, and is therefore classified as a flexible location project.

[0074] The location intent in the text description is converted into numerical parameters that can be processed by a computer, thereby supporting the differentiated avoidance strategy in S4.

[0075] If the project is a rigid site selection project, a time-based avoidance strategy should be implemented, specifically including: Keeping the resource pool associated with the Token unchanged, calculate the expected release time and postpone the ignition time interval of the Token to after the expected release time of the conflicting resources. If the number of postponements exceeds the preset threshold, mark the corresponding rigid site selection project as infeasible and terminate the simulation.

[0076] If the project is a flexible site selection project, a spatial dimension avoidance strategy will be implemented, specifically including: Search the knowledge graph of territorial spatial planning for alternative spatial unit nodes that have the same planning purpose as the current spatial unit node and are not occupied. Read the data attributes of the alternative spatial unit nodes and calculate the construction suitability score. Redirect the token to the resource storage corresponding to the alternative spatial unit node with the highest construction suitability score and update the token distribution status.

[0077] Calculate the suitability score for construction, redirect the Token to the target node with the highest score, add the spatial unit node to the taboo list of the Token, and prohibit redirection to the spatial unit node in the taboo list again within a preset iteration round.

[0078] Construction suitability score and cost estimate There is a negative correlation, meaning that the lower the estimated cost, the higher the suitability score for construction.

[0079] For timing logic conflicts, retrieve the upstream transition corresponding to the missing token, calculate the remaining minimum operation time of the upstream transition, and if the current simulation time allows for delay, automatically postpone the ignition time interval of the current transition. If the current simulation time does not allow for delay, generate a critical path warning and compress the duration parameter of the upstream transition.

[0080] Furthermore, the spatial dimension avoidance strategy in step S4 also includes a dynamic feedback verification sub-step, used to ensure that the dynamically adjusted planning scheme meets the ecological protection red line boundary and planning overall control index threshold of step S2. The dynamic feedback verification sub-step specifically includes: Step S4021: Before redirecting the Token to the resource repository corresponding to the alternative spatial unit node, the cost estimation function in step S203 is called. Using the data attributes of the replacement spatial unit nodes as input, the local cost estimate under the replacement state is calculated. .

[0081] Step S4022: Calculate the local cost estimate. Cost estimates of corresponding nodes in the original planned path The deviation rate between them is determined, and based on the geometric coordinates of the alternative spatial unit nodes, step S202 is performed again to determine the topological relationship between the spatial location of the state node and the boundary of the ecological protection red line.

[0082] Step S4023: If the replacement spatial unit node does not touch the ecological protection red line boundary, and the deviation rate is less than or equal to the preset dynamic tolerance threshold. If the spatial dimension avoidance strategy is effective, the token distribution state is updated and the deduction continues.

[0083] If the alternative spatial unit node touches the boundary of the ecological protection red line, or the deviation rate exceeds the dynamic tolerance threshold. If so, the current local avoidance plan is deemed invalid.

[0084] At this point, the system generates a global replanning instruction, which feeds back the conflicting tokens and their associated time attributes to the optimization module. This triggers the optimization module to re-execute step S2 after eliminating the currently infeasible paths, and generate a new land use index allocation scheme.

[0085] Preset dynamic tolerance threshold It is 5%.

[0086] Step S403: Update the current state of the hierarchical timed Petri net by the delayed ignition time interval or the redirected token distribution state, and return to execute step S401 until all resource occupation conflicts and timing logic conflicts are eliminated, and output the decision result.

[0087] It is important to note that this invention employs a two-tiered decision-making architecture—first static spatial optimization, then dynamic temporal deduction—to address the computational challenges of solving spatial complexity and temporal concurrency in land spatial planning within a single model. Specifically, the backtracking search algorithm and A* search algorithm in step S2 focus on resolving the optimal allocation of a single land parcel's function under static constraints of spatial resource and control rule dimensions, i.e., determining what is most cost-effective to do with this land. While the land use index allocation scheme generated at this stage satisfies the constraints in terms of total index quantity, it does not yet consider the temporal competition for the same spatial resource when multiple projects are implemented concurrently. The hierarchical time-assigned Petri net introduced in steps S3 and S4 focuses on resolving the temporal game problem of the implementation project dimension, i.e., determining when these projects can be implemented without conflict. Since multiple projects may request the same land parcel (resource repository) in actual planning implementation, and the actual project start time is dynamically affected by the approval process (i.e., changes), such non-linear concurrent conflicts cannot be directly predicted through the static graph search in step S2. Therefore, it is necessary to expose potential resource occupancy conflicts through discrete event simulation in step S4, and to make time-axis corrections or spatial fine-tunings to the static scheme generated in S2.

[0088] To address the problem that existing planning and simulation technologies cannot automatically resolve complex resource occupation conflicts and temporal logic conflicts based on project characteristics, this invention innovatively distinguishes between rigid and flexible site selection projects by reading the location dependency parameters carried within the token during the state evolution process of a hierarchical time-assigned Petri net. It replaces spatial unit nodes with semantic association retrieval from a land spatial planning knowledge graph, implementing a spatial dimension avoidance strategy based on redirection for flexible projects and avoiding infinite loops through a taboo list. For rigid projects, a temporal dimension avoidance strategy based on a delayed ignition time interval is implemented. This data attribute-driven, dynamic, differentiated avoidance mechanism successfully solves the problem of traditional methods lacking adaptive adjustment capabilities when facing multiple conflicts, effectively eliminating potential risks in planning implementation and ensuring that the final output decision result has both temporal feasibility and spatial compliance.

[0089] Example 2, refer to Figure 2 This paper presents a knowledge graph-based decision-making system for land spatial planning, which includes a mapping module, an optimization module, a modeling module, and a deduction module.

[0090] The mapping module is used to acquire spatial vector data and unstructured planning text data of land and space planning, parse the business logic of land and space planning and construct a multidimensional ontology network, map the spatial vector data and unstructured planning text data into the multidimensional ontology network, and generate a knowledge graph of land and space planning.

[0091] The optimization module is used to construct a land use status reachability map based on the land spatial planning knowledge graph, and to traverse the land use status reachability map using a backtracking search algorithm to obtain a land use index allocation scheme.

[0092] The modeling module is used to establish a hierarchical time-assigned Petri net based on the land use index allocation scheme.

[0093] The inference module is used to perform discrete-time step state evolution based on hierarchical time-assigned Petri nets. It detects resource occupation conflicts and temporal logic conflicts through hierarchical time-assigned Petri nets, adjusts projects based on project priority information extracted from the land spatial planning knowledge graph, and outputs decision results.

[0094] To address the common technical challenges in existing land spatial planning decision-making systems, such as the difficulty in deeply integrating spatial vector data with unstructured planning text data, and the lack of dynamic extrapolation capabilities for the implementation process in traditional static planning models, which frequently lead to resource occupation conflicts and temporal logic conflicts when planning schemes are implemented, this invention constructs a system architecture integrating a mapping module, an optimization module, a modeling module, and an extrapolation module. First, the mapping module parses business logic to construct a multi-dimensional ontology network, breaking down data barriers to generate a land spatial planning knowledge graph. Then, the optimization module uses backtracking search algorithms and pruning strategies in the land use status reachability graph to efficiently calculate land use index allocation schemes that meet rigid constraints. The system's unique modeling module instantiates static schemes into hierarchical time-assigned Petri nets, providing a dynamic simulation environment for the extrapolation module. The extrapolation module, through discrete time step state evolution, can accurately capture conflicts that are difficult to detect from a static perspective and adaptively adjust in conjunction with project priority information in the graph. This effectively compensates for the shortcomings of traditional systems that emphasize planning but neglect implementation, realizing intelligent decision support across the entire chain from knowledge integration and path optimization to dynamic extrapolation.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A knowledge graph-based land spatial planning decision-making method, characterized in that, Includes the following steps: Step S1: Obtain spatial vector data and unstructured planning text data of the land spatial planning, parse the business logic of the land spatial planning and construct a multi-dimensional ontology network, map the spatial vector data and unstructured planning text data into the multi-dimensional ontology network, and generate a land spatial planning knowledge graph. Step S2: Based on the land spatial planning knowledge graph, construct a land use status reachability map, and use a backtracking search algorithm to traverse the land use status reachability map to obtain a land use index allocation scheme; Step S3: Based on the land use index allocation scheme, establish a hierarchical time-assigned Petri net; Step S4: Based on the hierarchical time-assigned Petri net, perform state evolution, detect resource occupation conflicts and temporal logic conflicts, and output the decision result after eliminating resource occupation conflicts and temporal logic conflicts.

2. The knowledge graph-based land spatial planning decision-making method as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Analyze the business logic of land and space planning, extract business concept knowledge points, define class nodes, object attributes and data attributes, and construct a multi-dimensional ontology network; The business logic includes land space survey, land space planning, land use control land classification guidelines, land space planning data submission standards, and business process specifications for administrative approval and planning implementation management. The class nodes include spatial resource dimension, planning goal dimension, control rule dimension, implementation project dimension and evolution process dimension. The spatial resource dimension includes natural geographic entities and planned land parcels. The planning goal dimension includes population indicators, economic indicators and ecological indicators. The control rule dimension includes the three zones and three lines and legal and regulatory constraints. The implementation project dimension includes development and construction projects. The evolution process dimension includes land parcel status change events. The object attributes include inclusion and adjacency relationships describing spatial topology, constraint relationships describing rule restrictions, occupancy relationships describing project implementation, contribution relationships describing indicator achievement, and temporal evolution relationships describing state changes. The data attributes include the geometric coordinates and area values ​​of the land parcel, the semantic description string of the planning text, the values ​​of the control indicators, the time window parameters of the project, and the status identifier code of the approval process.

3. The knowledge graph-based land spatial planning decision-making method as described in claim 2, characterized in that, Step S1 further includes the following sub-steps: Step S102: parse the polygon vertex coordinate sequence and boundary information in the spatial vector data, generate spatial unit nodes, and extract the adjacency and inclusion relationships between spatial unit nodes based on coordinate operations to generate spatial topological edges connecting different spatial unit nodes. An entity recognition model is constructed, which includes a BERT embedding layer, a bidirectional long short-term memory network layer, and a conditional random field layer. The unstructured planning text data is input into the entity recognition model, and the entity labeling sequence is output. The entity labeling sequence includes spatial entity labels, control indicator labels, constraint operator labels, and threshold value labels. Based on dependency parsing, a syntactic parse tree is constructed for the entity annotation sequence. Semantic quadruples are extracted based on the dependency relationships between entities. The semantic quadruples include subject, index, operator, and value. Based on the semantic quadruple, class nodes in the multidimensional ontology network are retrieved, corresponding control rule nodes with constraint logic are generated, and the values ​​in the semantic quadruple are assigned to the data attributes of the control rule nodes, and an associated index edge is established from the spatial unit node to the control rule node. Step S103: Map the spatial unit nodes and control rule nodes to the multidimensional ontology network, traverse the temporal logic definition in the evolution process dimension, instantiate directed constraint edges between spatial unit nodes of land use change, and construct a knowledge graph of national land spatial planning.

4. The knowledge graph-based land spatial planning decision-making method as described in claim 3, characterized in that, Step S2 includes the following sub-steps: Step S201: Taking the current land use type of the spatial unit node as the initial state, and according to the planning evolution rules extracted from the control rule dimension, select variable spatial unit nodes with the potential for land use type change. The land use type change operation of the variable spatial unit node is defined as the first state transition edge. Based on the evolution of the variable spatial unit node from the initial state to the planned target land use type, a land use state reachability graph is constructed, wherein the graph nodes of the land use state reachability graph represent the land use type distribution state of the variable spatial unit node.

5. The knowledge graph-based land spatial planning decision-making method as described in claim 4, characterized in that, Step S2 further includes the following sub-steps: Step S202 involves extracting the ecological protection red line boundary from the control rules dimension and extracting the planning overall control index threshold from the planning target dimension. A backtracking search algorithm is used to traverse the land use status reachability graph from the initial state. For each graph node on the traversal path, each graph node is defined as a state node. The spatial location of the state node and its topological relationship with the ecological protection red line boundary are determined. Specifically, this includes: If the boundary of the ecological protection red line is touched, the state branch shall be removed; If the ecological protection red line boundary is not touched, the cumulative area of ​​all types of land use under the statistical state node is calculated to obtain the land use index occupancy. Based on the land use index occupancy, the time attribute of the current state node is determined. If the planning period is nearing its end and the land use quota does not meet the threshold of the overall planning control index, then this state branch will be removed. If the process is in the planning period and the land use quota exceeds the threshold of the overall planning control index, then the state branch is retained and the penalty cost for process violations is calculated. Step S203: Extract development and construction cost data from the implementation project dimension, obtain the cumulative construction cost based on the development and construction cost data, extract ecological value parameters from the planning target dimension, obtain the cumulative ecological value loss based on the ecological value parameters, construct a cost estimation function, and calculate the cost estimate using the cost estimation function. The mathematical expression for the cost estimate is: ； in, As the cost estimate, The land use status can be reached as shown in the current state in the diagram. To accumulate construction costs, To accumulate ecological value loss, and These are the preset normalized weighting coefficients. This serves as an index for the overall control indicators of the plan. The priority weight of the i-th overall planning control indicator defined in the planning objective dimension. For the first The target values ​​of the overall control indicators of the project plan, In the current state Next The land use quota occupancy of this indicator For heuristic search weights, The consequences of procedural violations; The A* search algorithm is used to search for the path that minimizes the cost estimate. The second state transition edges in the path with the minimum cost estimate are extracted and arranged in time sequence. All the second state transition edges are combined into a state transition sequence, and the state transition sequence is used as the land use index allocation scheme.

6. The knowledge graph-based land spatial planning decision-making method as described in claim 5, characterized in that, Step S3 includes the following sub-steps: Step S301: Based on the event logic and causal relationships in the state transition sequence, construct a hierarchical time-assigned Petri net. The hierarchical time-assigned Petri net includes a macro-level development time sequence layer and a micro-level land parcel approval layer, specifically including: In the macro-development time sequence layer, corresponding transitions are generated for each development and construction activity in the state transition sequence, and time sequence libraries and directed arcs connecting the transitions are established based on the predecessor and successor relationships of the development and construction activities. In the micro-plot approval layer, a corresponding resource repository is generated for each spatial unit node; Establish input and output arcs that point to the corresponding resource library from the transitions.

7. The knowledge graph-based land spatial planning decision-making method as described in claim 6, characterized in that, Step S3 further includes the following sub-steps: Step S302: Define the Token that flows between the time series repository and the resource repository in the hierarchical time-assigned Petri net as an object carrying data attributes. The data attributes include the project unique identifier code extracted from the multidimensional ontology network, project type priority information, location dependency parameters, and plot geometric coordinates and area values. Construct a data interaction interface, traverse the implementation project dimension in the multidimensional ontology network, extract the corresponding land use project instance according to the state transition sequence, obtain data attributes, write the obtained data attributes into the Token, determine the initial land use state according to the state transition sequence, place the Token in the corresponding resource repository, and generate the Token distribution state. Step S303: Extract the planned start time and construction period of the project from the implementation project dimension of the multidimensional ontology network, and use the planned start time and construction period of the project as the ignition time interval of the corresponding transition. The ignition condition for a transition is defined as follows: all input time series locations and resource locations of the transition contain the corresponding Token, and the current inference time is within the ignition time interval.

8. The knowledge graph-based land spatial planning decision-making method as described in claim 7, characterized in that, Step S4 includes the following sub-steps: Step S401: Based on the hierarchical time-assigned Petri net, perform state evolution according to discrete time steps, and monitor the number of tokens in the resource pool and the arrival status of the transition ignition time interval. When it is detected that the resource library representing the same spatial unit node contains two or more mutually exclusive tokens within the same time step, it is determined to be a resource occupation conflict. When it is detected that one of the transitions has reached the ignition time interval, but the input timing library or resource library of the transition does not contain a token that meets the ignition conditions, causing the transition to fail to ignite, it is determined to be a timing logic conflict.

9. A knowledge graph-based land spatial planning decision-making method as described in claim 8, characterized in that, Step S4 further includes the following sub-steps: Step S402: For the identified resource occupation conflict, read the data attributes encapsulated inside the conflicting Token and parse the position dependency parameter; Based on the location dependency parameter, projects that have resource occupation conflicts are divided into rigid location projects and flexible location projects. If the project is a rigid site selection project, a time-based avoidance strategy should be implemented, specifically including: Keeping the resource library associated with the Token unchanged, calculate the expected release time, and postpone the ignition time interval of the Token to after the expected release time of the conflicting resources. If the number of postponements exceeds a preset threshold, mark the corresponding rigid site selection project as infeasible and terminate the simulation. If the project is a flexible site selection project, a spatial dimension avoidance strategy will be implemented, specifically including: Search the knowledge graph of the national land spatial planning for alternative spatial unit nodes that have the same planning purpose as the current spatial unit node and are not occupied. Read the data attributes of the alternative spatial unit nodes and calculate the construction suitability score. Redirect the Token to the resource storage corresponding to the alternative spatial unit node with the highest construction suitability score and update the Token distribution status. Calculate the suitability score for construction. The suitability score is negatively correlated with the cost estimate. Redirect the token to the target node with the highest score. Add the spatial unit node to the token's taboo list. Within a preset number of iterations, redirect the token to the spatial unit node in the taboo list again. For timing logic conflicts, retrieve the upstream transition corresponding to the missing token, calculate the remaining minimum operation time of the upstream transition, and if the current simulation time allows for delay, automatically postpone the ignition time interval of the current transition; if the current simulation time does not allow for delay, generate a critical path warning and compress the duration parameter of the upstream transition. Step S403: Update the current state of the hierarchical timed Petri net by the delayed ignition time interval or the redirected token distribution state, and return to execute step S401 until all resource occupation conflicts and timing logic conflicts are eliminated, and output the decision result.

10. A knowledge graph-based land spatial planning decision-making system, applied in a knowledge graph-based land spatial planning decision-making method as described in any one of claims 1-9, characterized in that, It includes a mapping module, an optimization module, a modeling module, and a deduction module; The mapping module is used to acquire spatial vector data and unstructured planning text data of land spatial planning, parse the business logic of land spatial planning and construct a multidimensional ontology network, map the spatial vector data and unstructured planning text data to the multidimensional ontology network, and generate a land spatial planning knowledge graph. The optimization module is used to construct a land use status reachability map based on the land spatial planning knowledge graph, and to traverse the land use status reachability map using a backtracking search algorithm to obtain a land use index allocation scheme. The modeling module is used to establish a hierarchical time-assigned Petri net based on the land use index allocation scheme; The inference module is used to perform discrete-time step state evolution based on the hierarchical time-assigned Petri net, detect resource occupation conflicts and temporal logic conflicts through the hierarchical time-assigned Petri net, adjust the project based on the project priority information extracted from the land spatial planning knowledge graph, and output the decision results.

Images