Land space planning management method and system based on big data

Abstract

The invention provides a land space planning management method and system based on big data. The method comprises the following steps: acquiring original planning graph data; constructing an approval behavior event sequence; based on the prediction of the improved graph neural network, generating a prediction graph and a future purpose prediction result according to the path propagation of the improved graph neural network; according to the prediction map and a future use prediction result, identifying a use conflict trend possibly occurring in the next time period in the territorial space, quantifying the planning tension change intensity between the regions, and positioning a potential high-risk land parcel set; and according to the high-risk land parcel set, in combination with a future use prediction result, a use violation penalty value, an approval state vector, planning tension change intensity and original planning graph data, executing structural rule matching, and generating a structured management suggestion set.

Description

Technical Field

[0001] The present invention belongs to the technical field of territorial spatial planning management, and in particular relates to a territorial spatial planning management method and system based on big data. Background Art

[0002] As a basic institutional tool for coordinating natural resource utilization and national economic and social development, the scientificity, dynamics, and implementation effectiveness of territorial spatial planning directly affect the sustainability of urban development and the level of ecological environment protection. In recent years, with the deepening of the reform of the national natural resource governance system, territorial spatial planning has gradually realized the transformation from "single-use control" to "multi-objective collaborative governance". However, many technical limitations have emerged in the actual operation of this system. First, at the level of planning data management, various planning layers (including overall territorial spatial planning, detailed planning, regulatory planning, etc.) coexist at multiple scales and levels, lacking effective data fusion and logical consistency verification means, often resulting in problems such as overlaps, contradictions, and fuzzy boundaries in spatial use arrangements, especially obvious in the planning coordination at different levels such as provinces, cities, and counties. Second, in terms of planning implementation and dynamic update, the traditional territorial spatial planning management mode often relies on manual maintenance and periodic revision, and it is difficult to reflect the actual implementation of policies such as actual construction and land use approval in real time, resulting in a deviation between the planning layer and the actual use situation. For example, although the use of a certain plot has been approved for change, the original use code is still retained in the upper-level planning map, causing data distortion and management blind spots.

[0003] In addition, most existing systems are static data platforms. Although they have functions such as overlay analysis and compliance review, they essentially still remain in the paradigm of "graphical inspection + manual intervention", and cannot realize the prediction of the evolution trend of the planning map, the structural identification of the impact of policy actions, as well as the automatic diagnosis and feedback of conflict problems. Some existing studies have tried to introduce means such as remote sensing identification, knowledge graphs, and intelligent segmentation for land development status monitoring. However, these methods mostly focus on the "current situation identification" level, lack deep logical coupling with the planning map, and often ignore the internal driving force of dynamic data (such as approval logs, land transfer behaviors) on planning adjustment, and are difficult to meet the management requirements of the current territorial spatial planning of "data closed-loop, management closed-loop, and feedback closed-loop".

[0004] Therefore, how to construct a territorial spatial planning management system that can automatically identify cross-scale conflicts, dynamically respond to actual policy implementation behaviors, and possess the capabilities of use prediction and map evolution has become a key challenge in the current natural resource management technology system. Summary of the Invention

[0005] The purpose of the present invention is to design a territorial spatial planning management method and system based on big data to effectively solve the above problems.

[0006] In order to achieve the above object, a first aspect of the present invention provides a land space planning management method based on big data, the method comprising the following steps:

[0007] Step 1, obtain the original planning map data, divide the planning area of the original planning map data layer into non-overlapping basic plot units, analyze the basic plot units according to the spatial contact degree between the plot units and the distance of the hierarchical relationship, and obtain a standardized spatial planning map structure G = (V, E); wherein, in the spatial planning map structure G = (V, E), V is a node set, each node represents a plot unit, E is an edge set, and the edge represents the spatial contact degree and the connection relationship of the hierarchical relationship between the plot units, and the conflict intensity or coupling relationship is expressed by weight;

[0008] Step 2: Construct an approval behavior event sequence, where each behavior a in the event sequence k Defined as triple 9v i ,t k ,u k ), representing the plot unit v i At time t k Approved for use k , according to the triple (v i ,t k ,u k ) constructs an approval status vector for each plot unit, and designs a use violation penalty value based on the intensity of the cumulative use deviation of the plot unit. Finally, the approval status vector and the use violation penalty value are added to each node of the spatial planning graph structure G = (V, E) to form a dynamic graph G t =(V t ,E), where V t Represents a dynamic set of plots at time point t;

[0009] Step 3: According to the dynamic graph G t =(V t ,E) Performing prediction based on the improved graph neural network, ,generating a prediction graph and future usage prediction result according to the path propagation of the improved graph neural network;

[0010] Step 4: Identify the trend of possible use conflicts in the national land space in the next period based on the prediction map and the future use prediction results, quantify the intensity of planning tension changes between regions, and locate potential high-risk plots;

[0011] Step 5: Based on the high-risk land parcel set, combined with future use prediction results, use violation penalty values, approval status vectors, planning tension change intensity, and original planning map data, perform structural rule matching to generate a structured management recommendation set.

[0012] Further, the step 1 further includes: preprocessing the original planning map data; the preprocessing includes coordinate system unification, use classification standard mapping, and attribute field normalization;

[0013] In the graph structure G=(V, E), each plot unit will serve as a node in the graph structure, including: the basic information of the plot and its affiliated planning level; the spatial contact degree and hierarchical relationship between the plot units are used as spatial / logical connection relationships to construct edges. Among them, if two plot units are completely non-contact in space, even in the same layer, no connection relationship will be established; if they overlap in space and come from different hierarchical layers, an edge will be constructed and the weight of the edge will be calculated, and this weight combines the coincidence degree and hierarchical difference between the two plot units.

[0014] Further, according to the triple (v i , t k , u k ), an approval status vector is constructed for each plot unit, representing the dynamic behavior state of the node at time t, and the calculation is as follows:

[0015]

[0016] Among them, is the dynamic state value of plot v i at time t; is the time decay factor, γ∈(0, 1), reflecting that the longer the historical approval time, the smaller the impact on the current state; c k is the credibility weight of the approval behavior a k ; δ(u k ≠u i ) is the use deviation indicator function. If the behavior use u k is inconsistent with the original planned use u i of the plot, it is 1, otherwise it is 0; K is the length of the historical approval window.

[0017] Further, in the dynamic graph G t =(V t , E), V t represents the set of dynamic plots at time point t, and each node state has approval history memory, deviation recognition ability, and behavior density representation.

[0018] Further, based on the dynamic graph G t =(V t , E), a prediction based on an improved graph neural network is performed, and according to the path propagation of the improved graph neural network, a prediction result is generated. The specific calculation is as follows:

[0019]

[0020] Among them, s i,u represents the predicted path, is the behavior driving factor, and η1 and η2 are the weight coefficients of the node and use violation penalty values. represents the use activity drive of this node That is, the more frequent the approval, the more likely the use change will occur. It means that the more serious the historical deviation is, the more the continuation of the original use should be avoided, so as to promote the transformation of the predicted use. is the use violation penalty value; is the use co-evolution propagation factor, which represents the proportion of adjacent nodes that are consistent with the predicted use, indicating the degree of influence of the plot by the surrounding uses. j is node j, is the set of adjacent nodes, w ij is the edge weight, is the approval use status of node j; is the use inertia penalty factor, and ρ is the use deviation penalty intensity hyperparameter; I(u = u i ) indicates whether it is the original use; Penalty(u) represents the continuation cost of this use.

[0021] Furthermore, identifying the possible use conflict trends in the territorial space in the next time period according to the predicted map and the future use prediction results, quantifying the change intensity of the planning tension between regions, and locating the set of potential high-risk plots specifically include:

[0022] Taking whether there is an offset from the planned use as a constraint item to perform time-series state update calculation, generating a predicted tension score, which represents the spatial tension between the plot unit and the adjacent plot units in the future use state;

[0023] Aggregating the tensions of each node, identifying the tension concentration points, and generating the total predicted tension value of the current plot unit. The larger the value, the more it represents the boundary of the area under high use variation / conflict pressure;

[0024] According to the tension threshold set by experience, mark the nodes with the total predicted tension value greater than the tension threshold as high-risk plots to form a set of risk regions.

[0025] Furthermore, whether there is an offset from the planned use is determined by an indicator function.

[0026] Furthermore, the structural rule matching specifically includes:

[0027] Suggestions for use revision:

[0028] If the following conditions are all satisfied:

[0029] The predicted use is inconsistent with the planned use;

[0030] The cumulative value of the use deviation is higher than the set threshold;

[0031] The total predicted tension value is medium or above;

[0032] Then output: It is recommended to revise the planned use of the current plot unit from the original use to the predicted result of the future use, reason: The planned use has been invalid for a long time, and the predicted trend is stable;

[0033] Suggestions for use approval restrictions:

[0034] If the current predicted use of the current plot unit is consistent with the planned use, but there are use change trends in ≥2 nodes in the adjacent plots and the tension is high, then output:

[0035] This plot is in the edge transition zone of the structural use. It is recommended to conduct joint review and record control on the approval application, and give priority to guiding it to be consistent with the upper level Figure 1 consistent;

[0036] Suggestions for the conflict use buffer zone:

[0037] If the current plot unit is adjacent to at least two different types of plots and the conflict tension score > 0.7, then output:

[0038] It is recommended to delimit a use buffer or transition area to resolve the high-conflict boundary and avoid future layer breaks.

[0039] Furthermore, the conflict tension score is a similarity calculation.

[0040] In another embodiment of the present invention, a land spatial planning management system based on big data is provided. The system includes:

[0041] A planning map segmentation unit, configured to obtain the original planning map data, divide the planning area of the original planning map data layer into non-overlapping basic plot units, analyze the basic plot units according to the spatial contact degree and the distance of the hierarchical relationship between the plot units, and obtain a standardized spatial planning map structure G=(V, E); where, in the spatial planning map structure G=(V, E), V is the node set, each node represents a plot unit, E is the edge set, and the edge represents the connection relationship of the spatial contact degree and the hierarchical relationship between the plot units, and its conflict intensity or coupling relationship is expressed by a weight;

[0042] A dynamic node analysis unit, configured to construct an approval behavior event sequence, and each behavior a of the event sequence k is defined as a triple (v i , t k , u k ), indicating the plot unit v iAt time t k is approved for use u k , according to the triple (v i , t k , u k Construct an approval status vector for each plot unit, design a use violation penalty value based on the intensity of the cumulative use deviation of the plot unit, and finally add the approval status vector and the use violation penalty value to each node of the spatial planning map structure G=(V, E) to form a dynamic graph G t =(V t , E), where V t represents the set of dynamic plots at time point t;

[0043] A graph neural network prediction unit for making predictions based on the improved graph neural network according to the dynamic graph G t =(V t , E), and generating a prediction graph and a future use prediction result according to the path propagation of the improved graph neural network;

[0044] A conflict quantification unit for identifying the possible use conflict trends in the national land space in the next time period according to the prediction graph and the future use prediction result, quantifying the change intensity of the planning tension between regions, and locating the set of potential high-risk plots;

[0045] A planning management output unit for performing structural rule matching according to the set of high-risk plots, combining the future use prediction result, the use violation penalty value, the approval status vector, the change intensity of the planning tension, and the original planning map data, and generating a structured set of management suggestions.

[0046] The beneficial technical effects of the present invention are at least as follows:

[0047] The present invention precisely aims at the core pain points in the above-mentioned existing technologies and proposes a dynamic national land space planning management method system integrating identification, response and prediction. By establishing the interaction relationship between multi-scale planning map structures, this method can effectively quantify the conflict intensity between planning uses at different levels and identify the structural tension areas that may affect the spatial order; at the same time, construct a data-driven mechanism for approval behaviors and policy implementation processes, and use real-occurring land approval, project implementation, policy adjustment and other behaviors as dynamic inputs to promote the update of planning uses and the synchronous adjustment of management layers; on this basis, further introduce the ability of evolutionary deduction of spatial structures, combine historical evolution trends and behavior influencing factors, establish a prediction mechanism for future use change trends, and provide decision-making support for planning compilation, dynamic revision and planning risk warning.

[0048] Compared with existing methods, the present invention can not only systematically discover and diagnose spatial conflicts in planning data, but also has the ability to respond to the impact of management behaviors and the forward-looking judgment ability of the evolution trend of future uses, thus realizing the truly dynamic control and intelligent evolution of territorial spatial planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on the following drawings without creative efforts.

[0050] Figure 1 It is a flowchart of a method for territorial spatial planning management based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] The embodiments of the present invention are described in detail below. The examples of the embodiments are shown in the accompanying drawings, in which the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention.

[0052] In one or more embodiments, as Figure 1 shown, a method for territorial spatial planning management based on big data according to the present invention is disclosed. The method includes the following steps:

[0053] Step 1: Obtain the original planning map data, divide the planning area of the original planning map data layer into non-overlapping basic plot units, analyze the basic plot units according to the spatial contact degree and the distance of the hierarchical relationship between the plot units, and obtain a standardized spatial planning map structure G=(V, E); wherein, in the spatial planning map structure G=(V, E), V is a node set, each node represents a plot unit, E is an edge set, the edge represents the connection relationship of the spatial contact degree and the hierarchical relationship between the plot units, and its conflict intensity or coupling relationship is expressed by a weight.

[0054] Specifically, in the whole inventive method, this step serves as the starting point, responsible for standardizing the planning map data from different levels and sources and performing structured modeling, and outputting a unified spatial graph structure G=(V, E). This graph structure should not only be able to express the spatial and planning hierarchical relationships between plots, but also support subsequent dynamic behavior embedding and use evolution modeling, so as to enable the entire system to have a "dynamically inferable" spatial planning data basis.

[0055] In actual engineering, the original data often comes from a wide range of sources, including provincial master plans, municipal regulatory detailed plans, and county-level special plans. These data often have problems such as inconsistent coordinate systems, inconsistent field meanings, different use classification standards, and slight boundary misalignments. In this step, all original layers need to be unified in coordinate system, mapped to the use classification standard, normalized in attribute fields, and the planning area is divided into non-overlapping basic plot units. Each plot unit will be used as a node v in the graph structure. i , which not only contains the basic information of the plot (such as geographical boundary, use code u i , area a i ), but also records the planning level l i to which it belongs. For example, a forest reserve from the provincial ecological red line layer may spatially overlap with an expanded industrial area in the municipal construction land layer. After the two overlapping areas are merged, two nodes will be generated respectively, marked with different levels, and a connection relationship will be established between them. The level markers here (such as l i = 1 represents provincial level, l j = 2 represents municipal level) are not only used as influencing factors in subsequent modeling, but also provide a basis for the subsequent judgment of the logic of "who should obey whom".

[0056] The edges e of the graph structure ij are constructed based on two factors: the degree of spatial contact between plots (such as whether the boundaries are adjacent or partially overlapping) and the proximity of the hierarchical relationship. If two plots are completely non-contact in space, even in the same layer, no connection relationship will be established; while if they overlap in space and come from different hierarchical layers, an edge will be constructed and the weight w ij of the edge will be calculated. This weight combines the degree of overlap and the hierarchical difference between the two plots:

[0057] w ij = α·Overlap(v i , v j ) + β·|l i - l j |

[0058] Where:

[0059] w ij represents the weight of the connection edge between plots v i and v j , and is used to measure the strength of their spatial connection;

[0060] Overlap(v i , v j ) is the spatial overlap degree between the two plots, and the calculation formula is the ratio of the intersection area of the two to the smaller area;

[0061] li and l j are respectively the planning levels where plots v i and v j are located (e.g., provincial level = 1, municipal level = 2);

[0062] |l i -l j | represents the difference between two planning levels;

[0063] α and β are weight parameters set empirically to control the influence of the overlap degree and level difference on the edge weight (e.g., α = 0.7, β = 0.3).

[0064] For example, if there is an intersection of 4000 square meters between a municipal construction land plot with an area of 8000 square meters and a provincial ecological red line area with an area of 12000 square meters, and the level difference is 1, then Overlap = 4000 / 8000 = 0.5, and the edge weight is 0.7·0.5 + 0.3·1 = 0.65. This weight will be used in subsequent steps to identify the propagation relationship of land use evolution or the intensity of potential conflicts between plots.

[0065] After the above processing, the finally obtained graph structure G=(V, E), where each node v i represents a plot, and the edge e ij represents the spatial / logical connection relationship, and its conflict intensity or coupling relationship is expressed through the weight w ij This modeling method is different from the layer overlay in traditional GIS. It not only supports the logical modeling of land use but also supports subsequent behavior-driven updates and trend predictions.

[0066] Step 2: Construct the approval behavior event sequence. Each behavior a k in the event sequence is defined as a triple (v i , t k , u k ), which means that the plot unit v i is approved for the land use u k at time t k . According to the triple (v i , t k , u k ), construct an approval status vector for each plot unit, design a land use violation penalty value based on the intensity of the cumulative land use deviation of the plot unit, and finally add the approval status vector and the land use violation penalty value to each node of the spatial planning graph structure G=(V, E) to form a dynamic graph G t = (V t , E), where V t represents the set of dynamic plots at time point t.

[0067] Specifically, in Step 1, the present invention has constructed a standardized spatial planning map structure G = (V, E), where each node v i represents a basic plot unit, and the node attributes include its static planning use u i , level l i and area a i etc. In reality, the actual implementation of land use often does not strictly conform to the planning map. Some plots may change frequently due to multiple approvals in the short term, while others may remain unstarted for a long time. The timeliness and superposition of approval behaviors are the key sources of the "deviation between planning and reality". Therefore, relying solely on the layer is not enough to support the truly dynamic planning management logic, and it is necessary to introduce approval behaviors as the dynamic evolution driving force.

[0068] The present invention first constructs an approval behavior event sequence Each behavior a k is defined as a triple (v i , t k , u k ), indicating that plot v i is approved for use u k at time t k . These data usually come from government business systems such as land supply databases, land use approval systems, and project approval and record-filing, and can be matched to the node v i constructed in Step 1 through spatial coordinates.

[0069] Next, the present invention constructs an approval status vector i for each plot unit v to represent the dynamic behavior state of the node at time t. This state comprehensively considers the following three types of factors:

[0070] The behavior density (activity) of the last K approval behaviors;

[0071] The use deviation between the current approval use and the planned use;

[0072] The credibility weighting of the source of the approval behavior (for example, the approval weight of the provincial department is higher, and the weight of the township department is lower).

[0073] Model these three elements into a composite behavior embedding expression:

[0074]

[0075] Where:

[0076] The dynamic state value of plot v i at time t;

[0077] The time decay factor, γ∈(0,1), reflects that the longer the historical approval time is, the smaller the impact on the current state is;

[0078] c k : The credibility weight of the approval behavior a k is manually set. For example, it is 1.0 for provincial level, 0.7 for municipal level, and 0.5 for county and district level;

[0079] δ(u k ≠u i ): The use deviation indication function. If the use u of the behavior k is inconsistent with the original planned use u of the land parcel i , it is 1, otherwise it is 0;

[0080] K: The length of the historical approval window (such as the recent 3 years).

[0081] For example: If the original plan of a certain land parcel is "agricultural land", but it has been approved as "industrial" and "commercial and residential" by the city and county respectively within the recent 3 years, and the times are t-1 and t-2, then its If γ = 0.9, then The higher this value is, it indicates that the land parcel has frequent approvals and serious use deviations, and it is an object that needs to be focused on in the future.

[0082] In order to further enhance the ability to identify "planning failure", the present invention defines a dynamic use penalty term to represent the intensity of the cumulative use deviation of the land parcel:

[0083]

[0084] Among them:

[0085] The use violation penalty value of the land parcel v i at time t;

[0086] λ: The coefficient that adjusts the influence of the use deviation penalty;

[0087] The use inconsistency indication function;

[0088] (t - t k ): The time span. The longer the deviation is, the more it accumulates, and the higher the penalty is.

[0089] This value will be introduced as a regularization term into the model loss function in the subsequent prediction module to guide the model to identify the trend of long-term deviation from the planned use and avoid blindly continuing short-term approval behaviors.

[0090] Finally, the present invention adds two dynamic attributes to each node v of the original graph G=(V, E) i : (Behavior Embedding Index) and (Use Violation Penalty Value) to form a dynamic graph G t =(V t , E), where V t represents the set of dynamic plots at time point t, and each node state has approval history memory, deviation recognition ability, and behavior density representation.

[0091] Step 3: Perform prediction based on the improved graph neural network according to the dynamic graph G t =(V t , E), and generate a prediction graph and future use prediction results according to the path propagation of the improved graph neural network.

[0092] Specifically, this step is in a crucial position of connecting the preceding with the following in the system of the present invention. The input is the dynamic graph G t =(V t , E) with approval behavior information embedded in Step 2, and the output is the planned use predicted for each plot node v i at the future moment t + 1. Different from traditional land use planning that relies on static layers, manual overlay maps, and subjective judgments, this step not only attempts to give a certain use prediction, but also constructs a use evolution model with causal expression ability and spatial propagation through a modeling mechanism that combines "spatial structure + approval behavior + use evolution law", enabling the entire management system to predict the use change trend in the region in advance based on real behavior trends, plot spatial connections, and policy risk accumulation, so as to achieve intelligent intervention and proactive planning adjustment.

[0093] Specifically, in G t output in Step 2, each plot v i already has the following information:

[0094] The approval use status at the current moment (i.e., the most recent actual use);

[0095] Behavior Activity Index (indicating the approval density in recent years);

[0096] Use Violation Penalty Value (indicating the accumulation of deviation of historical approval uses from the plan);

[0097] The set of adjacent nodes and the edge weight w ij (from Step 1, expressing the spatial or planning level connection between plots).

[0098] The objective of the present invention is to design a prediction model f θ, while retaining the adjacent propagation characteristics of the graph structure, consider the history of approval behavior and the law of use evolution, so as to calculate the possible use states of each plot in the next stage The reason why traditional graph neural networks (such as GCN) or time series methods (such as RNN) cannot be directly applied to such tasks is that:

[0099] The land use status is a non - continuous variable, and the evolution path has asymmetry (for example, "ecological → commercial" is much more difficult than "commercial → residential");

[0100] The change of plot use is affected by the coupling of multiple factors including behavior drive + neighborhood co - evolution + planning constraints;

[0101] Especially in plots where the "planned use" and "approved use" have deviated for a long time, there is a trend of unstable use and it cannot be predicted by fitting the current situation.

[0102] Therefore, the present invention proposes a hybrid evolution model based on graph - structure drive, use co - evolution propagation and behavior - guided regularization. The present invention splits the evolution probability of plot use into a "self - drive part" and an "adjacent propagation part", and constructs the following prediction mechanism:

[0103]

[0104] The final prediction is:

[0105]

[0106] The description is as follows:

[0107] Represents the use activity drive of this node, that is, the more frequent the approval, the more likely the use change will occur;

[0108] The more serious the historical deviation, the more the continuation of the original use should be avoided, so as to promote the transformation of the predicted use;

[0109] The proportion of adjacent nodes that are consistent with the predicted use, indicating the degree of influence of the surrounding use on the plot. For example, if the surrounding of a plot gradually turns into commercial use, it is more likely to be evolved by the market or approval guidance itself;

[0110] An asymmetric use inertia penalty term, expressing that "certain uses should not be easily predicted to continue":

[0111] ρ is the hyper - parameter of the use deviation penalty intensity;

[0112] Indicates whether it is the original use;

[0113] Penalty(u) represents the continuation cost of this use. For example, set Penalty(Industrial) = 1.2, Penalty(Residential) = 0.9, Penalty(Ecological) = 0.6 to suppress the possibility of certain uses evolving continuously.

[0114] For example: plot v i The original use was industrial (u i = Industrial), and among the current adjacent nodes, 2 are commercial uses and 1 is industrial use, and the approval activity deviation Then:

[0115] If it is predicted that u = Industrial: the behavior score is 1.6·η1 - 0.7·η2, and the neighbor score is 1·w ij , and the penalty term is 1.2·ρ;

[0116] If it is predicted that u = Commercial: the behavior score is the same as above, and the neighbor score is 2·w ij , and the penalty term is 0;

[0117] By comprehensively comparing the scores of the two, the use prediction can be given.

[0118] This structure not only truly restores the evolution path between "policy - behavior - neighborhood - use", but also realizes the automatic suppression of unreasonable evolution trends (such as "ecological land being approved as industrial land year after year") through the special penalty term Penalty(u), without relying on external strong constraint rules, and maintains flexibility and automatic learning ability.

[0119] More importantly, the prediction result is not just a label, but can be further used for:

[0120] Comparing with the original planned use u i to find plots that may deviate from the use in the future;

[0121] Forming a use conflict index with adjacent plots to identify potential conflict areas;

[0122] Combined with the analysis of historical approval trends, judging which areas have the risk of planning failure.

[0123] Step 4: According to the predicted map and the future use prediction results, identify the possible use conflict trends in the national territorial space in the next time period, quantify the change intensity of the planning tension between regions, and locate the set of potential high-risk plots.

[0124] Specifically, this step aims to identify the possible use conflict trends in the national territorial space in the next time period based on the future use prediction results output in step three and quantify the change intensity of the planning tension between regions, so as to locate the set of potential high-risk plots The task of this step does not involve model modeling or output of management suggestions. Instead, as the "forward-looking identification layer", it realizes the transition from the "predicted use status" to the "use conflict evolution". Its positioning is similar to the "conflict radar on the temporal difference graph" and is the basis for the entire invention system to achieve dynamic monitoring and early warning.

[0125] The input includes the prediction map output by step three where each node v i contains:

[0126] the original planned use u i (static);

[0127] the current predicted use

[0128] the structure edge set E and edge weight w ij (inherited from step one);

[0129] The first step is to determine whether there may be a use deviation for each plot at time t + 1, that is The present invention defines such a deviation as the "potential use change tension", and further constructs a future tension evaluation function in combination with the use structure between adjacent plots

[0130]

[0131] where:

[0132] the predicted tension score, representing the plot v i and the adjacent plot v j in the future use status of the spatial tension;

[0133] w ij : the edge weight, representing the degree of spatial contact and hierarchical difference between plots (defined by step one);

[0134] the use conflict scoring matrix (such as industrial vs ecological set to 1.0, commercial vs residential set to 0.2); the indicator function, judging whether there is a deviation from the planned use;

[0135] λ: controls the influence degree of the deviation term on the overall tension improvement, and typical values are such as 0.5 - 1.0.

[0136] Based on the original "adjacent use conflict", this formula innovatively introduces a penalty term for deviation from the original plan, so that it can not only discover the direct conflicts between future uses, but also discover the regional evolution trends where the "structure seems coordinated but seriously deviates from the planning goals", which is the key technical highlight of this step.

[0137] The second step is to aggregate the tensions of each node and identify the tension accumulation points. In the present invention, for each node v i Calculate its total predicted tension exposure value S i :

[0138]

[0139] Where:

[0140] Plot v i The set of adjacent plots;

[0141] S i : The total predicted tension value of plot v i The larger the value, the closer it is to the boundary of the area with "high use variation / conflict pressure".

[0142] According to the empirically set tension threshold θ (for example, the median of all S i plus the standard deviation), the present invention marks the nodes with S i > θ as high-risk plots, forming a set of risk areas Each plot in the set is accompanied by its predicted use, planned use, and tension score for the next step of generating structured suggestions.

[0143] Step 5: According to the set of high-risk plots, combine the future use prediction results, use violation penalty values, approval status vectors, planned tension change intensities, and original planned map data to perform structured rule matching and generate a set of structured management suggestions.

[0144] Specifically, this step aims to generate a set of structured management suggestions based on the set of future high-tension areas identified in step four Combined with the use evolution trends of each plot, historical approval deviation situations, adjacent structure tension information, etc. For the territorial space planning system to be used as a direct reference or implementation basis in future layer revision and control strategy optimization.

[0145] The inputs for this step include:

[0146] The set of risk plots output from step four

[0147] Each plot The planned use u i ;

[0148] Predicted use (from step three);

[0149] Cumulative value of use deviation Approval activity (from Step 2);

[0150] Tension score S i (from Step 4);

[0151] Spatial attribute: area a i and planning level l i and edge weight w ij (from Step 1);

[0152] Conflict adjacency structure and its corresponding usage status.

[0153] The system first performs a structural rule match on each plot v i to determine its proposed category. The core rules are as follows:

[0154] Proposed land use revision: If the following conditions are simultaneously met:

[0155] (The predicted land use is inconsistent with the planned land use);

[0156] Higher than the set threshold (e.g., );

[0157] Tension value S i is medium or above (S i > μ S );

[0158] Then output: "It is recommended to revise the planned land use of v i from u i to Reason: The planned land use has been invalid for a long time, and the predicted trend is stable."

[0159] Proposed land use approval restriction: If the current predicted land use of plot v i is consistent with the planned land use, but there is a trend of land use change in ≥ 2 nodes among the adjacent plots (e.g., ), and the tension is high (S i > μ S ), then output:

[0160] "This plot is in the transitional zone of the structural land use edge. It is recommended to conduct a joint review and record control on the approval application, and give priority to guiding it to be consistent with the upper-level Figure 1 ."

[0161] Proposed conflict land use buffer zone: If v i is adjacent to at least two different types of plots (e.g., u j are residential and ecological respectively), and the conflict tension score Then output:

[0162] "It is recommended to delimit a buffer or transition area for specific uses (such as green spaces and public facilities) to resolve high-conflict boundaries and avoid future layer breaks."

[0163] The recommendations output by the system are represented in structured fields, including:

[0164] Plot number v i ;

[0165] Predicted use Compared with the original planned use u i ;

[0166] Type of recommendation (enumeration: revision, restriction, buffer);

[0167] Reason for recommendation (automatically generated according to rules);

[0168] Level of urgency for intervention (high, medium, low, graded according to the combined score of S i , );

[0169] Recommended operation method (such as amending the regulatory plan layer, approval restriction, use annotation, etc.);

[0170] Visualization of recommended layer markings (used to directly present the recommended area on the GIS platform).

[0171] An example output is as follows:

[0172]

[0173] Final set of recommendations As the system output interface, it can be docked with the layer management and approval control modules or the revision proposal system in the national territorial space planning platform to achieve an integrated closed-loop from data intelligent analysis to planning management decision-making.

[0174] In another embodiment of the present invention, a national territorial space planning management system based on big data is provided. The system includes:

[0175] A planning map segmentation unit for obtaining original planning map data, dividing the planning area of the original planning map data layer into non-overlapping basic plot units, analyzing the basic plot units according to the spatial contact degree and hierarchical relationship between the plot units, and obtaining a standardized spatial planning map structure G=(V, E); where, in the spatial planning map structure G=(V, E), V is the node set, each node represents a plot unit, E is the edge set, and the edge represents the connection relationship of the spatial contact degree and hierarchical relationship between the plot units, and expresses its conflict intensity or coupling relationship through weights;

[0176] A dynamic node analysis unit for constructing an approval behavior event sequence, where each behavior a of the event sequencek is defined as a triple (v i , t k , u k ), which represents that the land parcel unit v i is approved for the use u at time t k . According to the triple (v k , t i , u k ), an approval status vector is constructed for each land parcel unit, and a use violation penalty value is designed based on the intensity of the cumulative use deviation of the land parcel unit. Finally, the approval status vector and the use violation penalty value are added to each node of the spatial planning graph structure G = (V, E) to form a dynamic graph G k = (V t , E), where V t represents the set of dynamic land parcels at time point t; t

[0177] A graph neural network prediction unit for making predictions based on the improved graph neural network according to the dynamic graph G t = (V t , E), and generating a prediction graph and a future use prediction result according to the path propagation of the improved graph neural network;

[0178] A conflict quantification unit for identifying the possible trends of use conflicts in the national territorial space in the next time period according to the prediction graph and the future use prediction result, quantifying the intensity of the change in the planning tension between regions, and locating the set of potential high-risk land parcels;

[0179] A planning management output unit for performing structural rule matching according to the set of high-risk land parcels, combining the future use prediction result, the use violation penalty value, the approval status vector, the intensity of the change in the planning tension, and the original planning map data, and generating a structured set of management suggestions.

[0180] These are only some preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.​

Claims

1. A method for managing territorial spatial planning based on big data, characterized in that, The method includes the following steps: Step 1: Obtain the original planning map data, divide the planning areas in the original planning map data layer into non-overlapping basic plot units, analyze the basic plot units according to the degree of spatial contact and the proximity of the hierarchical relationships between the plot units, and obtain a standardized spatial planning map structure G = (V, E); where, in the spatial planning map structure G = (V, E), V is the set of nodes, each node represents a plot unit, E is the set of edges, and the edges represent the connection relationships of the spatial contact degree and hierarchical relationships between the plot units, and their conflict intensity or coupling relationship is expressed by weights; Step 2: Construct an approval behavior event sequence, where each behavior a in the event sequence k Defined as a triple (v i ,t k ,u k ), representing the plot unit v i At time t k Approved for use k , according to the triple (v i ,t k ,u k ) constructs an approval status vector for each plot unit, and designs a use violation penalty value based on the intensity of the cumulative use deviation of the plot unit. Finally, the approval status vector and the use violation penalty value are added to each node of the spatial planning graph structure G = (V, E) to form a dynamic graph G t =V t ,E), where V t Represents a dynamic set of plots at time point t; Step 3: Based on the dynamic graph G t =(V t , E), perform prediction based on the improved graph neural network, and generate a prediction graph and a future usage prediction result according to the path propagation of the improved graph neural network; Step 4: Identify the possible use conflict trends in the national territorial space in the next time period according to the prediction map and the future use prediction results, quantify the change intensity of the planning tension between regions, and locate the set of potential high-risk plots; Step 5: According to the set of high-risk plots, combine the future use prediction results, use violation penalty values, approval status vectors, change intensity of planning tension, and the original planning map data, perform structural rule matching, and generate a structured set of management suggestions.

2. The method for managing territorial spatial planning based on big data according to claim 1, characterized in that The following is also included in Step 1: Preprocess the original planning map data; the preprocessing includes coordinate system unification, use classification standard mapping, and attribute field normalization; In the graph structure G = (V, E), each plot unit will be used as a node in the graph structure, including: the basic information of the plot and its affiliated planning level; use the spatial contact degree and hierarchical relationship between the plot units as spatial / logical connection relationships to construct edges. Among them, if two plot units are completely not in spatial contact, even if they are in the same layer, no connection relationship will be established; if they overlap in space and come from different hierarchical layers, then construct an edge and calculate the weight of the edge, and this weight integrates the coincidence degree and hierarchical difference between the two plot units.

3. A method for managing territorial spatial planning based on big data according to claim 1, characterized in that, The approval status vector is constructed for each plot unit according to the triple (v i , t k , u k ), representing the dynamic behavior state of the node at time t, and is calculated as: Among them, is the dynamic state value of plot v i at time t; is the time decay factor, γ ∈ (0, 1), reflecting that the longer the historical approval time, the smaller the impact on the current state; c k is the credibility weight of approval behavior a k ; δ(u k ≠ u i ) is the use deviation indicator function. If the behavior use u k is inconsistent with the original planned use u of the plot i , it is 1, otherwise it is 0; K is the length of the historical approval window.

4. A method for managing territorial spatial planning based on big data according to claim 1, characterized in that, The dynamic graph G t =(V t , E), where V t represents the set of dynamic parcels at time point t, and each node state has the memory of the approval history, the ability to identify deviations, and the characterization of behavior density.

5. A method for managing territorial spatial planning based on big data according to claim 1, characterized in that, The prediction based on the improved graph neural network is performed according to the dynamic graph G t =(V t , E), and the prediction result is generated according to the path propagation of the improved graph neural network. The specific calculation is as follows: Among them, s i,u represents the predicted path, is the behavior driving factor, η1 and η2 are the weight coefficients of the node and use violation penalty values, represents the use activity drive of this node That is, the more frequent the approval, the more likely the use change will occur. It means that the more serious the historical deviation is, the more the continuation of the original use should be avoided, thus promoting the transformation of the predicted use. is the use violation penalty value; is the use co-evolution propagation factor, which represents the proportion of adjacent nodes that are consistent with the predicted use, indicating the degree of influence of the plot by the surrounding uses. j is node j, is the set of adjacent nodes, w ij is the edge weight, is the approval use status of node j; is the use inertia penalty factor, ρ is the use deviation penalty intensity hyperparameter; I(u = u i ) indicates whether it is the original use; Penalty(u) represents the continuation cost of this use.

6. A method for managing territorial spatial planning based on big data according to claim 1, characterized in that, The identifying the possible use conflict trends in the national territorial space in the next time period according to the prediction map and the future use prediction results, quantifying the change intensity of the planning tension between regions, and locating the set of potential high-risk plots specifically includes: Perform time-series state update calculation with whether there is a deviation from the planned use as a constraint item to generate a predicted tension score, which represents the spatial tension between the plot unit and its adjacent plot units in the future use state; Aggregate the tensions of each node, identify the tension concentration points, and generate the total predicted tension value of the current plot unit. The larger the value, the more it represents the boundary of the area under high use variation / conflict pressure; According to the tension threshold set by experience, mark the nodes with a total predicted tension value greater than the tension threshold as high-risk plots to form a set of risk areas.

7. A method for managing territorial spatial planning based on big data according to claim 6, characterized in that Whether there is a deviation from the planned use is determined by an indicator function.

8. A method for managing territorial spatial planning based on big data according to claim 1, characterized in that The structural rule matching specifically includes: Use revision suggestions: If the following conditions are all met simultaneously: The predicted use is inconsistent with the planned use; The cumulative value of use deviation is higher than the preset threshold; The total predicted tension value is medium or above; Then output: It is recommended to revise the planned use of the current plot unit from the original use to the future use prediction result, reason: The planned use has been invalid for a long time and the predicted trend is stable; Use approval restriction suggestions: If the current predicted use of the current land parcel unit is consistent with the planned use, but there is a trend of use change in ≥ 2 nodes in the adjacent land parcels and the tension is high, then output: This land parcel is in the marginal transition zone of structural use. It is recommended to conduct joint review and record control of the approval application, and give priority to guiding it to remain consistent with the upper-level map; Suggestions for the buffer zone of conflicting uses: If the current land parcel unit is adjacent to at least two land parcels of different types and the conflict tension score > 0.7, then output: It is recommended to delimit a use buffer or transition area to resolve the high-conflict boundary and avoid future layer breaks.

9. A method for managing territorial spatial planning based on big data according to claim 8, characterized in that, The conflict tension score is for similarity calculation.

10. A land spatial planning management system based on big data, characterized in that, The system includes: A planning map segmentation unit, which is used to obtain the original planning map data, divide the planning area of the original planning map data layer into non-overlapping basic land parcel units, analyze the basic land parcel units according to the spatial contact degree and the proximity of the hierarchical relationship between the land parcel units, and obtain the standardized spatial planning map structure G=(V,E); where, in the spatial planning map structure G=(V,E), V is the node set, each node represents a land parcel unit, E is the edge set, the edge represents the connection relationship of the spatial contact degree and the hierarchical relationship between the land parcel units, and its conflict intensity or coupling relationship is expressed by the weight; Dynamic node analysis unit, used to construct approval behavior event sequence, each behavior a of the event sequence k Defined as a triple (v i ,t k ,u k ), representing the plot unit v i At time t k Approved for use k , according to the triple (v i ,t k ,u k ) constructs an approval status vector for each plot unit, and designs a use violation penalty value based on the intensity of the cumulative use deviation of the plot unit. Finally, the approval status vector and the use violation penalty value are added to each node of the spatial planning graph structure G = (V, E) to form a dynamic graph G t =(V t ,E), where V t Represents a dynamic set of plots at time point t; Graph neural network prediction unit, for making predictions based on the improved graph neural network according to the dynamic graph G t =(V t , E), generating a prediction graph and a future usage prediction result according to the path propagation of the improved graph neural network; A conflict quantification unit, which is used to identify the possible use conflict trends in the territorial space in the next time period according to the prediction map and the future use prediction results, quantify the change intensity of the planning tension between regions, and locate the set of potential high-risk land parcels; A planning management output unit, which is used to perform structural rule matching according to the set of high-risk land parcels, combined with the future use prediction results, the use violation penalty value, the approval status vector, the change intensity of the planning tension, and the original planning map data, and generate a structured set of management suggestions.

Images