CN-121998456-A - Space suitability optimization method and system based on three-dimensional space re-identification

CN121998456ACN 121998456 ACN121998456 ACN 121998456ACN-121998456-A

Abstract

The invention relates to the technical field of homeland space planning, in particular to a space suitability optimization method and a space suitability optimization system based on three-dimensional space re-identification, wherein the method takes three-dimensional space as a space carrier, constructs a three-dimensional space re-identification matrix through a space-time semantic graph network and extracts potential producing area space units; the method comprises the steps of constructing a three-dimensional suitability tensor by taking space value evaluation as a research kernel, respectively representing ecological space bottom line constraint values, production space value matching characteristics and living space requirement adaptation characteristics by each channel of the three-dimensional suitability tensor, constructing a producing place space evolution model of multi-agent reinforcement learning on the basis, forming a dynamic punishment feedback mechanism by combining real-time monitoring data, outputting a corresponding space suitability layout result and realizing dynamic closed-loop updating. The invention realizes dynamic and automatic upgrading on the basis of the evaluation result of the existing static index system, and can provide quantitative decision support for planning, evaluation and dynamic adjustment of the homeland space.

Inventors

HE MEIZHEN
LI FENG
REN RUI
YANG YUYING
WANG BAIKUN
TANG XIN
ZHANG WEI
ZHANG SHUNCHENG

Assignees

四川省核地质调查研究所
四川省国土空间规划研究院

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (10)

1. The space suitability optimization method based on three-dimensional space re-identification is characterized by comprising the following steps of: acquiring multisource space-time data of a target area, constructing a space-time semantic graph network, aggregating spatial features of grid nodes through a graph convolution neural network, and outputting a three-dimensional space re-identification matrix containing fuzzy membership of the grid nodes belonging to production, living and ecological spaces; based on the three-generation space re-identification matrix, calculating the carbon sink opportunity cost of the target area and applying an ecological rigidity constraint mask to extract a potential production place space unit set of the target area; The method comprises the steps of merging space value evaluation data, industry evolution demand data and natural background data of a potential producing area space unit set, constructing a three-dimensional suitability tensor taking space unit coordinates as planes and taking ecological constraint, value matching and demand adaptation as characteristic channels, and normalizing characteristic values of all channels of the three-dimensional suitability tensor; And (3) inputting the three-dimensional suitability tensor as an environment state into the multi-agent game model, converting the action value output by the multi-agent game model into state transition probability, driving the cellular automaton to carry out spatial evolution, and outputting a corresponding spatial suitability optimization layout.
2. The space suitability optimization method based on three-dimensional space re-identification according to claim 1, wherein the three-dimensional space re-identification matrix containing the fuzzy membership degree of each grid node belonging to the production, living and ecological space is output, which specifically comprises: Defining a spatiotemporal semantic graph network as , wherein, As a set of grid nodes, For the collection of edges, An edge weight matrix based on the intensity of the material flow and the information flow; Node characteristics are aggregated through a graph convolution neural network, and any node is output Fuzzy membership vector belonging to production, living and ecological space The calculation formula is as follows: Wherein, the Is the neighbor node set of node i, D is the degree matrix, As an input feature vector for node j, As a matrix of weight parameters, Is an activation function; The extraction satisfies As a three-dimensional spatially blurred transition zone, wherein, Is a preset confidence threshold.
3. The space suitability optimization method based on three-dimensional space re-identification according to claim 2, wherein the carbon sink opportunity cost of the target area is calculated and an ecological rigidity constraint mask is applied, which is specifically: defining the carbon sink opportunity cost for any cell k in a set of potential producing space cells The calculation formula is as follows: Wherein, the The carbon sink per unit area maintains a natural background state for cell k, For the estimated carbon sink after the unit k is converted into the production space, For the carbon trade market price per unit equivalent, A sensitivity amplification factor for ecology; constructing an ecological rigidity constraint mask matrix M, if the unit k is located in the area Greater than a preset cost threshold or within legal ecological red line, mask matrix elements Otherwise And extracting a set of potential producing space units by carrying out Hadamard product operation on the space grid and the mask matrix M.
4. A method of optimizing spatial suitability based on three-dimensional spatial re-identification as claimed in claim 3, characterized by constructing a three-dimensional suitability tensor with spatial unit coordinates as planes and with ecological constraints, value matching and demand adaptation as feature channels, which is specifically: Building tensors , wherein, Longitude and latitude grid coordinates of the space unit; First channel of tensor Filling as units The ecological constraint index of (2) is obtained by carrying out weighted summation normalization on water loss sensitivity and biological diversity maintenance indexes; Second channel of tensor Filling as units The value matching index of (2) is obtained by mapping the economic value deduction carbon sink opportunity cost of basic production; Third channel of tensor Filling as units The demand adaptation index of (2) is obtained by fusing the industrial scale connection potential index and the resource footprint matching degree.
5. The method for spatial suitability optimization based on three-dimensional space re-recognition of claim 4, wherein a third channel of tensors Demand adaptation index in (a) The calculation formula is as follows through the calculation of the compound industry adaptation evaluation model: Wherein, the And Respectively units The physical and chemical suitability of soil and the bearing capacity of agricultural water resources; And Respectively obtaining an optical temperature radiation resource index and a grid access node capacity margin; And Respectively an agricultural demand weight and a new energy access weight; is a compound industrial space compatibility regulating factor, and 。
6. The space suitability optimization method based on three-dimensional space re-identification according to claim 5, wherein the three-dimensional suitability tensor is input as an environmental state into a multi-agent game model, which is specifically: Instantiating a producing area developer and an ecological protection party as agents which are mutually game, wherein the state space of each agent is a three-dimensional suitability tensor T; Defining local joint rewarding function of developing agent at any producing place at time t The calculation formula is as follows: Wherein, the A set of potential production space units occupied by the agent at time t; 、、 Respectively units Corresponding three channel characteristic values in the tensor; For the forward excitation weight to be the same, For the ecological constraint out-of-limit penalty factor, A physiological sensitivity parameter.
7. The space suitability optimization method based on three-dimensional space re-recognition as set forth in claim 6, wherein the motion value output by the multi-agent game model is converted into a state transition probability, and the state transition probability of the cellular automaton is the state transition probability The calculation formula of (2) is as follows: Wherein, the The action value Q function value for the agent output to convert element i to origin type j, Neighbor node set representing element i The cell proportion for which the internal state is already j, To indicate the state of the function, when the cell k Time of day Otherwise , As a global space suitability regulating function, when the unit i corresponds to the mask matrix element In the time-course of which the first and second contact surfaces, Otherwise, 1.
8. The space suitability optimization method based on three-dimensional space re-identification according to claim 7, wherein the corresponding space suitability optimization layout is output, which specifically comprises: extracting feature vectors of potential origin space units evolving to stable states in a three-dimensional suitability tensor ; Setting spatially adapted positive ideal solution vectors Calculating a weighted Euclidean distance , wherein, Weighting coefficients for each dimension; Based on weighted Euclidean distance And (3) ascending arrangement is carried out on the space units, and a first suitability scale development area, a second suitability feature compatible area and a third suitability limiting backoff area are sequentially defined according to the total amount control index distribution proportion of the target area.
9. The method for optimizing space suitability based on three-dimensional space re-identification of claim 8, further comprising the steps of acquiring real-time ecological monitoring data of the Internet of things, generating a dynamic penalty factor through a nonlinear attenuation model, feeding back the dynamic penalty factor to a value matching characteristic channel of a three-dimensional suitability tensor to update a characteristic value so as to trigger the re-modeling of space layout and complete dynamic closed loop control of space suitability; The method comprises the steps of generating a dynamic penalty factor through a nonlinear attenuation model, and feeding back the dynamic penalty factor to a value matching characteristic channel of a three-dimensional suitability tensor to update a characteristic value, wherein the method specifically comprises the following steps: Calculation unit Ecological stress index at time t ; If unit The ecological stress index of (2) is continuous Triggering and calculating dynamic penalty factors of the unit if the monitoring period exceeds a preset ecological tolerance threshold The calculation formula is as follows: Wherein, the Is a unit The dynamic penalty factor at the current instant t, In order to penalize the decay rate constant, For the summation index variable of the history monitoring time, For a set number of consecutive monitoring cycles, Is a unit At the time of history monitoring Is characterized by an ecological stress index of (a), Is an ecological tolerance threshold; applying a dynamic penalty factor to the second channel of the three-dimensional suitability tensor to update the local tensor value to be To suppress the game expansion probability of the unit in the spatial evolution model.
10. Space suitability optimizing system based on three-dimensional space re-identification, which is characterized by comprising: the map network space recognition module is used for acquiring multi-source space-time data of a target area and constructing a space-time semantic map network, and outputting a three-dimensional space re-recognition matrix containing fuzzy membership of each grid node belonging to production, living and ecological space through a map convolution neural network; The mask constraint and collection extraction module is used for calculating the opportunity cost of the carbon sink, applying an ecological rigidity constraint mask and extracting a potential space unit collection of the production place; The three-dimensional suitability tensor construction module is used for fusing multidimensional data, constructing a three-dimensional suitability tensor taking ecological constraint, value matching and requirement adaptation as characteristic channels, and carrying out characteristic value normalization; The game evolution optimizing module is used for inputting a three-dimensional suitability tensor as an environment state into the multi-agent game model, converting the action value into a state transition probability to drive cellular automaton space evolution so as to output a suitability optimizing layout; and the digital twin closed loop feedback module is used for generating a dynamic penalty factor based on the real-time ecological monitoring data and feeding back and updating the value matching characteristic channel of the three-dimensional suitability tensor to trigger the dynamic correction of the space layout.

Description

Space suitability optimization method and system based on three-dimensional space re-identification Technical Field The invention relates to the technical field of homeland space planning, in particular to a space suitability optimization method and system based on three-dimensional space re-identification. Background With the rapid development of the economic society and the deep advancement of ecological civilization construction, the regional space is used as a core carrier for bearing production, living and ecological activities, and the scientific configuration of the spatial pattern becomes a key link for realizing the regional sustainable development. In a macroscopic planning system of a three-dimensional space (production, living and ecological space), the space of a production place is a core element for guaranteeing grain safety and promoting industrial aggregation and supporting entity economy. With the rising of modern agriculture, new energy industry and compound industry, the connotation and boundary of the space of the production place are increasingly rich, and the traditional single element evaluation has been developed to multi-dimensional element fusion. In recent years, with the wide application of the Internet of things, big data and artificial intelligence technology, space planning is gradually transformed from static macro regulation to dynamic fine treatment. By introducing leading edge algorithms such as a graphic neural network, a multi-agent model, a cellular automaton and the like, deep semantic association behind multi-source space-time data is excavated, the coupling process of multiple elements in space evolution is simulated, and intelligent calculation of ecological bearing capacity is combined, so that the method has become an important development direction of space adaptability optimization and digital twin space management of the current production place. However, in the process of compiling and implementing the homeland space planning, the space suitability of the production place is influenced by factors such as multi-main game, element flow, ecological risk disturbance, real-time change of monitoring data and the like, and the single static evaluation result has certain applicable limitation in the dynamic management scene of the planning implementation period. Therefore, on the basis of the existing static index system and evaluation results, a dynamic automatic optimization method system facing to the implementation period is further formed, namely, a static evaluation-result output is expanded to a technical path of static evaluation bottoming, dynamic evolution optimization, monitoring feedback closed loop and continuous updating, so that the practicability and dynamic treatment capacity of the homeland space planning are improved. Disclosure of Invention The invention aims to provide a space suitability optimization method and a space suitability optimization system based on three-dimensional space re-identification, which not only realize deep fusion of multi-dimensional space data, but also finish technical crossing of producing space from static planning to dynamic self-adaptive game optimization by relying on a carbon sink constraint and dynamic punishment closed-loop mechanism by constructing a three-dimensional suitability tensor and multi-agent reinforcement learning evolution model. The invention is realized by the following technical scheme: the space suitability optimization method based on three-dimensional space re-identification comprises the following steps: acquiring multisource space-time data of a target area, constructing a space-time semantic graph network, aggregating spatial features of grid nodes through a graph convolution neural network, and outputting a three-dimensional space re-identification matrix containing fuzzy membership of the grid nodes belonging to production, living and ecological spaces; based on the three-generation space re-identification matrix, calculating the carbon sink opportunity cost of the target area and applying an ecological rigidity constraint mask to extract a potential production place space unit set of the target area; The method comprises the steps of merging space value evaluation data, industry evolution demand data and natural background data of a potential producing area space unit set, constructing a three-dimensional suitability tensor taking space unit coordinates as planes and taking ecological constraint, value matching and demand adaptation as characteristic channels, and normalizing characteristic values of all channels of the three-dimensional suitability tensor; And (3) inputting the three-dimensional suitability tensor as an environment state into the multi-agent game model, converting the action value output by the multi-agent game model into state transition probability, driving the cellular automaton to carry out spatial evolution, and outputting a corresponding spatial suitability optimiz