Ecological carrying capacity evolution prediction and cross-region scheduling method based on spatial correlation

Abstract

The present application belongs to the technical field of natural resource management and control scheduling, and provides an ecological carrying capacity evolution prediction and cross-regional scheduling method based on spatial correlation, which comprises the following steps: collecting spatial correlation data streams of each spatial node in a target region, screening a plurality of adjacent spatial node groups through a dynamic time warping algorithm, analyzing the connectivity of each group of data streams and constructing an intra-group connectivity coupling model, taking the shared spatial nodes of each node group as a cascading conduction relay to obtain a global connectivity coupling model, inputting the target node data stream into the model when a cross-regional scheduling event is triggered, deducing the next control period data stream of each spatial node, calculating the ecological pressure index, comparing the ecological pressure index with the carrying capacity red line threshold, completing the ecological carrying capacity evolution prediction, and if there is a node index exceeding the limit, regulating the global ecological parameters to maintain the pressure index of all nodes within the threshold range, completing the cross-regional ecological scheduling, and improving the accuracy of regional ecological management and control and resource scheduling.

Description

Technical Field

[0001] This invention belongs to the field of natural resource management and scheduling technology, specifically a method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation. Background Technology

[0002] Current ecological early warning schemes mostly rely on time-series monitoring data from various independent monitoring nodes, using time-series models such as ARIMA and LSTM to conduct single-point trend predictions; after monitoring indicators exceed limits, pollution control and resource allocation plans are formulated based on human experience. However, natural ecosystems are characterized by spatial connectivity and cascading propagation of disturbances, and existing technologies still have three core shortcomings:

[0003] The traditional approach neglects the cross-regional causal relationships of ecological factors. Pollutant migration, eutrophication diffusion, and carbon emission transfer exhibit spatial lag characteristics. Treating each spatial node as an independent unit, the traditional approach lacks causal tests to distinguish between linear and nonlinear spatial connectivity, making it difficult to identify the transmission pathways of ecological risks. Consequently, the carrying capacity predictions deviate significantly from the actual evolution patterns.

[0004] Traditional coupled models are limited to a single jurisdiction or a small watershed; disturbances caused by localized pollution control and water resource scheduling will propagate outward layer by layer along the geographical topology. Existing technologies lack a shared relay node networking architecture, cannot encapsulate the global coupled network, cannot deduce the chain reaction of scheduling disturbances across the entire region, and lack global analysis capabilities.

[0005] Existing equipment can only output alarms for exceeding limits, and cannot trace the source nodes of ecological overload by following the causal transmission network. It also cannot decouple the calculation of the amount of regulation and compensation required for the source area. Scheduling decisions rely on manual negotiation and cannot output quantitative instructions to drive sewage valves, water conservancy facilities, carbon capture equipment and other terminals to perform regulation. The scheduling accuracy and automation level are low.

[0006] In summary, there is an urgent need to address the technical challenges of accurately constructing cross-regional ecological causal transmission networks, predicting carrying capacity evolution, tracing the sources of overload, and implementing closed-loop intelligent scheduling across the entire region. To this end, this invention provides a method for predicting ecological carrying capacity evolution and cross-regional scheduling based on spatial correlation. Summary of the Invention

[0007] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0008] The technical solution adopted by this invention to solve its technical problem is: This invention provides a method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation, including: Acquire the spatial correlation data stream of each spatial node in the target area, and filter out multiple adjacent spatial node groups through a dynamic time warping algorithm based on causality test; Analyze the connectivity between spatially correlated data streams within any adjacent spatial node group, and establish a connectivity coupling model for adjacent spatial node groups based on the connectivity analysis results; Obtain the shared spatial nodes between each adjacent spatial node group, and use the shared spatial nodes as cascade transmission relays to encapsulate the connectivity coupling model of all adjacent spatial node groups into a global connectivity coupling model; When a candidate cross-regional scheduling event is generated, the spatial correlation data stream of the target spatial node corresponding to the candidate cross-regional scheduling event is obtained, and the spatial correlation data stream is input into the global connectivity coupling model. Through the cascade transmission relay of shared spatial nodes, the spatial correlation data stream of each global spatial node in the next control cycle is output. Based on the spatial correlation data stream of each spatial node in the next regulation cycle, calculate the ecological pressure index of each spatial node and compare it with the corresponding carrying capacity red line threshold to complete the prediction of ecological carrying capacity evolution. If the evolution prediction results show that the ecological pressure index of any spatial node exceeds the carrying capacity red line threshold, then the ecological pressure index of all spatial nodes will be adjusted to be within the carrying capacity red line threshold, thus completing the cross-regional scheduling.

[0009] Preferably, the specific process of obtaining the spatial correlation data stream of each spatial node to which the target area belongs is as follows: The target area is divided into multiple independent spatial nodes. By deploying physical monitoring equipment at each spatial node or calling historical ecological databases, multidimensional ecological indicator data of each spatial node within a preset time window are obtained. The multidimensional ecological indicator data includes air pollutant concentration, water eutrophication index and carbon emissions. The multidimensional ecological indicator data are arranged in time series to construct a spatial correlation data stream for each spatial node.

[0010] Preferably, the specific screening process for the adjacent spatial node group is as follows: The dynamic time warping algorithm is used to calculate the warping distance between the spatial correlation data streams of any two spatial nodes. If the warping distance is less than a preset warping distance threshold, the Granger causality test is used to calculate the causal significance P-value and causal propagation direction between the spatial correlation data streams of the two spatial nodes. If the causal significance P-value is less than a preset significance threshold, the two spatial nodes are determined to be adjacent spatial node groups.

[0011] Preferably, the specific process of analyzing the connectivity between spatially correlated data streams within any adjacent spatial node group is as follows: Based on the causal transmission direction determined by the causal significance P-value, the two spatial nodes within an adjacent spatial node group are marked as the source node and the sink node, respectively. Calculate the linear Pearson correlation coefficient and nonlinear mutual information value between the spatially correlated data streams corresponding to the source and sink nodes; If the absolute value of the linear Pearson correlation coefficient is greater than or equal to the preset linearity threshold, then it is determined that there is linear connectivity within the adjacent spatial node group. If the absolute value of the linear Pearson correlation coefficient is less than the linear determination threshold, and the nonlinear mutual information value is greater than or equal to the preset mutual information determination threshold, then it is determined that there is nonlinear connectivity within the adjacent spatial node group.

[0012] Preferably, the specific process of establishing the connectivity coupling model of the adjacent spatial node group is as follows: If the connectivity type is linear connectivity, then an autoregressive distributed lag model based on the linear transfer matrix is ​​established to fit the linear numerical transmission from the source node to the sink node. If the connectivity type is nonlinear connectivity, a nonlinear dynamic mapping function based on the multilayer perceptron network layer is established to fit the nonlinear response transmission of the source node to the sink node.

[0013] Preferably, the specific encapsulation process of the global connectivity coupling model is as follows: Obtain the spatial node identifiers of each adjacent spatial node group, perform set intersection comparison, and mark the spatial nodes that exist in at least two different adjacent spatial node groups as shared spatial nodes. A node state cascade transfer equation is established using shared space nodes as cascade transmission relays, and the output of the connectivity coupling model of the upstream adjacent space node group is mapped and connected to the input of the connectivity coupling model of the downstream adjacent space node group. Based on the node state cascade transfer equation, tensor splicing technology is used to dimensionally align and splice the linear transfer matrix or multilayer perceptron network layer of the connectivity coupling model of independent adjacent spatial node groups. Taking each spatial node as a graph node, the causal transmission direction as the direction of the directed edge, and the causal significance P-value or nonlinear mutual information value as the weight of the directed edge, the spliced ​​overall structure is integrated into a global directed graph transmission network and encapsulated as a whole to generate a global connectivity coupling model.

[0014] Preferably, the specific process of outputting the spatial correlation data stream of all global spatial nodes in the next control cycle is as follows: Obtain the control location information and control intensity parameters of candidate cross-regional scheduling events to determine the target spatial nodes; The control intensity parameter is set as the expected value change as the disturbance variable, and the corresponding time series node in the initial spatial correlation data stream is located according to the preset effective time. At time series nodes and subsequent nodes, numerical addition and subtraction operations are performed independently on the corresponding atmospheric pollutant concentrations, water eutrophication indices, or carbon emissions. The data sequence after superimposed perturbation variables is smoothed and normalized to generate an updated spatial correlation data stream; The updated spatial correlation data stream of the target spatial node, as well as the initial spatial correlation data stream of other non-target spatial nodes, are synchronously input into the global connectivity coupling model, and the spatial correlation data stream of each global spatial node in the next regulation cycle is output.

[0015] Preferably, the specific process for calculating the ecological pressure index of each spatial node is as follows: Based on the spatial correlation data stream of the next regulation cycle, obtain multi-dimensional ecological indicator data of spatial nodes under the next regulation cycle; The range standardization method is used to eliminate the dimensional differences in ecological indicator data of each dimension, and standardized values ​​are obtained. Obtain the feature weight coefficients corresponding to the ecological indicator data of each dimension, and calculate the weighted sum of the standardized values ​​of each dimension and their corresponding feature weight coefficients to obtain a dimensionless comprehensive scalar value, which serves as the ecological pressure index of the spatial node.

[0016] Preferably, the specific process for completing the ecological carrying capacity evolution prediction is as follows: The calculated ecological pressure index of each spatial node is compared with the carrying capacity red line threshold corresponding to each spatial node. If the ecological pressure index is less than or equal to the corresponding carrying capacity red line threshold, then the corresponding spatial node is marked as ecologically safe. If the ecological pressure index is greater than the corresponding carrying capacity red line threshold, the corresponding spatial node is marked as ecological overload, and the difference exceeding the carrying capacity red line threshold is calculated as the overload risk.

[0017] Preferably, the specific process for completing the cross-regional scheduling is as follows: Using spatial nodes marked as ecological overload as the starting point for tracing the source, a graph theory search algorithm is used to traverse the connected paths composed of directed edges in the reverse direction of the global directed graph transmission network. Based on the weights of each directed edge on the connected path, the comprehensive transmission contribution of each connected path is calculated. The connected path with the largest comprehensive transmission contribution is identified as the key transmission path, and the node at the forefront of the key transmission path is identified as the upstream source space node. The overload risk is decoupled into the over-limit of a single ecological indicator by using the feature weight coefficient, and the parameter compensation of the upstream source spatial node is calculated by combining the comprehensive transmission contribution. The regulation intensity parameters are updated based on the parameter compensation amount to generate new expected numerical changes in the direction of increasing the reduction or gain. When the ecological pressure index of all spatial nodes is less than or equal to their corresponding carrying capacity red line threshold, the corresponding cross-regional scheduling event parameters are obtained as a balance scheduling scheme, which is then converted into control execution instructions and sent to the corresponding physical execution terminals in the target area. The physical execution terminals include water conservancy gate control systems, industrial sewage valve controllers or intelligent power limiting gateways and carbon capture equipment controllers.

[0018] The beneficial effects of this invention are as follows:

[0019] This invention integrates causal testing and dynamic time warping algorithms to achieve precise screening of spatial node groups. It combines Pearson correlation coefficient and nonlinear mutual information value to distinguish between linear and nonlinear spatial connectivity relationships. This invention breaks through the technical bottleneck of traditional single correlation analysis, which cannot take into account temporal characteristics and causal transmission mechanisms and is difficult to characterize ecological associations in complex regions. It significantly improves the refinement and accuracy of spatial ecological association identification.

[0020] This invention constructs a linear autoregressive distributed lag model and a nonlinear multilayer perceptron mapping model to address the differences in connectivity types. By using shared space node cascade relay, tensor dimension splicing, and global directed graph network encapsulation, it builds a unified connectivity coupling model across the entire domain. This effectively solves the problems of fragmented regional nodes, incomplete characterization of linkage and transmission relationships, and low accuracy of multi-regional ecological coupling modeling in traditional technologies, and realizes accurate extrapolation of ecological data flow across nodes and regions. Attached Figure Description

[0021] The invention will now be further described with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart of the steps of the ecological carrying capacity evolution prediction and cross-regional scheduling method based on spatial correlation of the present invention;

[0023] Figure 2 This is a schematic diagram of the causal significance connectivity matrix provided in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the global directed graph transmission network topology provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0026] Example 1

[0027] like Figure 1 As shown in the embodiments of the present invention, the method for predicting the evolution of ecological carrying capacity based on spatial correlation and cross-regional scheduling includes: Step 1: Obtain the spatial correlation data stream of each spatial node in the target area, and filter out multiple adjacent spatial node groups through a dynamic time warping algorithm based on causality test; In some specific embodiments, the process of obtaining the spatial correlation data stream of each spatial node to which the target area belongs in step one is as follows: Based on geographic coordinates, the target area is divided into multiple independent spatial nodes; By deploying monitoring equipment at various spatial nodes (e.g., air quality monitoring micro-stations for collecting atmospheric pollutant concentrations, online automatic water quality monitoring buoys for collecting eutrophication indices, and physical sensing terminals such as carbon flux observation towers for collecting carbon emissions), or by accessing historical ecological databases, multidimensional ecological indicator data for each spatial node within a preset time window can be obtained.

[0028] The multidimensional ecological indicator data includes: air pollutant concentration, water eutrophication index, and carbon emissions. Multidimensional ecological indicator data refers to parameters that can characterize the state of ecological carrying capacity. It should be noted that the preset time window refers to the long-term historical observation span used to obtain historical evolution patterns. The preset time window length is usually set to a long-term span that includes at least one complete natural year (e.g., the past 12 months to 36 months) to ensure that the obtained multidimensional ecological indicator data can fully cover the complete ecological seasonal fluctuation cycle and capture the longest physical lag time of cross-regional transmission. The multidimensional ecological index data are arranged in chronological order according to time series, and then denoised and normalized to eliminate the dimensional differences between different indicators. The resulting multidimensional ecological flux time series with timestamps is defined as the spatial correlation data stream of each spatial node.

[0029] The spatial correlation data stream obtained through the above process not only includes the static ecological capacity of spatial nodes, but also records the dynamic trajectory of ecological indicator data fluctuations over time, providing basic data for subsequent capture of cross-regional transmission delays.

[0030] In some specific embodiments, the process of selecting multiple adjacent spatial node groups in step one using a dynamic time warping algorithm based on causality testing is as follows: The dynamic time warping algorithm is used to calculate the warped distance between spatially correlated data streams of any two spatial nodes.

[0031] Because the transmission of ecological impacts in space has a physical lag effect (for example, it takes several days for pollutants from spatial node A to drift to spatial node B by the wind), traditional synchronous correlation analysis is prone to missing results.

[0032] The dynamic time warping algorithm allows for non-linear distortion and local scaling of the time axis, enabling the alignment of data peaks with time phase shifts and thus accurately calculating the warping distance including hysteresis effects. For example, the specific process for calculating the normalized distance between spatially correlated data streams of any two spatial nodes is as follows: Obtain the first spatial correlation data stream of the first spatial node, denoted as time series. And the second spatial correlation data stream of the second spatial node, denoted as time series. , where n and m are the total number of sampling points contained in the two time series, respectively; Construct a local distance matrix D of size n×m, and calculate the distance for each data point in the time series Q. With each data point in time series C The Euclidean distance between them is calculated using the following formula: ; i is the index of the sampling point of the first time series Q, ;j is the index of the sampling point of the second time series C, ; Will The coordinates in the local distance matrix D are The element values ​​are used to characterize the instantaneous ecological state differences between two spatial nodes at different timestamps; Based on the local distance matrix D, a regular path from the starting point (1,1) to the ending point (n,m) is found using a dynamic programming algorithm. Where K is the total number of nodes contained in the regularized path, and the path nodes k represents the path step index. This represents the Q-side sampling index corresponding to the k-th step of the path. This represents the sampling index on the C side corresponding to the k-th step of the path.

[0033] In the process of finding regular paths, monotonicity constraints and continuity constraints are set, i.e., the boundary conditions are satisfied. and and step constraints This ensures that time evolution can only proceed in one direction and cannot be skipped; Simultaneously introduce global window constraints based on physical hysteresis limits (e.g., Sakoe-Chiba band constraints), with the following constraints: ; Where r is the maximum number of time steps required for the regularized path to deviate from the diagonal of the matrix. The value of the maximum number of time steps r depends on the maximum physical limit time for the ecological indicator data to propagate in space (e.g., the longest transmission time required for air pollutants to cross the distance between two spatial nodes), to prevent the algorithm from performing excessive time-distorted matching without physical meaning; Dynamic programming is used to traverse the matrix grid to find the optimal regularized path with the lowest cumulative cost; a cumulative cost function is defined. x is the horizontal index of the matrix corresponding to time series Q, and y is the vertical index of the matrix corresponding to time series C. Only when the aforementioned Sakoe-Chiba band constraint is satisfied Cost iteration calculation is performed within the matrix coordinate range; Set the initial conditions for dynamic programming: ; The state transition equation is: ; in, This represents the journey from the initial coordinates (1,1) of the matrix to the current coordinates. The minimum cumulative distance cost; coordinates The corresponding local Euclidean distance.

[0034] Minimize the cumulative cost of the endpoint corresponding to the optimal regularized path. Defined as the dynamic time-warped distance between the spatially correlated data streams of the first spatial node and the second spatial node, i.e. DTW is an abbreviation for Dynamic-Time Warping.

[0035] The calculated regularization distance is compared with a preset regularization distance threshold. The regularization distance threshold is used to characterize the similarity tolerance of two spatial nodes in the ecological evolution waveform.

[0036] If the regularization distance is less than the preset regularization distance threshold, it is initially determined that the two spatial nodes have potential connectivity in the ecological fluctuation trend, and the two spatial nodes are selected as candidate node pairs. For the initially identified candidate node pairs, a Granger causality test is performed on the spatial correlation data streams of the candidate node pairs. Specifically, by constructing an autoregressive model, it is tested whether the historical lag term of the spatial correlation data stream of one spatial node can substantially reduce the prediction error of the spatial correlation data stream of the other spatial node, thereby outputting a causal significance P-value that characterizes the direction and strength of causal transmission. It should be noted that the causal transmission direction points from the side that substantially reduces the prediction error to the side whose prediction error is reduced. The causal transmission direction clearly defines the transmission initiation node and transmission receiving node in the process of ecological impact transfer in physical space. The magnitude of the causal significance P-value is used to quantify the causal transmission intensity in reverse. That is, the smaller the causal significance P-value, the more severe the ecological evolution impact of the transmission initiation node on the transmission receiving node. For example, the specific process of outputting the causal significance P-value, which characterizes the direction and strength of causal transmission, is as follows: The spatial correlation data stream of the first spatial node in the candidate node pair is set as the target sequence. The spatial correlation data stream of the second spatial node is the prediction sequence. ,in This represents the total number of samples in the time series. Based on the maximum number of time steps set in the aforementioned dynamic time warping calculation process, the maximum lag order of the autoregressive model is determined. ; Constructing only the target sequence The constrained autoregressive model with its own historical lags is expressed as follows: ; in, For the current time step, This represents the predicted value of the target sequence at the current time step. For constant terms, These are the autoregressive coefficients. These are the regression residuals of the constrained model. For the target sequence... The current value is used to fit and predict the model, and the sum of squared residuals of the restricted autoregressive model is calculated, denoted as the first sum of squared residuals. ; Constructing a structure that simultaneously contains the target sequence Its own historical lag and the predicted sequence The expression for the unrestricted autoregressive model with historical lags is: ; in, For constant terms, and For regression coefficients, These are the regression residuals of an unrestricted model. For the target sequence... The current value is fitted for prediction, and the sum of squared residuals of the unrestricted autoregressive model is calculated, denoted as the second sum of squared residuals. ; Based on the first residual sum of squares With the second residual sum of squares The difference, combined with the total sample size of the time series. With the maximum hysteresis order Calculate The test statistic is calculated using the following formula: ; According to the above Test statistic in The cumulative probability corresponding to the distribution is used to calculate the test prediction sequence. Is it the target sequence? The first causal significance p-value of Granger causality; Swap target sequences With predicted sequence Given the given conditions, repeat the above steps to calculate the second causal significance P-value; For example, the specific process of outputting the causal significance P-value, which characterizes the direction and strength of causal transmission, is as follows: The spatial correlation data stream of the first spatial node in the candidate node pair is set as the target sequence. The spatial correlation data stream of the second spatial node is the prediction sequence. Where N is the total number of samples in the time series. This represents the observed value of the target sequence at time step t. This represents the observed value of the predicted sequence at time step t; Based on the maximum time step set in the aforementioned dynamic time warping calculation process, the maximum lag order p of the autoregressive model is determined; specifically, the maximum lag order p is set to be equal to the maximum time step r that the path is allowed to deviate from the diagonal in the Sakoe-Chiba constraint. The range of values ​​for the time step index t is as follows: Discard the first p observation times when it is impossible to obtain all the lagged samples; Constructing only the target sequence The constrained autoregressive model with its own historical lags is expressed as follows: ; in, For the regression constant term, The autoregressive coefficients corresponding to the lagged terms of the target sequence are... Let be the regression residual of the restricted autoregressive model at time t.

[0037] The target sequence is fitted with a restricted autoregressive model at the current time step, and the sum of squared residuals of the restricted autoregressive model is calculated, denoted as the first sum of squared residuals. , ; Constructing a structure that simultaneously contains the target sequence Its own historical lag and the predicted sequence The expression for the unrestricted autoregressive model with historical lags is: ; in, For the regression constant term, The regression coefficients corresponding to the lagged terms of the target sequence are... To predict the regression coefficients corresponding to the lag terms of the sequence, Let be the regression residual of the unrestricted autoregressive model at time t.

[0038] For the target sequence The current value is fitted for prediction, and the sum of squared residuals of the unrestricted autoregressive model is calculated, denoted as the second sum of squared residuals. , ; Based on the first residual sum of squares With the second residual sum of squares The difference, combined with the total number of valid samples The F-test statistic is calculated based on the maximum lag order p. Molecular degrees of freedom equal to the number of additional parameters to be estimated in the unconstrained model compared to the constrained model, p; The denominator degrees of freedom equal the number of effective observation samples minus the total number of parameters in the unconstrained model, which is... ; The calculation formula is: ; Based on the cumulative probability of the F-distribution corresponding to the F-test statistic, calculate the first causal significance P-value corresponding to the Granger cause of whether the predicted sequence X is the target sequence Y. Swap the target sequence and the predicted sequence, set X as the new target sequence and Y as the new predicted sequence, repeat all the above regression F test steps, and calculate the second causal significance P value used to test whether Y is a Granger cause of X.

[0039] The causal significance P-value is compared with a preset significance threshold. The significance threshold is typically set to 0.05 to filter out spurious correlations caused by chance factors.

[0040] If the causal significance P-value is less than the preset significance threshold, then a clear source-sink causal transmission relationship is confirmed between the two spatial nodes; Two spatial nodes that simultaneously satisfy the conditions of having a normalization distance less than a preset normalization distance threshold and a causal significance P-value less than a preset significance threshold are paired and bound together to generate multiple adjacent spatial node groups.

[0041] It should be noted that the adjacent spatial node group mentioned in this invention is not limited to physical adjacency in geographical location, but refers to topological adjacency in which ecological indicator data have high similarity and clear causal connectivity in the transmission network.

[0042] Step 2: Analyze the connectivity between spatially correlated data streams within any adjacent spatial node group, and establish a connectivity coupling model for adjacent spatial node groups based on the connectivity analysis results; In some specific embodiments, the process of analyzing the connectivity between spatially correlated data streams within any adjacent spatial node group in step two is as follows: Based on the causal transmission direction determined in step one based on the causal significance P-value, the two spatial nodes inside the candidate node pair are marked as the source node and the sink node, respectively. The starting node of the transmission is marked as the source node, and the receiving node of the transmission is marked as the sink node. The spatial correlation data streams corresponding to the source node and sink node are obtained, the linear Pearson correlation coefficient between them is calculated, and the absolute value of the linear Pearson correlation coefficient is compared with a preset linear judgment threshold. The linear judgment threshold is used to measure whether there is a significant proportional trend of change in the same or opposite direction between the data streams. In some embodiments, the linear judgment threshold can be 0.85. The nonlinear mutual information value between the spatially correlated data streams corresponding to the source node and sink node is calculated, and the nonlinear mutual information value is compared with a preset mutual information judgment threshold. The nonlinear mutual information value is used to characterize the amount of nonlinear information contained in one spatially correlated data stream, which can capture complex and non-proportional ecological response relationships; For example, the specific process for calculating the nonlinear mutual information value between the spatially correlated data streams corresponding to the source node and the sink node is as follows: Obtain the spatial correlation data stream corresponding to the source node, denoted as discrete time series. And the spatial correlation data stream corresponding to the sink node, denoted as discrete time series. ,in This represents the total number of samples after time-phase alignment. Using state-space partitioning methods (such as histogram estimation), the time series is divided into... and The numerical range is divided into multiple discrete state intervals; Statistical Time Series The frequency of data points falling into each state interval is used to calculate the time series. Marginal probability distribution Similarly, the time series is calculated. Marginal probability distribution ; Statistical Time Series and Data pairs at the same time position The frequency of both falling into the corresponding state interval is used to calculate the joint probability distribution of the source node and the sink node. ; Based on Shannon's information theory, the nonlinear mutual information value between the spatially correlated data streams corresponding to the source node and the sink node is calculated. The calculation formula is: ; in, This represents the nonlinear mutual information value, and its unit is bit. Representing time series Discrete states after partitioning Representing time series Discrete states after partitioning; The source node data is in a state The marginal probability, The data at the sink node is in a state. The marginal probability; The source node data is in a state And the sink node data is in state. The joint probability; This is a base-2 logarithmic operation. By calculating the divergence between the product of the joint probability distribution and the marginal probability distribution, the nonlinear coupling strength between the source and sink nodes, independent of any specific functional form, is quantified.

[0043] For example, the specific process for calculating the nonlinear mutual information value between the spatially correlated data streams corresponding to the source node and the sink node is as follows: Obtain the spatial correlation data stream corresponding to the source node, denoted as discrete time series. And the spatial correlation data stream corresponding to the sink node, denoted as discrete time series. , where N is the total number of samples after time phase alignment; Using a state-space partitioning method (such as histogram estimation), the numerical ranges of time series X and Y are divided into multiple discrete state intervals; all discrete states of X constitute the state set. The set of all discrete states of Y constitutes the state set. .

[0044] By statistically analyzing the frequency of data points in time series X falling into each state interval, the marginal probability distribution of time series X can be calculated. Similarly, the marginal probability distribution of the time series Y is calculated. ; Statistical time series X and Y data pairs at the same time position The frequency of both falling into the corresponding state interval is used to calculate the joint probability distribution of the source node and the sink node. ; Based on Shannon's information theory, the nonlinear mutual information value between the spatially correlated data streams corresponding to the source node and the sink node is calculated. The calculation formula is: ; When the joint probability p(x,y)=0, the corresponding summation term takes the value of 0; in, This represents the nonlinear mutual information value, and its unit is bit. This represents the discrete state after the time series X is divided. This represents the discrete state after the time series Y is divided; The source node data is in a state The marginal probability, The marginal probability of the sink node data being in state y; The source node data is in a state And the joint probability that the sink node data is in state y; This is a logarithmic operation with base 2.

[0045] By calculating the divergence between the product of the joint probability distribution and the marginal probability distribution, the nonlinear coupling strength between the source node and the sink node, which is independent of a specific functional form, is quantified; the higher the value of mutual information, the stronger the nonlinear correlation between the two sets of data streams.

[0046] Connectivity analysis results are generated based on the comparison results: If the absolute value of the linear Pearson correlation coefficient is greater than or equal to the linearity determination threshold, then it is determined that there is linear connectivity within the adjacent spatial node group. If the absolute value of the linear Pearson correlation coefficient is less than the linearity threshold, and the nonlinear mutual information value is greater than or equal to the mutual information threshold, then it is determined that nonlinear connectivity exists within the adjacent spatial node group. Based on the above determination, a connectivity analysis result including the connectivity type (linear or nonlinear) is output.

[0047] In some specific embodiments, the process of establishing a connectivity coupling model for adjacent spatial node groups based on connectivity analysis results in step two is as follows: Based on the connectivity types included in the connectivity analysis results, select the corresponding basic model architecture for the adjacent spatial node group: If the connectivity type is linear connectivity, then a linear autoregressive distributed lag model is chosen as the basic model architecture. By constructing a linear transfer matrix, the process of spatial correlation data flow from the source node being transmitted to the sink node at a fixed ratio and a fixed time difference is fitted. If the connectivity type is nonlinear connectivity, then a nonlinear dynamic mapping function based on a multilayer perceptron is chosen as the basic model architecture. The nonlinear fitting capability of the activation function in the multilayer perceptron network layer is used to fit the abrupt or exponential response process of the spatial correlation data flow of the source node to the spatial correlation data flow of the sink node. The historical spatial correlation data streams of the source nodes and sink nodes are divided into training sets and validation sets; The training set is input into the selected basic model architecture for parameter fitting and iterative training. By minimizing the prediction error between the source node input data and the sink node actual data, the coupling weight coefficients in the basic model architecture are determined, completing the model training. The trained model is then encapsulated as the adjacent spatial node group connectivity coupling model. The adjacent spatial node group connectivity coupling model can accurately quantify and simulate the specific numerical evolution law of the transfer of source node ecological indicator data fluctuations to sink nodes within the adjacent spatial node group.

[0048] Step 3: Obtain the shared spatial nodes between each adjacent spatial node group, and use the shared spatial nodes as cascade transmission relays to encapsulate the connectivity coupling model of all adjacent spatial node groups into a global connectivity coupling model; In some specific embodiments, the process of obtaining the shared space nodes between each adjacent space node group in step three, using the shared space nodes as cascade relays, and encapsulating the connectivity coupling model of all adjacent space node groups into a global connectivity coupling model is as follows: Obtain all established adjacent spatial node groups, extract the spatial node identifiers contained in each adjacent spatial node group; perform set intersection comparison on the extracted spatial node identifiers, and extract and mark the spatial nodes that exist in at least two different adjacent spatial node groups at the same time as shared spatial nodes. It should be noted that the shared space nodes play a pivotal role in the ecological transmission network, connecting upstream and downstream. For example, a certain shared space node is both the sink node of the upstream adjacent space node group and the source node of the downstream adjacent space node group. Using the shared spatial nodes as cascade relays, a node state cascade transfer equation is established. Specifically, the output of the upstream adjacent spatial node group connectivity coupling model (i.e., the calculated value of the spatial correlation data stream received by the shared spatial node as the sink node) is directly mapped to the input of the downstream adjacent spatial node group connectivity coupling model (i.e., the initial state value output by the same shared spatial node as the source node) as a dynamic input variable. The node state cascade transfer equation breaks down the data gaps between independent adjacent spatial node group connectivity coupling models, realizing the cross-level continuous transmission of spatial correlation data streams. Based on the established node state cascade transfer equation, tensor splicing technology is used to dimensionally align and splice the linear transfer matrices of the connectivity coupling model of all independent adjacent spatial node groups or the activation functions in the multilayer perceptron network layer; taking each spatial node as a graph node, the causal transmission direction determined within each adjacent spatial node group as the direction of the directed edge, and the corresponding causal significance P-value or nonlinear mutual information value as the weight of the directed edge, the spliced ​​overall structure is integrated into a global directed graph transmission network; The global directed graph transmission network is subjected to global connectivity and loop closure verification to ensure that all spatial nodes are included in the network and that the data flow transmission path conforms to the causal transmission direction determined in step one. After the verification passes, the global directed graph transmission network is encapsulated as a whole to generate the global connectivity coupling model. The global connectivity coupling model breaks the limitations of local space and can simulate and output the ripple-like cascading response results of local ecological scheduling events throughout the entire target area in one go.

[0049] Combination Figure 2 As shown, Figure 2 This is a schematic diagram of the causal significance connectivity matrix provided in an embodiment of the present invention. When constructing the connectivity coupling model of adjacent spatial node groups, the Granger causal significance P-values ​​between each node were calculated. For example... Figure 2 As shown in the heatmap matrix, the darker the color, the smaller the P-value, and the more significant the causal transmission relationship. For example, the P-value of node 0 (source node) to node 2 (sink node) is 0.01, which is much smaller than the preset linear significance threshold (such as 0.05). Based on this, it is determined that there is a very strong directed data flow transmission relationship between the two, and the directed edges in graph theory are extracted accordingly. Thus, the hidden ecological transmission topology network in the real geographical environment is objectively restored in a data-driven manner.

[0050] Combination Figure 3 As shown, Figure 3 The schematic diagram of the global directed graph transmission network topology provided in the embodiments of the present invention illustrates the physical process of encapsulating a locally independent model into a globally connected coupled model.

[0051] like Figure 3 As shown, the target region was initially divided into multiple local adjacent spatial node groups (groups A, B, and C, indicated by shaded ellipses in the figure). By comparing the intersection of sets, nodes that exist in multiple groups simultaneously were extracted and marked as shared spatial nodes (nodes N4 and N7 highlighted in the figure).

[0052] During the encapsulation process, a cascaded transmission equation for node states was established using shared space nodes as cascaded transmission relays. For example, shared node N4 is both the sink node of upstream group A (receiving data streams from N1, N2, and N3) and the source node of downstream group B (outputting data streams to N5 and N6). Using tensor splicing technology, the linear transmission matrices of group A and group B are dimensionally aligned and connected at node N4, thereby breaking down data gaps. The directed edges in the graph represent definite causal transmission directions. Through this topology reconstruction mechanism of local grouping-shared relay-global splicing, this invention successfully constructs a global directed graph transmission network covering the entire domain. This allows local scheduling disturbances of a single node (such as implementing pollution reduction at node N1) to ripple like arrows, accurately propagating across levels to the end nodes (such as nodes N8 and N9), providing a solid network architecture foundation for subsequent cross-regional global scheduling.

[0053] Step 4: When a candidate cross-regional scheduling event is generated, the spatial correlation data stream of the target spatial node corresponding to the candidate cross-regional scheduling event is obtained and updated. The updated spatial correlation data stream is input into the global connectivity coupling model, and the spatial correlation data stream of each global spatial node in the next control cycle is output. In some specific embodiments, the process of obtaining and updating the spatial correlation data stream of the target spatial node corresponding to the candidate cross-region scheduling event when a candidate cross-region scheduling event is generated is as follows: Obtain the generated candidate cross-regional scheduling events, extract the control location information and control intensity parameters involved in the candidate cross-regional scheduling events, and determine the corresponding target space node based on the control location information; It should be noted that the candidate cross-regional scheduling events are virtual ecological intervention schemes, which correspond to the aforementioned multi-dimensional ecological indicator data, such as emission quota reduction schemes for air pollutant concentrations, cross-basin ecological water replenishment schemes for water eutrophication, and carbon quota adjustment schemes for carbon emissions. The initial spatial correlation data stream of the target spatial node in the current time window is obtained from the historical ecological database or the monitoring equipment deployed at each spatial node, and used as the baseline state data. The regulation intensity parameter is set to the expected numerical change that matches the ecological indicators in the initial spatial correlation data stream, and the expected numerical change is added to the initial spatial correlation data stream as a perturbation variable. Based on the preset effective time of the candidate cross-regional scheduling event, the corresponding timestamp in the initial spatial correlation data stream is located, and this timestamp is used as the time series node where the perturbation takes effect.

[0054] Specifically, for the aforementioned multidimensional ecological indicator data, the superposition process of the perturbation variables is carried out independently and without interference, including: If the candidate cross-regional scheduling event is a emission quota reduction scheme for atmospheric pollutant concentration, then the concentration reduction amount corresponding to the scheme is obtained as a disturbance variable, and at the time series node where the disturbance takes effect and its subsequent nodes, the atmospheric pollutant concentration values ​​in the initial spatial correlation data stream are subtracted. If the candidate cross-regional scheduling event is a cross-basin ecological water replenishment scheme for water eutrophication, then the expected decrease in the index brought about by the water replenishment scheme is obtained as a disturbance variable. At the time series node where the disturbance takes effect and its subsequent nodes, the water eutrophication index in the initial spatial correlation data stream is subtracted. If the candidate cross-regional scheduling event is a carbon quota adjustment scheme for carbon emissions, then the carbon emission increase or decrease corresponding to the adjustment scheme is obtained as a disturbance variable. At the time series node where the disturbance takes effect and its subsequent nodes, the carbon emission values ​​in the initial spatial correlation data stream are added or subtracted accordingly. The initial spatial correlation data stream after superimposing perturbation variables is re-smoothed and normalized to generate an updated spatial correlation data stream. This updated spatial correlation data stream serves as the immediate response input source for the target spatial node after experiencing the candidate cross-regional scheduling event.

[0055] In some specific embodiments, the process of inputting the updated spatial correlation data stream into the global connectivity coupling model and outputting the spatial correlation data stream of each global spatial node in the next regulation cycle in step four is as follows: The updated spatial correlation data stream of the target spatial node, and the initial spatial correlation data stream of other non-target spatial nodes are synchronously input into the corresponding input end of the global connectivity coupling model as the initial boundary conditions of the global directed graph transmission network. The forward propagation inference is initiated within the global connectivity coupling model. The updated spatial correlation data stream first performs first-level numerical propagation to the adjacent sink node within the local adjacent spatial node group to which the target spatial node belongs, based on the connectivity type (linear transfer matrix or nonlinear dynamic mapping function) established in step two. When the spatially correlated data stream of the first-level numerical transmission reaches the shared spatial node, it triggers the node state cascade transmission equation established in step three. At this time, the shared spatial node plays the role of a cascade transmission relay. Its actual physical meaning is that after aggregating and updating the received changes in the upstream spatially correlated data stream, it is seamlessly converted into the source node input variables of the downstream adjacent spatial node group, thereby crossing the local model boundary and triggering the second-level and deeper-level chain transmission. The above-mentioned transmission and relay process is continuously iterated in the entire global connectivity coupling model according to the preset time step (i.e., the evolution interval on the time series) until the predicted time span reaches the preset next regulation cycle. When the simulation reaches the next control cycle, forward propagation stops, and the final state tensors of all graph nodes in the global connectivity coupling model are obtained. These tensors are then denormalized to restore data with actual physical dimensions, thereby outputting the spatial correlation data stream of all spatial nodes in the next control cycle. The output fully presents the waveform of the real impact of local scheduling events on the global ecological pattern in future cycles after fermentation through the spatial correlation network.

[0056] Step 5: Based on the spatial correlation data stream of each spatial node for the next regulation cycle, calculate the ecological pressure index of each spatial node and compare it with the corresponding carrying capacity red line threshold to complete the prediction of ecological carrying capacity evolution. In some specific embodiments, the specific process of calculating the ecological pressure index of each spatial node in step five, based on the spatial correlation data stream of each spatial node in the next regulation cycle, is as follows: For any given spatial node, obtain multidimensional ecological indicator data for that time node from the spatial correlation data stream of its next regulation cycle; It should be noted that, since the spatial correlation data stream contains multi-dimensional ecological indicators such as atmospheric pollutant concentration, water eutrophication index and carbon emissions, in order to assess the overall ecological carrying capacity of the spatial node from a macro perspective and construct a unified global scheduling optimization benchmark, these indicators with different dimensions and different degrees of impact on the overall ecology need to be standardized and weighted and integrated. The range standardization method is used to eliminate dimensional differences in the ecological indicator data of each dimension. Specifically, for the first... Characteristic values ​​of ecological indicators Using the preset historical maximum value of this indicator and historical minimum Perform normalization calculations to obtain standardized values. The calculation formula is as follows: ; Obtain the feature weight coefficients corresponding to the pre-defined ecological indicator data of each dimension. And satisfy ,in This represents the total number of dimensions for the ecological indicators. The feature weighting coefficients are used to objectively quantify the disaster-causing contribution or pressure percentage of each ecological indicator in the current spatial node's ecosystem. Standardized values ​​for each dimension Its corresponding feature weight coefficient By performing a weighted summation, a dimensionless comprehensive scalar value is obtained. The calculation formula is as follows: ; The comprehensive scalar value Defined as the ecological pressure index of the spatial node in the next regulatory cycle. This ecological pressure index intuitively and quantitatively reflects the comprehensive ecological load level that the spatial node will bear in future cycles after cross-regional cascading transmission.

[0057] In some specific embodiments, the specific process of step five, and comparing it with the corresponding carrying capacity red line threshold to complete the prediction of ecological carrying capacity evolution, is as follows: Obtain the carrying capacity red line threshold corresponding to each spatial node; the carrying capacity red line threshold is a safety extreme value pre-set based on the geographical environment base, historical ecological monitoring limit value and statutory environmental protection standards of each spatial node, and its physical meaning represents the upper limit of the maximum ecological pressure index that the spatial node can withstand without irreversible ecological degradation (such as black and odorous water bodies and severe air pollution); since the ecological capacity of each node is different, the carrying capacity red line threshold of different spatial nodes can have different independent values; The calculated ecological pressure index of each spatial node is compared with its corresponding carrying capacity red line threshold one by one. Based on the comparison results, the future ecological status of each spatial node is classified and marked: If the ecological pressure index of a spatial node is less than or equal to the corresponding carrying capacity red line threshold, then the evolution prediction state of the spatial node is marked as ecologically safe. If the ecological pressure index of a certain spatial node is greater than the corresponding carrying capacity red line threshold, the evolution prediction state of the spatial node is marked as ecological overload, and the difference exceeding the threshold is calculated as the overload risk quantity. The system acquires the predicted evolution status and overload risk of all spatial nodes, generating a global ecological risk distribution map or early warning list for the target area in the next regulation cycle, thereby completing the ecological carrying capacity evolution prediction. The evolution prediction results not only indicate which nodes will experience crises in the future, but also clarify the severity of the crises, providing clear targets for subsequent scheduling adjustments.

[0058] Step Six: If the evolution prediction results show that the ecological pressure index of any spatial node exceeds the carrying capacity red line threshold, then adjust the ecological pressure index of all spatial nodes to be within the carrying capacity red line threshold, and complete the cross-regional scheduling.

[0059] In some specific embodiments, in step six, if the evolutionary prediction results show that the ecological pressure index of any spatial node exceeds the carrying capacity red line threshold, then the process of adjusting until the ecological pressure index of all spatial nodes is within the carrying capacity red line threshold and completing the cross-regional scheduling is as follows: Based on the evolution prediction results output in step five, spatial nodes whose state is marked as ecological overload are obtained, and the specific difference between their ecological pressure index and the carrying capacity red line threshold is calculated. The specific difference is defined as the overload risk amount. Based on the global directed graph transmission network topology in the global connectivity coupling model established in step three, reverse source analysis is performed with the overload risk as the target to locate the upstream source space node that causes the space node overload and the key transmission path.

[0060] For example, the specific process of performing reverse source tracing analysis is as follows: In the global directed graph transmission network, the spatial node whose state is marked as ecological overload is set as the source tracing starting node. ; Based on a graph theory search algorithm, the algorithm traverses the reverse path of the global directed graph transmission network, checking all upstream paths that can flow to the overloaded target node, and obtaining all nodes that can reach the source node. The set of connected paths , where L is the total number of connected paths; For any connected path in the set Assume the path contains a sequence of K nodes. ,in ,and The starting source node of the path; obtain each directed edge on the connected path. weight , namely, the conduction strength weight; the conduction strength weight is obtained by normalizing the causal significance P-value or the nonlinear mutual information value obtained in the previous steps; Optional weight construction methods include: Granger causality weights: P represents the causal significance p-value of the corresponding edge; Nonlinear mutual information weights: Perform extreme value normalization on the mutual information values ​​and map them to the range of 0-1.

[0061] Calculate the connected path based on the weights of each directed edge in the connected path. Comprehensive transmission contribution The calculation formula is the product of the propagation intensity weights along the path: ; If the path contains only a single node as the traceability starting point, the comprehensive conduction contribution is assigned a default value of 1; Repeat the above calculation to obtain the comprehensive conduction contribution of all paths in the connected path set; then find the target path that maximizes the comprehensive conduction contribution , the mathematical expression is: ; When multiple paths have the same maximum conduction contribution, the path with the least number of nodes is selected as the optimal key conduction path.

[0062] The obtained is marked as a key conduction path, and the foremost node of the key conduction path is positioned as the upstream source spatial node that causes the overload of said spatial node.

[0063] According to the result of the reverse traceability analysis, parameter adjustment is performed on the candidate cross-regional scheduling events initially generated in step four.

[0064] Specifically, since the overload risk amount of the end node is calculated based on the weighted fused comprehensive scalar value, first, using the feature weight coefficient in step five, said overload risk amount is inversely decoupled into the index excess amount corresponding to a single ecological indicator (i.e., air pollutant concentration, water eutrophication index or carbon emission); Then, combined with the comprehensive conduction contribution of the key conduction path obtained from the aforementioned reverse traceability analysis, the parameter compensation amount that the upstream source spatial node needs to adjust in order to offset the index excess amount is calculated; the regulation intensity parameter corresponding to said upstream source spatial node is updated according to said parameter compensation amount, so as to generate an updated candidate cross-regional scheduling event.

[0065] During the parameter update process, the linkage is carried out in strict accordance with the disturbance superposition mechanism in step four: If the original candidate cross-regional scheduling event is sewage discharge quota reduction or carbon quota reduction, said regulation intensity parameter is corrected in the direction of increasing the reduction range; If it is cross-basin ecological water replenishment, said regulation intensity parameter is corrected in the direction of increasing the water replenishment gain range. The updated regulation intensity parameter will be directly used as the new expected numerical change (i.e., the new disturbance variable) for subsequent iterative deduction; ​The updated candidate cross-regional scheduling events are re-input into the global connectivity coupling model for a new round of forward cascading propagation simulation, and the ecological pressure index of each spatial node is recalculated and compared with the corresponding carrying capacity red line threshold. Based on the aforementioned source tracing compensation mechanism, the parameter adjustment and comparison verification process constitutes a closed-loop directional iterative approximation process. Simultaneously, to prevent the system from falling into an infinite loop due to physical carrying capacity limitations, a maximum iteration count threshold is pre-set. The above-described directional iterative approximation process continues until the evolutionary prediction results show that the ecological pressure index of all spatial nodes is less than or equal to their corresponding carrying capacity threshold. At this point, the iteration stops, and the network equilibrium point of global ecological carrying capacity is reached. If, when the maximum number of iterations threshold is reached, there are still nodes whose ecological pressure index exceeds the threshold, the set of parameters with the smallest sum of global overload risk in all iterations is obtained as the suboptimal compromise solution, and a system warning is triggered. The cross-regional scheduling event parameters (or the parameters corresponding to the suboptimal compromise solution) corresponding to the equilibrium point are obtained as the final balanced scheduling scheme. This balanced scheduling scheme is transformed into specific control execution instructions and sent to the physical execution terminals corresponding to each spatial node in the target area (e.g., hydraulic gate control systems for eutrophication regulation, industrial sewage valve controllers for air pollutant reduction, and intelligent power-limiting gateways or carbon capture device controllers for carbon emission reduction). The scheduling parameters are implemented through the actions of the physical execution terminals, thereby completing the actual cross-regional scheduling.

[0066] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation, characterized in that, include: Acquire the spatial correlation data stream of each spatial node in the target area, and filter out multiple adjacent spatial node groups through a dynamic time warping algorithm based on causality test; Analyze the connectivity between spatially correlated data streams within any adjacent spatial node group, and establish a connectivity coupling model for adjacent spatial node groups based on the connectivity analysis results; Obtain the shared spatial nodes between each adjacent spatial node group, and use the shared spatial nodes as cascade transmission relays to encapsulate the connectivity coupling model of all adjacent spatial node groups into a global connectivity coupling model; When a candidate cross-regional scheduling event is generated, the spatial correlation data stream of the target spatial node corresponding to the candidate cross-regional scheduling event is obtained, and the spatial correlation data stream is input into the global connectivity coupling model. Through the cascade transmission relay of shared spatial nodes, the spatial correlation data stream of each global spatial node in the next control cycle is output. Based on the spatial correlation data stream of each spatial node in the next regulation cycle, calculate the ecological pressure index of each spatial node and compare it with the corresponding carrying capacity red line threshold to complete the prediction of ecological carrying capacity evolution. If the evolution prediction results show that the ecological pressure index of any spatial node exceeds the carrying capacity red line threshold, then the ecological pressure index of all spatial nodes will be adjusted to be within the carrying capacity red line threshold, thus completing the cross-regional scheduling.

2. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process of obtaining the spatial correlation data stream of each spatial node to which the target region belongs is as follows: The target area is divided into multiple independent spatial nodes. By deploying physical monitoring equipment at each spatial node or calling historical ecological databases, multidimensional ecological indicator data of each spatial node within a preset time window are obtained. The multidimensional ecological indicator data includes air pollutant concentration, water eutrophication index and carbon emissions. The multidimensional ecological indicator data are arranged in time series to construct a spatial correlation data stream for each spatial node.

3. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific screening process for the adjacent spatial node group is as follows: The dynamic time warping algorithm is used to calculate the warping distance between the spatial correlation data streams of any two spatial nodes. If the warping distance is less than a preset warping distance threshold, the Granger causality test is used to calculate the causal significance P-value and causal propagation direction between the spatial correlation data streams of the two spatial nodes. If the causal significance P-value is less than a preset significance threshold, the two spatial nodes are determined to be adjacent spatial node groups.

4. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process for analyzing the connectivity between spatially correlated data streams within any adjacent spatial node group is as follows: Based on the causal transmission direction determined by the causal significance P-value, the two spatial nodes within an adjacent spatial node group are marked as the source node and the sink node, respectively. Calculate the linear Pearson correlation coefficient and nonlinear mutual information value between the spatially correlated data streams corresponding to the source and sink nodes; If the absolute value of the linear Pearson correlation coefficient is greater than or equal to the preset linearity threshold, then it is determined that there is linear connectivity within the adjacent spatial node group. If the absolute value of the linear Pearson correlation coefficient is less than the linear determination threshold, and the nonlinear mutual information value is greater than or equal to the preset mutual information determination threshold, then it is determined that there is nonlinear connectivity within the adjacent spatial node group.

5. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process for establishing the connectivity coupling model of the adjacent spatial node group is as follows: If the connectivity type is linear connectivity, then an autoregressive distributed lag model based on the linear transfer matrix is ​​established to fit the linear numerical transmission from the source node to the sink node. If the connectivity type is nonlinear connectivity, a nonlinear dynamic mapping function based on the multilayer perceptron network layer is established to fit the nonlinear response transmission of the source node to the sink node.

6. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific encapsulation process of the global connectivity coupling model is as follows: Obtain the spatial node identifiers of each adjacent spatial node group, perform set intersection comparison, and mark the spatial nodes that exist in at least two different adjacent spatial node groups as shared spatial nodes. A node state cascade transfer equation is established using shared space nodes as cascade transmission relays, and the output of the connectivity coupling model of the upstream adjacent space node group is mapped and connected to the input of the connectivity coupling model of the downstream adjacent space node group. Based on the node state cascade transfer equation, tensor splicing technique is used to dimensionally align and splice the linear transfer matrix of the connectivity coupling model of independent adjacent spatial node groups or the multilayer perceptron network layer. Using each spatial node as a graph node, the causal transmission direction as the direction of the directed edges, and the causal significance P-value or nonlinear mutual information value as the weight of the directed edges, the spliced ​​overall structure is merged into a global directed graph transmission network, and the whole is encapsulated to generate a global connectivity coupling model.

7. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process of outputting the spatial correlation data stream of all global spatial nodes in the next control cycle is as follows: Obtain the control location information and control intensity parameters of candidate cross-regional scheduling events to determine the target spatial nodes; The control intensity parameter is set as the expected value change as the disturbance variable, and the corresponding time series node in the initial spatial correlation data stream is located according to the preset effective time. At time series nodes and subsequent nodes, numerical addition and subtraction operations are performed independently on the corresponding atmospheric pollutant concentrations, water eutrophication indices, or carbon emissions. The data sequence after superimposed perturbation variables is smoothed and normalized to generate an updated spatial correlation data stream; The updated spatial correlation data stream of the target spatial node, as well as the initial spatial correlation data stream of other non-target spatial nodes, are synchronously input into the global connectivity coupling model, and the spatial correlation data stream of each global spatial node in the next regulation cycle is output.

8. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process for calculating the ecological pressure index of each spatial node is as follows: Based on the spatial correlation data stream of the next regulation cycle, obtain multi-dimensional ecological indicator data of spatial nodes under the next regulation cycle; The range standardization method is used to eliminate the dimensional differences in ecological indicator data of each dimension, and standardized values ​​are obtained. Obtain the feature weight coefficients corresponding to the ecological indicator data of each dimension, and calculate the weighted sum of the standardized values ​​of each dimension and their corresponding feature weight coefficients to obtain a dimensionless comprehensive scalar value, which serves as the ecological pressure index of the spatial node.

9. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process for completing the ecological carrying capacity evolution prediction is as follows: The calculated ecological pressure index of each spatial node is compared with the carrying capacity red line threshold corresponding to each spatial node. If the ecological pressure index is less than or equal to the corresponding carrying capacity red line threshold, then the corresponding spatial node is marked as ecologically safe. If the ecological pressure index is greater than the corresponding carrying capacity red line threshold, the corresponding spatial node is marked as ecological overload, and the difference exceeding the carrying capacity red line threshold is calculated as the overload risk.

10. The method for predicting the evolution of ecological carrying capacity and cross-regional scheduling based on spatial correlation according to claim 1, characterized in that, The specific process for completing the cross-regional scheduling is as follows: Using spatial nodes marked as ecological overload as the starting point for tracing the source, a graph theory search algorithm is used to traverse the connected paths composed of directed edges in the reverse direction of the global directed graph transmission network. Based on the weights of each directed edge on the connected path, the comprehensive transmission contribution of each connected path is calculated. The connected path with the largest comprehensive transmission contribution is identified as the key transmission path, and the node at the forefront of the key transmission path is identified as the upstream source space node. The overload risk is decoupled into the over-limit of a single ecological indicator by using the feature weight coefficient, and the parameter compensation of the upstream source spatial node is calculated by combining the comprehensive transmission contribution. The regulation intensity parameters are updated based on the parameter compensation amount to generate new expected numerical changes in the direction of increasing the reduction or gain. When the ecological pressure index of all spatial nodes is less than or equal to their corresponding carrying capacity red line threshold, the corresponding cross-regional scheduling event parameters are obtained as a balance scheduling scheme, which is then converted into control execution instructions and sent to the corresponding physical execution terminals in the target area. The physical execution terminals include water conservancy gate control systems, industrial sewage valve controllers or intelligent power limiting gateways and carbon capture equipment controllers.

Images