A carbon emission key node searching method of a power system

Abstract

The application discloses a kind of carbon emission key node searching method of power system, the method steps include: S1, determine the several characteristics of all nodes in power system, several The characteristics are divided into several characteristic groups, in several The characteristics in the characteristic group, the correlation between characteristics is not greater than correlation threshold value;S2, obtain the characteristic key coefficient of the node in the power system in several The characteristics in the characteristic group;S3, according to the several characteristic key coefficients of the node, obtain the weighted characteristic key coefficient of the node, and construct connection diagram, determine carbon emission key node according to the connection diagram;Through the method involved in the application, the efficiency and accuracy of carbon emission key node acquisition are improved.

Description

Technical Field

[0001] This invention relates to a method for searching for critical carbon emission nodes in power systems, specifically for searching for critical carbon emission nodes in complex power systems. Background Technology

[0002] The power system is the core hub of energy transition. With the large-scale integration of renewable energy and the increasing proportion of power electronic devices, the spatiotemporal distribution of system carbon emissions exhibits high heterogeneity and dynamic coupling characteristics. Traditional "homogenization across the entire domain" emission reduction strategies are ill-equipped to handle this complexity—failure to accurately pinpoint key carbon emission nodes will lead to resource misallocation and efficiency losses. For example, some high-energy-consuming load centers in the power grid rely on long-distance coal-fired power transmission, and their electricity consumption behavior indirectly pushes up system carbon emissions through the amplification effect of network losses. However, existing technologies lack the ability to identify such hidden hotspots, causing clean energy investment to concentrate on low-efficiency segments.

[0003] The physical structure and operation of the power system determine the "leverage effect" of carbon emissions. Minor disturbances at key nodes (such as hub substations and important transmission channels) can cause significant fluctuations in the overall carbon intensity of the power grid through power flow redistribution. Taking a regional interconnected power grid as an example, a blockage in a critical interconnection line may force the receiving-end grid to call up local high-carbon units, leading to a surge in marginal carbon emissions.

[0004] The high proportion of renewable energy connected to the grid further highlights the necessity of identifying key nodes. The volatility of wind and solar power necessitates frequent adjustments to thermal power units, resulting in significant differences in carbon emission intensity under different operating conditions. If the spatiotemporal migration patterns of carbon emissions caused by flexible adjustments in thermal power cannot be identified, simply shutting down units may disrupt the system's flexible supply and instead stimulate a rebound in carbon emissions at other nodes. Furthermore, the convergence and interaction of distributed resources on the user side and electric vehicles are reshaping the boundaries of carbon emission responsibility. Only by identifying key carbon emission nodes can a coordinated carbon reduction mechanism integrating "source-grid-load-storage" be constructed.

[0005] In existing technologies, key carbon emission nodes in power systems are identified through the coefficient of variation weighted average (TOPSIS) method and the construction of a two-stage model. For example, the China Patent Network discloses a method for identifying key carbon emission nodes in power systems, with application number CN202311631818.9. In this patent's technical solution, key carbon emission nodes are identified through the coefficient of variation weighted average (TOPSIS) method and the construction of a two-stage model. However, in practical applications, the various characteristics of nodes are somewhat correlated, so the coefficient of variation weighted average (TOPSIS) method cannot reflect the importance of each node in complex power systems. Furthermore, in constructing the two-stage model, this patent uses the Dijkstra algorithm and particle swarm optimization algorithm to obtain the model, and designs a high penalty coefficient when the graph involved in the model is not connected. However, this patent does not involve a mechanism for constructing graph connectivity, resulting in a slow convergence speed in practical applications, making it impossible to identify key carbon emission nodes in complex power systems within a suitable timeframe. Summary of the Invention

[0006] The purpose of this invention is to provide a method for accurately identifying key carbon emission points in complex power systems.

[0007] Another objective of this invention is to provide a method for improving the search speed for key carbon emission nodes in complex power systems.

[0008] To achieve the above objectives, the present invention provides a method for searching for key carbon emission nodes in a power system. The method includes the following steps: S1, determining several features of all nodes in the power system, dividing the several features into several feature groups, wherein the correlation between the features in the several feature groups is not greater than a correlation threshold; S2, obtaining the feature key coefficients of the nodes in the power system in the several feature groups; S3, obtaining the weighted feature key coefficients of the nodes based on the several feature key coefficients of the nodes, constructing a connection graph based on the weighted feature key coefficients, and determining the key carbon emission nodes based on the connection graph.

[0009] Preferably, step S1 includes: constructing a feature map using the correlation between the features as a distance metric; segmenting the feature map to obtain feature sub-maps; and segmenting the feature sub-maps if, in the feature sub-maps, at least two of the features have a correlation greater than the correlation threshold.

[0010] Preferably, step S1 includes: using the correlation between the features as a distance metric, performing K-means clustering to obtain several clusters; if in the cluster, at least two of the features have a correlation greater than the correlation threshold, then performing K-means clustering on the cluster; wherein, in the K-means clustering, the number of clusters is at least 2.

[0011] Preferably, step S3 includes: for a certain feature group, the weight value corresponding to the feature group is related to the number of feature types in the feature group; the sum of the weight values ​​corresponding to the feature group is 1.

[0012] Preferably, step S1 includes: dividing the plurality of features into two feature groups: a topology-related feature group and a topology-unrelated feature group.

[0013] Preferably, step S3 includes: the weight value corresponding to the topology-related feature group is related to the network complexity of the power system; the sum of the weight values ​​of the topology-related feature group and the topology-unrelated feature group is 1.

[0014] Preferably, the correlation between the features includes a weighted sum of linear correlations and nonlinear correlations between the features.

[0015] Preferably, in step S3, the process for obtaining the key feature coefficients of the node includes: Based on the topology-related feature group and the topology-irrelevant feature group, construct a topology-related feature matrix and a topology-irrelevant feature matrix; perform KPCA processing on the topology-related feature matrix and the topology-irrelevant feature matrix to obtain the processed feature matrix; and obtain the feature key coefficients of the node based on the processed feature matrix.

[0016] Preferably, in step S3, the process of obtaining the key feature coefficients of the node includes: determining the coefficient of variation of each feature in the feature group, determining the corresponding weight value of each feature based on the coefficient of variation, processing the feature of the data corresponding to each node based on the weight value, and obtaining the probability function value of the processed data to obtain the key feature coefficients of the node.

[0017] Preferably, the process of constructing the connection graph includes: in the topology graph of the power system, the probability of selecting an edge is positively correlated with the magnitude of the weighted feature key coefficients of the nodes at both ends of the edge.

[0018] The beneficial effect of this invention is that it can more accurately determine the importance of nodes in complex power systems, thereby enabling more precise identification of key carbon emission nodes in power systems. Another beneficial effect of this invention is that it is more efficient in finding key carbon emission nodes in complex power systems, enabling the search for key carbon emission nodes in a shorter time. Attached Figure Description

[0019] Figure 1 This invention relates to a flowchart of a method.

[0020] Figure 2 This is a network topology diagram of a power system.

[0021] Figure 3 It is a topology graph of a set of feature nodes.

[0022] Figure 4 This is a schematic diagram of a subgraph of the node-related graph.

[0023] Figure 5 It is a fully connected subgraph of the node-related graph.

[0024] Figure 6 This is a subgraph of the node-related graph. Detailed Implementation

[0025] Example 1: This example discloses a specific implementation flow of the method involved in the present invention. (See attached document.) Figure 1 .

[0026] The specific implementation process of the method involved in this embodiment is as follows.

[0027] S1. Obtain the topology network information of the power system, obtain several features of all nodes in the topology network information, divide the features into several feature groups, and ensure that the correlation between features in the feature groups is not greater than the correlation threshold.

[0028] In the topology information of a power system, the characteristics of a node include: the node's carbon flow intensity, the node's carbon inflow, the node's carbon flow rate standard deviation, the node's load carbon flow rate, the node's unit injected carbon flow rate, the node's carbon flow rate correlation, the node's standard deviation correlation balance, the node's improved clustering coefficient, the node's Shell contribution, and the difference between the node's maximum and minimum carbon flow rates.

[0029] The carbon flow intensity of a node reflects the relationship between electrical power flow and carbon emission flow; specifically, it is a physical quantity related to the carbon emission factor, which measures the amount of carbon emissions generated per unit of electricity consumed. A higher carbon flow intensity at a node indicates a greater amount of carbon emissions generated per unit of electricity consumed.

[0030] For the i-th node in the topology network information of the power system, the expression for the carbon flow intensity of node i is as follows.

[0031] Its expression is: Among them, P Bj This represents the active power of branch j; P represents the carbon flux density of branch j; Gi This represents the active power generated by the generator set at node i. This represents the carbon flux density of the generator set connected to node i; I * This represents the several layers of adjacent nodes of node i.

[0032] The carbon inflow to a node is caused by the carbon emission flow flowing into the node, while the carbon emission flow flowing out has no contribution to it. The contribution of generator units and other nodes to the carbon potential of a node can be characterized by the carbon inflow to the node.

[0033] For the i-th node in the topology network information of the power system, the expression for the carbon inflow of node i is as follows.

[0034] Its expression is: Among them, R Bj R represents the carbon flow rate from branch j into node i. Gi Let be the carbon flow rate of the branch from the generator set at node i into node i.

[0035] The standard deviation of carbon flow rate at a node reflects the degree of carbon flow rate fluctuation at that node. The higher the standard deviation of carbon flow rate at a node, the higher the degree of carbon-sulfur rate fluctuation at that node; the lower the standard deviation of carbon flow rate at a node, the lower the degree of carbon flow rate fluctuation at that node.

[0036] The expression for the standard deviation of carbon flow rate of the i-th node in the topology network information of the power system is as follows.

[0037] Its expression is: Where T is the total operating time of the power system; CLF i,t Let CLF be the carbon flux at node i during time period t; i,avg Let be the average carbon flow rate of node i during time T.

[0038] The load carbon flow rate of a node, specifically, represents the total greenhouse gas emissions generated by the power supply system per unit time period to meet the energy demands of the node's electrical load. This indicator aims to quantify the carbon intensity associated with node load consumption.

[0039] The expression for the load carbon flow rate of the i-th node in the topology network information of the power system is as follows.

[0040] Its expression is: μL i =Load i *α i , where Load i Let be the load of node i.

[0041] The carbon injection rate of a node is expressed as the amount of carbon emissions generated by the generator units connected to the node per unit time.

[0042] For the i-th node in the topology network information of the power system, the expression for the carbon injection rate of the unit at the i-th node is as follows.

[0043] Its expression is:

[0044] The carbon flow rate correlation of a node indicates the carbon relationship between that node and a certain number of other nodes. Specifically, the higher the carbon flow rate correlation of a node, the more nodes it has a carbon relationship with; conversely, the lower the carbon flow rate correlation of a node, the fewer nodes it has a carbon relationship with.

[0045] The expression for the carbon flow rate correlation of the i-th node in the topology network information of the power system is as follows.

[0046] Its expression is: in, d represents the average carbon flow rate of the adjacent machine nodes of node i; j Let be the degree of node j, which is the neighboring node of node i.

[0047] The standard deviation correlation balance of a node represents the degree to which a particular node influences the carbon flow rate fluctuations of other nodes within a certain range in a power system. A smaller standard deviation correlation balance indicates a smaller influence of that node on the carbon flow rate fluctuations of other nodes within that range; conversely, a larger standard deviation correlation balance indicates a larger influence of that node on the carbon flow rate fluctuations of other nodes within that range.

[0048] For the i-th node in the topology network information of the power system, the expression for the standard deviation of the i-th node related to the balance is as follows.

[0049] Its expression is: Among them, h ij S is the minimum distance between node i and node j; d σ is the set of nodes whose minimum distance to node i is not greater than d; j Let be the standard deviation of the carbon flux at node j; Let be the standard deviation after node i fails.

[0050] The following example will provide a detailed explanation of the variables involved in this feature.

[0051] See Figure 2 ,exist Figure 2 There are nodes 1, 2, 3, 4, 5, 6, 7 and 8. Let’s take node 1 as an example.

[0052] If the minimum distance between node 1 and node 2 is 1, then h 12 The value of h is 1; the minimum distance between node 1 and node 6 is 2, then h 16 The value of h is 2; the minimum distance between node 1 and node 5 is 1, then h 15 The value of h is 1; the minimum distance between node 1 and node 7 is 1, then h 17 The value of h is 1; the minimum distance between node 1 and node 3 is 1, then h 13 The value of h is 1; the minimum distance between node 1 and node 4 is 1, then h 14 The value is 1.

[0053] If the variable d is 1, then nodes 2, 3, 4, 5, and 7 belong to the node set S. d If variable d is greater than 1, then nodes 2 to 8 belong to node set S. d .

[0054] The improved clustering coefficient of nodes is derived from the improvement of local structural entropy. It mainly uses the influence of the local network on the whole network to replace the influence of the nodes on the whole network, and to a certain extent reflects the degree of influence of the nodes on the whole network.

[0055] The expression for the improved clustering coefficient of the i-th node in the topology network information of the power system is as follows.

[0056] Its expression is: PLE i =LE i -CC i ,in, It should be noted that Tc i Let be the number of ternary closures formed by the first-order neighbor nodes of node i.

[0057] The Shell contribution of a node is determined based on the K-shell algorithm. The specific steps of the K-shell algorithm are as follows: First, nodes with a degree of 1 and the edges connected to them are removed until there are no nodes with a degree of 1 left in the topology. During this process, the K-shell value of the removed nodes is 1.

[0058] After eliminating nodes with a degree of 1 from the existing topology, nodes with a degree of 2 and the edges connected to them are removed until no nodes with a degree of 2 remain in the topology. During this process, the K-shell value of the removed nodes is 2.

[0059] This process continues until all nodes in the network topology are removed, and the K-shell value of each node is determined.

[0060] For the i-th node in the detailed topology of the power system, the expression for the shell contribution of the i-th node is as follows.

[0061] Its expression is: Where, k s (i) is the K-shell value of node i; k s (j) is the K-shell value of node j.

[0062] The difference between the maximum and minimum carbon flow rates of a node represents the difference between the peak and low carbon flow rates of the node, and to a certain extent reflects the fluctuation range of the node's carbon flow rate.

[0063] For the i-th node in the detailed topology of the power system, the expression for the difference between the maximum and minimum carbon flow rates of the i-th node is as follows.

[0064] Its expression is: Rg i =CLF i,max -CLF i,min CLF i,max The maximum carbon flux at the i-th node; CLF i,min Let be the minimum carbon flow rate of the i-th node.

[0065] In this embodiment, there are 10 types of node features, and there is a certain correlation between each feature. The correlation includes linear correlation and non-linear correlation.

[0066] If only linear or nonlinear correlation is considered, the linear or nonlinear correlation between features may be small when there is nonlinear or linear correlation between features. This can lead to inaccurate key feature coefficient values ​​of nodes when identifying key carbon emission nodes.

[0067] Therefore, when it is necessary to obtain the correlation between two features more accurately, both linear and nonlinear correlations between the features need to be considered.

[0068] In this embodiment, the correlation between features is the weighted sum of linear correlation and nonlinear correlation between features.

[0069] The linear correlation and non-linear correlation of data are inversely proportional (in this embodiment, outliers are removed without considering them). The higher the linear correlation between features, the lower the non-linear correlation between features; and vice versa.

[0070] Therefore, when the linear correlation between features is high, we should mainly consider the linear correlation between features; when the nonlinear correlation between features is high, we should also consider the nonlinear correlation between features.

[0071] That is, for feature x and feature y, the expression for the correlation between these two features is as follows.

[0072] Its expression is: Where person(x,y) is the Pearson coefficient between feature x and feature y; σ is the adjustment coefficient.

[0073] We obtain the correlation between features and use the correlation between features as a distance metric to construct a fully connected graph. The following example illustrates this.

[0074] See Figure 3 ,exist Figure 3 In the given information, there are features 1 to 4. If the correlation between feature 1 and feature 2 is 0.92, then the edge length between feature 1 and feature 2 is 0.92; if the correlation between feature 1 and feature 3 is 0.12, then the edge length between feature 1 and feature 3 is 0.12; if the correlation between feature 1 and feature 4 is 0.71, then the edge length between feature 1 and feature 4 is 0.71; if the correlation between feature 2 and feature 3 is 0.81, then the edge length between feature 2 and feature 3 is 0.81; if the correlation between feature 2 and feature 4 is 0.42, then the correlation between feature 2 and feature 4 is 0.42; if the correlation between feature 3 and feature 4 is 0.86, then the edge length between feature 3 and feature 4 is 0.86.

[0075] The constructed fully connected graph is segmented according to the correlation threshold to obtain several subgraphs. The specific steps are as follows.

[0076] The fully connected graph is normalized and segmented. After segmentation, the Ncut value of the fully connected graph is minimized. In this embodiment, the fully connected graph is divided into two subgraphs, and the expression for the Ncut value is as follows.

[0077] Its expression is: Where A and B represent the two subgraphs to be divided; cut(A,B) is represented as: ∑ i∈A,j∈B w ij w ij Let A be the length distance between node i and node j; assoc(A,V) is represented as: ∑ i∈A,j∈V w ij V is the set of nodes in a fully connected graph.

[0078] After obtaining the segmented subgraph, traverse all subgraphs. If there is at least one edge between two feature nodes in the subgraph with a length greater than the correlation threshold, then the subgraph is segmented. Repeat the subgraph segmentation process until there are no two feature nodes in the segmented subgraph with a length greater than the correlation threshold. If there are no two feature nodes in the subgraph with a length greater than the correlation threshold, then the subgraph is not segmented.

[0079] This way, several feature groups are obtained, ensuring that the correlation between features in these feature groups does not exceed a correlation threshold.

[0080] S2. Obtain the key feature system of the node in the power system from the plurality of feature groups, the specific process of which is as follows.

[0081] Obtain the key feature coefficients of each node under each feature group.

[0082] If the number of features in a feature group is one, then the key coefficient of each node is obtained according to the one-dimensional Gaussian distribution function. If the number of features in a feature group is not less than one, then the coefficient of variation weight value corresponding to each feature in the feature group is obtained, the feature data of each node is multiplied by the corresponding coefficient of variation weight value, and the key coefficient of each node is obtained according to the multi-dimensional Gaussian distribution function.

[0083] The specific process is as follows.

[0084] In a power system, there are N nodes. If each node has m features under each feature group, then construct an m-row N-column feature matrix.

[0085] For feature groups with more than one feature, the process for determining the key coefficients of node features is as follows.

[0086] In each feature group, the coefficient of variation weight value corresponding to each feature is obtained. Specifically, the data corresponding to each feature in N data points is obtained. For the k-th feature, the obtained data includes: Z k =[x k1 ,x k2 ,x k3 ,...,x kNFor this dataset, the expression for the coefficient of variation weight value corresponding to the k-th feature is: Where, μ k σ is the average value of the dataset corresponding to the k-th feature; k is the standard deviation of the dataset corresponding to the k-th feature; m is the number of feature categories.

[0087] After determining the weight values ​​of the coefficient of variation for each feature, the dataset corresponding to the node is processed. The processed data for node i is as follows:

[0088] Determine the optimal solution for each feature in each feature group, and denot the optimal solution for each feature data as follows: The optimal solution of the feature data is processed as follows: the optimal solution of each feature data is multiplied by the corresponding coefficient of variation weight value. The processed optimal solution of the feature data is:

[0089] It should be noted that the optimal solution for the feature data can be provided by those skilled in the art based on the actual situation of the power system, and will not be elaborated here.

[0090] Based on the multidimensional Gaussian distribution function, the key coefficients of the nodes are obtained, and the characteristic covariance matrix of the nodes is denoted as ∑ m,m , where ∑ i,j =Cov(Z) i Z j )=E(Z i *Z j )-E(Z i )*E(Z j ).

[0091] The numerical value of the multidimensional probability function represents the key feature coefficient of a node. For the key feature coefficient of the i-th node, its expression is: in, Let be the key feature coefficient of node i.

[0092] For a feature group with only one feature, the process for determining the node critical coefficient is as follows.

[0093] In a power system, there are N nodes, and each node has only one feature under this feature group. Construct a feature matrix with 1 row and N columns.

[0094] Determine the optimal solution for each feature in the feature group, denoted as z.

[0095] To obtain the standard deviation corresponding to this feature, the expression for the standard deviation of this feature is: Where, xavg This represents the average value of the data at each node.

[0096] The numerical value of the one-dimensional probability function represents the key characteristic coefficient of a node. For the i-th node in a power system, the expression for the key characteristic coefficient of that node is:

[0097] After determining the key coefficients of each feature group, determine the weight value corresponding to each feature group, and obtain the sum of the key coefficients of the feature group multiplied by the weight value corresponding to the feature group to get the weighted key coefficients.

[0098] In a feature group, if there are more corresponding features, the feature group contains more information, and the confidence level of the node's feature key coefficient is higher. If there are fewer corresponding features, the feature group contains less information, and the confidence level of the node's feature key coefficient is lower.

[0099] Therefore, the same weight cannot be applied to the key feature coefficients of nodes in each feature group. This would distort the weighted key feature coefficients of nodes, making it impossible to measure the importance of nodes in carbon emissions. Instead, a weight should be assigned to the response based on the credibility of each feature group.

[0100] For the k-th feature group, the number of feature types in this feature group is s. k The expression for the weight value corresponding to this feature group is as follows.

[0101] Its expression is: For the i-th feature,

[0102] As can be seen from the above formula, for a certain feature group, the more information the feature group contains, the greater the weight value of the feature group; the less information the feature group contains, the less the weight value of the feature group.

[0103] After obtaining the weight value corresponding to each feature group, the feature key coefficient of the node in each feature group is multiplied by the weight value corresponding to the feature group to obtain the weighted feature key coefficient of the node in each feature group. The weighted key feature coefficients of the node in each feature group are added together to obtain the weighted feature key coefficient of the node.

[0104] In a power system, for the i-th node, the expression for the weighted characteristic key coefficient of that node is as follows.

[0105] Its expression is: Among them, CCc j (x i) represents the key feature coefficient of the i-th node in the j-th feature group; L is the number of feature groups.

[0106] S3. Based on several key feature coefficients of the node, obtain the weighted key feature coefficients of the node, construct a connection graph, and determine the key carbon emission nodes based on the connection graph. The specific process is as follows.

[0107] After determining the weighted key coefficients of the nodes, a connection graph in the topology network information of the power system is constructed based on the weighted key coefficients of each node. This connection graph is a subgraph of the topology network graph in the topology network information.

[0108] The construction process of the connection graph in the topology network information of a power system is as follows.

[0109] In the topology information of the power system, the weighted key characteristic coefficients of the nodes at both ends of each edge are obtained, and the probability of each edge being selected is obtained based on the weighted characteristic coefficients of the nodes at both ends of each edge.

[0110] For a given edge, the probability of selecting that edge is calculated by exponentiating the geometric mean of the weighted key coefficients of the two endpoints of the edge with the natural base as the base. The first probability value of the edge is then divided by the sum of the first probability values ​​of all edges in the power system network to obtain the second probability value of the edge.

[0111] The second probability value for this edge is the probability of selecting this edge.

[0112] In a power system, the expression for the probability of selecting the i-th edge is as follows.

[0113] Its expression is: Where E is the set of all edges in the topology of the power system; WCCc i1 With WCCc i2 It is represented as the weighted key feature coefficient of the nodes at both ends of the i-th edge.

[0114] In the network topology information of the power system, after edge selection, the connection graph is updated for the first time. After updating the connection graph, the edges that are directly connected to the connection graph and have not been selected are traversed to obtain the endpoints of the traversed edges that are not connected to the connection graph.

[0115] After obtaining the endpoints of the traversed edges that are not connected to the connected graph, obtain the nodes whose shortest distance to the endpoints is no greater than depth D and are not located in the connected graph. Obtain the carbon emission-related complexity of the topology network composed of these nodes, and then obtain the carbon emission-related complexity of the traversed edge-related adjacency topology network.

[0116] The carbon emission-related complexity of the topological network is related to the degree of all nodes in the network and the weighted key feature coefficients. Specifically, the product of the degree of all nodes in the image network and the weighted key feature coefficients is obtained, and the geometric mean of the product values ​​of all nodes is obtained to obtain the carbon emission-related complexity of the topological network.

[0117] Select the edge with the highest carbon emission-related complexity in the adjacency topology network, connect this edge to the connection graph, update the connection graph, and update the depth parameter D. The way to update the depth parameter D is: D = D - 1.

[0118] In the power system topology network information, the number of edges and nodes are related to the depth parameter D. In this embodiment, the relationship between the depth parameter D and the number of edges and nodes in the power system topology network information is as follows.

[0119] Its expression is: Where Sum(E) is the number of edges in the topology of the power system; Sum(E) is the number of nodes in the topology of the power system; and D0 is an adjustment parameter. For example, the floor function, D_avg(V) is the average of the shortest distances between nodes in the topology network information of the power system.

[0120] Repeat the above process until the value of the depth parameter D is 0, then end the connection graph construction process and obtain the connection graph.

[0121] By repeating the above connection graph construction process several times, several different connection graphs are obtained. Based on the obtained different connection graphs, a set of edges for each of the different connection graphs is obtained. For the i-th connection graph, the set of edges is denoted as E. sub,i Find the intersection of the sets of edges from several distinct connected graphs, denoted as E. sub .

[0122] Find the intersection of the sets of edges from several distinct connected graphs, denoted as E. sub .

[0123] Its expression is: E sub =E sub,1 ∩E sub,2 ∩.....∩E sub,K , where K is the number of connected graphs.

[0124] After obtaining the intersection of the sets of edges of several connected graphs, a node-related graph is constructed based on the intersection, and it is determined whether the node-related graph is connected. If the node-related graph is connected, then the node-related graph is a connected graph, and all nodes in the connected graph are key nodes for carbon emissions; if the node-related graph is not connected, then a subgraph of the node-related graph is obtained, and a subgraph connection matrix is ​​constructed. In the subgraph connection matrix A, if the i-th subgraph is not connected to the j-th subgraph, then the element in the i-th row and j-th column of the subgraph is 0, represented as: A i,j =0; If the i-th subgraph is directly connected to the j-th subgraph, then the element in the i-th row and j-th column of the subgraph is 1, represented as: A i,j =1.

[0125] It should be noted that for elements with the same row and column, the value of that element is 1, represented as: A i,j =1, i=j.

[0126] It should be noted that when constructing the subgraph connection matrix, the carbon emission-related complexity of the corresponding topology network of the subgraph is obtained and arranged according to the carbon emission-related complexity of the subgraph.

[0127] Traverse the elements in the subgraph connection matrix. If the value of an element in the subgraph connection matrix is ​​0, obtain the row and column of the element, find the shortest path between the subgraph node corresponding to the row and the subgraph node corresponding to the column, and connect the two subgraphs through the shortest path. After connecting the two subgraphs, update the elements in the subgraph connection matrix. If the element in the subgraph connection matrix is ​​1, the subgraph connection process is not performed.

[0128] After traversing the graph, the first connection graph is obtained, and the shortest path is removed from the first connection graph. The specific process is as follows.

[0129] Obtain the shortest distance between all different subgraph nodes in the subgraph connection process of the first connection graph, obtain the geometric mean of the weighted feature key coefficients of all nodes that do not belong to the node-related graph in the shortest path between all different subgraph nodes, and traverse the shortest path according to the magnitude of the geometric mean.

[0130] It should be noted that during the traversal of the shortest path, the shortest path between different subgraph nodes with the smallest geometric mean is traversed first.

[0131] In the shortest path traversal process, if the node-related graph becomes disconnected after removing the traversed shortest path, the shortest path is not removed; if the node-related graph becomes connected after removing the traversed shortest path, the shortest path is removed, and the first connectivity graph is updated. The shortest path traversal process continues until all shortest paths have been removed, at which point the shortest path traversal process ends, a second connectivity graph is obtained, and this second connectivity graph is used as the final connectivity graph. All nodes in the connectivity graph are key carbon emission nodes.

[0132] To further illustrate the connectivity process of the node-related graph, an example will be used below to explain the process in detail.

[0133] See Figure 4 ,exist Figure 4 In the above, there exists a sub-node-related graph. Figure 1 ,son Figure 2 ,son Figure 3 and son Figure 4 ,son Figure 1 ,son Figure 2 ,son Figure 3 and son Figure 4 They are not interconnected, and the sub-units are not connected. Figure 1 The carbon emission-related complexity is highest for the corresponding topological network; sub- Figure 2 The carbon emission-related complexity of the corresponding topological network is second only to that of the sub-network. Figure 3 The carbon emission-related complexity of the corresponding topological network is greater than that of the sub-network. Figure 4 The carbon emission-related complexity of the corresponding topology network, and the sub- Figure 3 The carbon emission-related complexity of the corresponding topology network is less than that of the sub-network. Figure 2 The carbon emission-related complexity of the corresponding topology network.

[0134] Construct a 4x4 subgraph connection matrix, where the first to fourth rows correspond to the subgraphs. Figure 1 To the Son Figure 4 The first to fourth columns correspond to the sub-columns respectively. Figure 1 To the Son Figure 4 Set the values ​​of the first row and first column to the fourth row and fourth column of the subgraph connection matrix to 1, and set the rest to 0.

[0135] Iterate through the elements of the subgraph connection matrix in the first row and first column. If the value of this element is 1, the subgraph connection process is not performed.

[0136] Iterate through the elements of the subgraph connection matrix in the first row and second column. If the element's value is 0, then retrieve the subgraph corresponding to the first row. Figure 1 The sub-line corresponding to the second row Figure 2 , get child Figure 1 With you Figure 2 The shortest path between nodes, through which the child node's shortest path is made possible. Figure 1 With you Figure 2 Connect the subgraphs and update the subgraph connection matrix.

[0137] Iterate through the elements of the subgraph connection matrix in the first row and third column. If the element's value is 0, then retrieve the subgraph corresponding to the first row. Figure 1 The sub-line corresponding to the third line Figure 3 , get child Figure 1 With you Figure 3 The shortest path between nodes, through which the child node's shortest path is made possible. Figure 1 With you Figure 3 Connect the subgraphs and update the subgraph connection matrix.

[0138] Iterate through the elements of the subgraph connection matrix in the first row and fourth column. If the value of this element is 0, then retrieve the subgraph corresponding to the first row. Figure 1 The sub-line corresponding to the third line Figure 4 , get child Figure 1 With you Figure 4 The shortest path between nodes, through which the child node's shortest path is made possible. Figure 1 With you Figure 4 Connect the subgraphs and update the subgraph connection matrix.

[0139] Following a similar process, after traversing the subgraph connectivity matrix, the subgraph... Figure 2 With you Figure 3 Connect nodes via the shortest path; sub Figure 2 With you Figure 4 Connect nodes via the shortest path; sub Figure 3 With you Figure 4 Connect nodes using the shortest path between them.

[0140] Note Figure 1 With you Figure 2 The shortest path between them is path 1; [Note: The last part is a typo and can be left as is.] Figure 1 With you Figure 3 The shortest path between them is path 2; note Figure 1 With you Figure 4 The shortest path between them is path 3; note Figure 2 With you Figure 3 The shortest path between them is path 4; note Figure 2 With you Figure 4 The shortest path between them is path 5; note Figure 3 With you Figure 4 The shortest path between them is path 6.

[0141] For the specific connection relationships between subgraphs after traversing the subgraph connection matrix, please refer to [link / reference]. Figure 5 ,exist Figure 5 neutron Figure 1 With you Figure 2 Connection, sub Figure 1 With you Figure 3 Connection, sub Figure 1 With you Figure 4 Connection, sub Figure 2 With you Figure 3 Connection, sub Figure 2 With you Figure 4 Connection, sub Figure 3 With you Figure 4 connect.

[0142] If the geometric mean values ​​of the shortest paths are in the following order: Path 2 > Path 3 > Path 1 > Path 6 > Path 5 > Path 4.

[0143] Then start traversing from path 4, from Figure 4 As can be seen from the topology, after removing path 4, the first connection graph becomes connected. Therefore, remove path 4 and update the first connection graph.

[0144] Traverse path 5, by Figure 4 As can be seen from the topology, after removing path 5, the first connection graph becomes connected. Therefore, remove path 5 and update the first connection graph.

[0145] Traverse path 6, by Figure 4 As can be seen from the topology, after removing path 6, the first connection graph becomes connected. Therefore, remove path 6 and update the first connection graph.

[0146] Traverse path 1, by Figure 4 As can be seen from the topology, after removing path 1, the first connection graph is not connected, so path 1 is not removed.

[0147] Traverse path 3, by Figure 4 As can be seen from the topology, after removing path 3, the first connection graph is not connected, so path 3 is not removed.

[0148] Traverse path 2, by Figure 4 From the topology, we can see that after removing path 2, the first connection graph is not connected, so path 2 is not removed.

[0149] See Figure 6 , Figure 6 To obtain the connection relationships of each subgraph in the second connection graph, the connection relationships of each subgraph are as follows: Figure 1 With you Figure 2 Connection, sub Figure 1 With you Figure 3 Connection, sub Figure 1 With you Figure 4 connect.

[0150] The connection graph obtained by the method in this embodiment covers key node clusters in the power system whose carbon emission contribution is significantly higher than that of other subnets due to its topological connectivity. Furthermore, the dynamic change range of its internal carbon flow rate covers a high proportion of the main fluctuation range of the entire network. In complex power system networks, the efficiency of the obtained connection graph is high. It can obtain the connection graph in the topological network information of the power system within a reasonable or shorter time range, thereby identifying key carbon emission nodes.

[0151] Based on this characteristic, targeted monitoring and control of this connection diagram can provide the following advantages in practical applications.

[0152] By replacing full network traversal calculations with local topology analysis, active carbon flow regions can be quickly located, reducing the computational cost of real-time carbon tracing.

[0153] By leveraging the cluster of key carbon emission nodes, the carbon intensity of the entire grid can be nonlinearly affected by adjusting the power generation plans or power flow distribution of a few hubs, thereby improving the marginal benefits and response speed of carbon reduction strategies.

[0154] Example 2: Based on Example 1, this example discloses a specific implementation process of the method involved in the present invention.

[0155] In this embodiment, the specific implementation process of the method is as follows.

[0156] S1. Obtain the topology network information of the power system, obtain several features of all nodes in the topology network information, divide the features into several feature groups, and ensure that the correlation between features in the feature groups is not greater than the correlation threshold.

[0157] It should be noted that the types of features involved in this embodiment are the same as those involved in Embodiment 1; the correlation involved in this embodiment is the same as that involved in the embodiment.

[0158] However, in the application scenario of this embodiment, the correlation between each feature is not greater than the correlation threshold. Therefore, each feature needs to be divided into two groups: one group is the topology-related feature group, and the other group is the topology-irrelevant feature group.

[0159] In this embodiment, the topology-related feature set includes: the carbon flow rate correlation of nodes, the standard deviation correlation balance of nodes, the improved clustering coefficient of nodes, and the Shell contribution of nodes.

[0160] In this embodiment, the topology-irrelevant feature set includes: the carbon flow intensity of a node, the carbon inflow of a node, the standard deviation of the carbon flow rate of a node, the load carbon flow rate of a node, the unit injection carbon flow rate of a node, and the difference between the maximum and minimum carbon flow rates of a node.

[0161] If there are N nodes in a power system, then construct a 4-row N-column feature matrix and a 6-row N-column feature matrix for the topology-dependent feature group and the topology-independent feature group, respectively, denoted as the topology-dependent feature matrix and the topology-independent feature matrix.

[0162] KPCA processing is performed on the topological feature correlation matrix and the topological uncorrelated feature matrix to obtain the eigenvalues ​​corresponding to the kernel function skew variance matrix. Several eigenvalues ​​are obtained such that the sum of the obtained eigenvalues ​​accounts for at least 90% of the sum of all eigenvalues, and KPCA features corresponding to the several eigenvalues ​​are obtained.

[0163] When performing KPCA processing, the kernel function expression is: Where person(x,y) is the Pearson coefficient between feature x and feature y; σ is the adjustment coefficient.

[0164] After determining the KPCA features of the topology-related feature group and the topology-unrelated feature group, the topology-related KPCA feature matrix and the topology-unrelated KPCA feature matrix are constructed based on the KPCA features of the topology-related feature group and the topology-unrelated feature group.

[0165] For a topology-dependent feature matrix, if the number of KPCA features is k1, then construct a topology-dependent KPCA feature matrix with k1 rows and N columns.

[0166] For a topologically unrelated feature matrix, if the number of KPCA features is k2, then construct a topologically unrelated KPCA feature matrix with k2 rows and N columns.

[0167] S2. Obtain the key feature coefficients of the nodes in the power system from the plurality of feature groups; Obtain the key feature coefficients of the node in the topology-related KPCA feature matrix and the topology-unrelated KPCA feature matrix, thereby obtaining the key feature coefficients of the node in the topology-related feature group and the topology-unrelated feature group.

[0168] It should be noted that the process for obtaining the key feature coefficients of nodes in the topology-related KPCA feature matrix and the topology-unrelated KPCA feature matrix is ​​the same as the process for obtaining the key feature coefficients of nodes in the feature matrix in Example 1, and will not be repeated here.

[0169] After obtaining the key feature coefficients of nodes in the topology-related feature group and the topology-irrelevant feature group, determine the weight values ​​corresponding to the topology-related feature group and the topology-irrelevant feature group.

[0170] The weight values ​​corresponding to the topology-related feature groups are positively correlated with the network complexity in the topology network information of the power system, while the weight values ​​corresponding to the topology-unrelated feature groups are negatively correlated with the network complexity in the topology network information of the power system.

[0171] The expression for the weights corresponding to the topology-related feature groups is: w Tr =1-e -Nc .

[0172] The weights corresponding to the topologically irrelevant feature groups are expressed as: w Tn =e -Nc .

[0173] Where Nc represents the network complexity in the topology information of the power system, and its expression is: The variables Sum(E) and Sum(V) have been described in Example 1 and will not be repeated here.

[0174] In this embodiment, the weighted feature key coefficient of a node is obtained by multiplying the feature key coefficient of the node in the topology-related feature group by the weight value corresponding to the topology-related feature group, and adding the feature key coefficient of the node in the topology-unrelated feature group by multiplying the weight value corresponding to the topology-unrelated feature group.

[0175] For the i-th node in the topology network information of the power system, the expression for the weighted key characteristic coefficient of the i-th node is as follows.

[0176] Its expression is: WCCc i =w Tr *CCc i,Tr +w Tn *CCc i,Tn Among them, CCc i,Tr CCc represents the key feature coefficient of the i-th node in the topologically relevant feature group. i,Tn is the key feature coefficient of the i-th node in the topologically irrelevant feature group.

[0177] S3. Based on several key feature coefficients of the node, obtain the weighted key feature coefficients of the node, construct a connection graph, and determine the key carbon emission nodes based on the connection graph.

[0178] In this embodiment, the execution process of step S3 is the same as that of step S3 in embodiment 1, and will not be described again here.

[0179] Example 3: Based on Example 1, this example discloses a specific implementation process of the method involved in the present invention.

[0180] The specific implementation process of the method involved in this embodiment is as follows.

[0181] S1. Obtain the topology network information of the power system, obtain several features of all nodes in the topology network information, divide the features into several feature groups, and ensure that the correlation between features in the feature groups is not greater than the correlation threshold.

[0182] It should be noted that in this embodiment, the types of node features include those of the nodes involved in Embodiment 1 and Embodiment 2, and the features of the nodes will not be described in detail here.

[0183] It should be noted that, in this embodiment, the correlation between features is the same as the correlation between features involved in Embodiments 1 and 2.

[0184] Using the correlation between features as a distance metric, the feature set is randomly divided into two feature sets, and the cluster center points of the two feature sets are obtained. The total distance between the cluster center point and all feature points in the two feature sets is minimized.

[0185] After determining the cluster centers of the two feature sets, the feature point with the smallest distance from the cluster centers of the two feature sets is obtained, and the two feature sets and the cluster centers of the two feature sets are updated. The process is repeated until the cluster centers tend to stabilize. Specifically, in the two feature sets, the Eulerian distance between the cluster centers before and after the update is less than the convergence threshold.

[0186] After the cluster centers of the two feature sets have stabilized, the updated two feature sets are obtained. If there are no two features in the two feature sets with a correlation less than the correlation threshold, then the two feature sets are respectively regarded as two feature groups of a node. If there are two features in the two feature sets with a correlation greater than the correlation threshold, then K-means clustering is performed on the feature sets with a correlation greater than the correlation threshold, and the K-means clustering process is repeated until there are no two features in the feature sets with a correlation less than the correlation threshold.

[0187] The above process ensures that the correlation between features in several feature groups does not exceed the correlation threshold.

[0188] S2. Obtain the key feature coefficients of the nodes in the power system from among several feature groups.

[0189] In this embodiment, the execution process of step S2 is the same as that of step S2 in embodiment 1, and will not be described again here.

[0190] S3. Based on several key feature coefficients of the node, obtain the weighted key feature coefficients of the node, construct a connection graph, and determine the key carbon emission nodes based on the connection graph.

[0191] In this embodiment, the execution process of step S3 is the same as that of step S3 in embodiment 1, and will not be described again here.

[0192] This invention has illustrated its purpose, technical solutions, and beneficial effects through specific embodiments. However, these embodiments are merely examples to demonstrate the application of the invention and do not constitute a limitation on the scope of protection of this invention. We explicitly state that any reasonable modifications, equivalent substitutions, or technical improvements guided by the spirit and principles of this invention should be included within the scope of protection of this invention. This means that as long as these changes do not deviate from the core idea and basic function of the invention, they should be protected by patent rights. The scope of protection of this invention should be broad, including all directly obvious variations as well as non-obvious innovations that a technical expert can reasonably deduce from the disclosure of this invention. This broad protection aims to promote further research and development based on this invention, while ensuring that its innovativeness and practicality are fully protected by law.

Claims

1. A method for searching critical nodes of carbon emissions in a power system, characterized in that, The method steps include: S1. Determine several features of all nodes in the power system, divide the several features into several feature groups, and in the several feature groups, the correlation between the features is not greater than a correlation threshold. S2. Obtain the key feature coefficients of the nodes in the power system from the plurality of feature groups; S3. Based on several key feature coefficients of the node, obtain the weighted key feature coefficients of the node, construct a connection graph, and determine the key carbon emission nodes based on the connection graph.

2. The method for searching key carbon emission nodes in a power system according to claim 1, characterized in that, Step S1 includes: A feature map is constructed using the correlation between the features as a distance metric. The feature map is segmented to obtain feature sub-maps; If, in the feature subgraph, there exists at least two features whose correlation is greater than the correlation threshold, then the feature subgraph is segmented.

3. The method for searching key carbon emission nodes in a power system according to claim 1, characterized in that, Step S1 includes: Using the correlation between the features as a distance metric, K-means clustering is performed to obtain several clusters; If, within the cluster, there exist at least two features whose correlation is greater than the correlation threshold, then K-means clustering is performed on the cluster. In the K-means clustering, the number of clusters is at least 2.

4. A method for searching key carbon emission nodes in a power system according to claim 2 or 3, characterized in that, Step S3 includes: For a given feature group, the weight value corresponding to the feature group is related to the number of feature types in the feature group; The sum of the weight values ​​corresponding to the feature group is 1.

5. The method for searching key carbon emission nodes in a power system according to claim 1, characterized in that, Step S1 includes: The features are divided into two feature groups: a topology-related feature group and a topology-unrelated feature group.

6. The method for searching key carbon emission nodes in a power system according to claim 5, characterized in that, Step S3 includes: The weight values ​​corresponding to the topology-related feature groups are related to the network complexity of the power system; The sum of the weights of the topology-related feature group and the topology-unrelated feature group is 1.

7. A method for searching critical carbon emission nodes in a power system according to claim 1, 2, 3, 5, or 6, characterized in that, The correlation between the features includes: The weighted sum of the linear correlations and nonlinear correlations between the features.

8. A method for searching key carbon emission nodes in a power system according to claim 5 or 6, characterized in that, In step S3, the process of obtaining the key feature coefficients of the node includes: Based on the topology-related feature group and the topology-irrelevant feature group, construct the topology-related feature matrix and the topology-irrelevant feature matrix; KPCA processing is performed on the topology-related feature matrix and the topology-unrelated feature matrix to obtain the processed feature matrix; Based on the processed feature matrix, the key feature coefficients of the node are obtained.

9. The method for searching key carbon emission nodes in a power system according to claim 4, characterized in that, In step S3, the process of obtaining the key feature coefficients of the node includes: In the set of features, determine the coefficient of variation for each of the features, root The corresponding weight value for each feature is determined based on the coefficient of variation, and the feature of the data corresponding to each node is processed according to the weight value; Obtain the probability function value of the processed data, thereby obtaining the key feature coefficients of the node.

10. A method for searching critical carbon emission nodes in a power system according to claim 1, 2, 3, 5, 6, or 9, characterized in that, In step S1, the process of constructing the connection graph includes: In the topology graph of the power system, the probability of selecting an edge is positively correlated with the magnitude of the weighted feature key coefficients of the nodes at both ends of the edge.

Images