A power market multi-agent power transmission congestion early warning method and system

Abstract

The application discloses a kind of power market multi-agent power transmission congestion early warning method and system, it is related to electric power system automation technical field, the method includes: obtaining power grid topology structure, operating parameter and market agent transaction data;According to the first graph structure of topology structure construction;According to the injection power variation of the transaction data associated with transaction behavior grid node, according to injection power variation determines the influence coefficient of power flow to transmission line, and then constructs second graph structure;The adjacency matrix of first graph structure and second graph structure is fused, and the fusion adjacency matrix is obtained;The fusion adjacency matrix is input into graph convolution network to extract node features with operating parameter, determines line congestion risk probability in combination with line end node features, and outputs early warning signal.The application realizes the cross-domain mapping of power grid physical constraint and market transaction behavior and dynamic heterogeneous graph fusion, effectively improves the prediction accuracy and interpretability of power transmission congestion risk under the background of multi-agent transaction behavior.

Description

Technical Field

[0001] This invention relates to the field of power system automation technology, and more specifically, to a method and system for early warning of transmission congestion among multiple entities in the power market. Background Technology

[0002] Transmission congestion is a significant issue affecting the safe and stable operation of power systems. In an electricity market environment, changes in the trading behavior of multiple market participants, including power generation companies, electricity retailers, and load-side users, can alter the power distribution at grid nodes, thereby affecting the power flow status of transmission lines. Therefore, accurately predicting the transmission congestion risks caused by changes in market participant behavior is of great importance.

[0003] Existing transmission congestion early warning methods mainly include physical analysis methods based on power flow calculations and prediction methods based on graph convolutional networks. Among them, the graph convolutional network-based methods typically abstract the physical topology of the power grid into a graph model, use the electrical connections between nodes as the basis for information propagation, and achieve transmission congestion risk prediction through neighborhood feature aggregation.

[0004] However, the physical topology of the power grid does not correlate with the propagation of the impact of market participant behavior. The physical topology reflects the electrical connections between nodes, while the impact of changes in market participant trading behavior on transmission congestion typically affects the entire network through power transmission characteristics. There may not be a direct electrical connection between the market participant influencing the congested transmission line. Therefore, existing methods based on physical topology-based graph models struggle to accurately characterize the propagation of the impact between market participant behavior and transmission lines, resulting in an inability to effectively capture cross-regional behavioral impacts and a difficulty in quantifying the degree of influence of different market participant behaviors on transmission congestion.

[0005] Therefore, how to establish a transmission congestion early warning method that can reflect the characteristics of the influence and spread of market entities' behavior is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and system for early warning of transmission congestion in the power market involving multiple stakeholders. This addresses the technical problem that existing transmission congestion early warning methods based on graph convolutional networks use the physical topology of the power grid as the basis for information propagation between nodes, failing to consider the impact of market participants' trading behavior on the propagation of power flow changes in transmission lines, thus making it difficult to accurately identify transmission congestion risks caused by multi-stakeholder trading behavior.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A multi-entity transmission congestion early warning method for the power market includes the following steps: acquiring the power grid topology, operating parameters, and market participant transaction data; constructing a first graph structure representing the electrical connection relationships between nodes based on the topology; determining the injected power change of power grid nodes associated with the market participant transaction behavior based on the transaction data, determining the impact coefficient on transmission line power flow based on the injected power change, and constructing a second graph structure based on the impact coefficient; fusing the adjacency matrices of the first and second graph structures to obtain a fused adjacency matrix; extracting node features from the fused adjacency matrix and the operating parameter input graph convolutional network, determining the line congestion risk probability by combining the node features at both ends of the line, and outputting an early warning signal.

[0008] In a preferred embodiment, determining the influence coefficient on the power flow of the transmission line based on the change in injected power includes: determining the influence coefficient of the change in injected power on the power flow of each transmission line based on the power transfer distribution factor.

[0009] In a preferred embodiment, the step of fusing the adjacency matrices of the first graph structure and the second graph structure to obtain a fused adjacency matrix includes: extracting the weight distribution features of the adjacency matrices of the first graph structure and the second graph structure respectively; normalizing the adjacency matrices of the first graph structure and the second graph structure based on the weight distribution features to obtain normalized first and second adjacency matrices; and weighting and fusing the normalized first and second adjacency matrices according to preset fusion weights to obtain the fused adjacency matrix.

[0010] In a preferred embodiment, constructing the second graph structure based on the influence coefficients includes: dividing the influence coefficients into a set of positive influence coefficients and a set of negative influence coefficients according to the signs of the influence coefficients; constructing a positive influence subgraph based on the set of positive influence coefficients and a negative influence subgraph based on the set of negative influence coefficients, wherein the positive influence subgraphs and negative influence subgraphs are used to characterize the increasing and decreasing influence relationships of market participants' trading behavior on power transmission line flow, respectively; and combining the positive influence subgraphs and negative influence subgraphs to form the second graph structure.

[0011] In a preferred embodiment, constructing the second graph structure based on the influence coefficients includes: obtaining the influence coefficient amplitude distribution corresponding to the positive influence coefficient set and the negative influence coefficient set; determining the influence relationship screening conditions based on the influence coefficient amplitude distribution; screening the edges in the positive influence subgraph and the negative influence subgraph according to the influence relationship screening conditions, retaining the edges that meet the influence relationship screening conditions, and constructing the second graph structure based on the screened positive influence subgraph and negative influence subgraph.

[0012] In a preferred embodiment, the step of weighted fusion of the normalized first adjacency matrix and the second adjacency matrix according to preset fusion weights includes: determining the degree of change in market entity trading behavior based on the change in injection power corresponding to the market entity trading data within the current time period; adjusting the fusion weights corresponding to the first adjacency matrix and the second adjacency matrix according to the degree of change in market entity trading behavior; and using the adjusted fusion weights to weightedly fuse the normalized first adjacency matrix and the second adjacency matrix to obtain a fused adjacency matrix.

[0013] In a preferred embodiment, determining the probability of line congestion risk by combining the characteristics of nodes at both ends of the line and outputting an early warning signal includes: acquiring the node characteristics corresponding to the nodes at both ends of the transmission line to be predicted; constructing line characteristics based on the differences and correlations between the node characteristics of the nodes at both ends of the line; inputting the line characteristics into a risk prediction model to obtain the probability of line congestion risk; determining the congestion early warning result based on the probability of congestion risk and preset risk conditions, and outputting an early warning signal.

[0014] In a preferred embodiment, the method further includes: obtaining an impact coefficient corresponding to the target transmission line that generates the congestion warning result; determining the degree of influence of the trading behavior of each market participant on the congestion risk of the target transmission line based on the amplitude and direction of the impact coefficient; and sorting the market participants according to the degree of influence to generate a list of entities affected by congestion.

[0015] In a preferred embodiment, determining the injected power change of the grid node associated with the trading behavior of market participants based on the transaction data includes: determining the transaction power change of each market participant based on the transaction data; mapping the transaction power change to the corresponding grid node based on the grid access node corresponding to each market participant; and determining the injected power change of the grid node based on the transaction power change of each grid node.

[0016] A multi-entity transmission congestion early warning system for the power market includes: a data acquisition module for acquiring power grid topology, operating parameters, and market participant transaction data; a first graph construction module for constructing a first graph structure representing the electrical connection relationships between nodes based on the topology; a second graph construction module for determining the injected power change of power grid nodes associated with market participant transaction behavior based on the transaction data, determining the impact coefficient on transmission line power flow based on the injected power change, and constructing a second graph structure based on the impact coefficient; a graph fusion module for fusing the adjacency matrices of the first and second graph structures to obtain a fused adjacency matrix; and a risk early warning module for extracting node features from the fused adjacency matrix and the operating parameter input graph convolutional network, determining the probability of line congestion risk by combining the features of nodes at both ends of the line, and outputting an early warning signal.

[0017] This invention discloses a multi-entity transmission congestion early warning system for the power market, along with its technical effects and advantages: 1. This invention constructs a first graph structure representing the electrical connection relationship of power grid nodes and a second graph structure representing the influence relationship of market participants' transaction behavior, and merges the two graph structures. This allows the information propagation relationship of the graph convolutional network to simultaneously include power grid physical topology information and market behavior influence information, avoiding the problem in existing methods that cannot accurately describe the influence path of market participants' behavior due to feature propagation based solely on physical topology relationships.

[0018] 2. This invention determines the influence coefficient based on the relationship between the change in node injected power and the change in power flow of transmission lines, and uses the influence coefficient to construct a second graph structure, thereby realizing a quantitative representation of the influence relationship between the trading behavior of market participants and the power flow of transmission lines, and improving the ability to identify the congestion risk caused by market behavior.

[0019] 3. The adjacency matrices of the first and second graph structures are fused to enable feature propagation of different types of relationships in a unified graph structure, thereby improving the adaptability of transmission congestion risk prediction in complex power market environments. Attached Figure Description

[0020] Figure 1 A flowchart illustrating the multi-entity transmission congestion early warning method for the power market provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the structure of a multi-entity transmission congestion early warning system for the power market provided in an embodiment of the present invention; Figure 3 The second diagram provided in this embodiment of the invention is a schematic diagram of the structure construction process; Figure 4 The topology diagram of the IEEE 39-node standard test system provided in this embodiment of the invention; Figure 5 This is a comparison chart showing the accuracy of blocking risk identification between the embodiments of the present invention and the comparative methods. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0023] Example 1, Figure 1 This is a flowchart illustrating a multi-entity transmission congestion early warning method for the power market provided in an embodiment of the present invention. Figure 1 As shown, the method mainly includes the following steps: S1, acquires power grid topology, operating parameters, and market participant transaction data; In this embodiment, the power grid topology data includes power grid node information, transmission line connection relationships, and information on nodes at both ends of the lines, which are used to characterize the connection relationships between various power grid nodes and transmission lines in the power system. The power grid nodes are bus nodes, and the transmission lines are electrical channels connecting different bus nodes.

[0024] The operating parameter data is used to characterize the current operating status of the power grid, including node active power, node reactive power, voltage amplitude, voltage phase angle, and transmission line power flow data, providing input features for subsequent node feature extraction of graph convolutional networks and transmission congestion risk prediction.

[0025] The market participant transaction data is used to characterize changes in the transaction behavior of each market participant during the electricity market transaction process. It includes market participant identification information, grid access node information corresponding to the market participant, and transaction volume change information. The market participants include generation-side market participants, sales-side market participants, and load-side market participants.

[0026] In one implementation, the aforementioned data is acquired at preset time intervals, and data associations are established for the power grid topology data, operating parameter data, and market participant transaction data acquired within the same time segment. Furthermore, the acquired data undergoes preprocessing, including data time synchronization, format conversion, and abnormal data removal, to ensure data consistency among the various data sets.

[0027] S2, Construct a first graph structure representing the electrical connection relationship between nodes based on the topology; In this embodiment, the first graph structure is constructed using power grid buses as nodes and transmission lines connecting different buses as edges, and is represented as follows: (1) in, For a set of nodes, The total number of power grid nodes. Let the set of edges be defined as follows: when node With nodes When there is a power transmission line connection, then , This is the adjacency matrix corresponding to the structure of the first graph, used to represent the electrical connection relationships between nodes.

[0028] The adjacency matrix Defined as: (2) And satisfy This means that the node has no connection to itself.

[0029] The first graph structure constructed in the above manner is used to describe the physical topology of the power grid itself. The adjacency matrix of the first graph structure... It is a sparse matrix, and its non-zero elements correspond to node pairs that have direct transmission line connections.

[0030] Furthermore, the node set Feature matrix of associated nodes:

[0031] in, This represents the node feature dimension. The node features are composed of the operating parameters obtained in step S1, including but not limited to node active power, node reactive power, voltage amplitude, and voltage phase angle.

[0032] The node feature matrix Used as input to subsequent graph convolutional networks to extract correlation features between power grid nodes.

[0033] It should be noted that the node set in the first graph structure The node set is kept consistent with the node set in the subsequently constructed second graph structure so that the first and second graph structures can be merged in the same node space using adjacency matrices.

[0034] S3, determine the injected power change of the grid node associated with the market entity's trading behavior based on the transaction data, determine the impact coefficient on the power flow of the transmission line based on the injected power change, and construct the second graph structure based on the impact coefficient; In this embodiment, the second graph structure shares the same set of nodes as the first graph structure. Its adjacency matrix It is used to characterize the relationships between different power grid nodes under the influence of market participants' trading behavior. Figure 3 The second diagram provided in this embodiment of the invention illustrates the structural construction process. Figure 3 As shown, the construction of the second graph structure includes the following sub-steps: S31, Based on the market participant transaction data obtained in step S1, determine the change in transaction volume for each market participant. ( , (This refers to the total number of market participants). The change in trading volume represents the change in trading volume among market participants within a specific time period.

[0035] Based on the grid access node information corresponding to each market participant, the change in the traded electricity volume is mapped to the corresponding grid node. For the grid node... The amount of its injected power change Determined in the following ways: (3) in, Indicates the connection to the power grid node The set of market participants. If a power grid node is not connected to any market participant, then... .

[0036] By using the above methods, changes in the trading behavior of market participants are transformed into changes in the injected power of grid nodes, enabling trading data to be represented at the physical level of the power grid.

[0037] S32, based on the power transfer distribution factor, determine the influence coefficient of the power injection changes at each grid node on the power flow of the transmission line. The power transfer distribution factor describes the linear mapping relationship between the power injection changes at grid nodes and the power flow changes of the transmission line.

[0038] In this embodiment, the power transfer distribution factor is calculated based on a DC power flow model. Assume the power grid system includes... Each node and For a transmission line, the nodal susceptance matrix is ​​represented as follows: The transmission line-node correlation matrix is ​​represented as follows: Then the power transfer distribution factor matrix Represented as: (4) in, This represents the pseudo-inverse of the nodal susceptance matrix.

[0039] Elements in the power transfer distribution factor matrix Represents a power grid node The effect of injection unit power change on the first The degree of impact of power flow changes on a transmission line, expressed in per-unit values. Grid node. The effect of injection power changes on transmission lines The influence coefficient of trends .

[0040] Furthermore, based on the influence coefficients of each node, the actual impact of the change in injected power at each node on the power flow of the transmission line is calculated. For the transmission line... Its tidal current change It is determined by the changes in injected power at all nodes: (5) Through the above methods, a mapping relationship was established between changes in market participants' trading behavior, changes in power injection at grid nodes, and changes in power flow on transmission lines.

[0041] S33, using the node set of the first graph structure As the node set of the second graph structure, based on the influence coefficient obtained in step S32 Constructing the adjacency matrix of the second graph structure .

[0042] For each power grid node Based on its influence coefficient on all transmission lines, construct the influence vector corresponding to this node: (6) in, Representation Nodes The overall impact pattern of changes in injected power on the power flow of each transmission line. This represents the total number of transmission lines.

[0043] Furthermore, the association relationships between nodes are determined based on the similarity of their influence patterns. For any two nodes... and Calculate the degree of correlation between the corresponding influence vectors: (7) in, This describes the method for calculating the degree of correlation between influence vectors.

[0044] In this embodiment, the degree of association is calculated using cosine similarity: (8) When the changes in injected power at two nodes have similar impact patterns on the power flow of the transmission line, the directions of their impact vectors tend to be consistent. In this case, the two nodes have a high correlation strength in the second graph structure.

[0045] In another implementation, the degree of correlation can also be determined using the Pearson correlation coefficient, which reflects the similarity of influence patterns between nodes by calculating the linear correlation between two influence vectors. Those skilled in the art will understand that other methods capable of characterizing the degree of correlation between vectors can also be used in practical applications, including but not limited to Euclidean distance and mutual information.

[0046] Based on the degree of correlation Determine the adjacency matrix of the second graph structure : (9) That is, the adjacency matrix of the second graph structure elements in Represents a node and The degree of correlation between them under the influence of the transaction behavior of market entities.

[0047] S34. Because the power transfer distribution factor is directional, its sign reflects the different directions of influence of market participants' trading behavior on the power flow of transmission lines. The direction of power flow in transmission lines is determined according to a preset reference direction. Specifically, a positive influence coefficient indicates that an increase in injected power at that node will lead to an increase in the power flow of the corresponding transmission line along the reference direction, i.e., it has an effect of exacerbating congestion; a negative influence coefficient indicates that an increase in injected power at that node will lead to a decrease in the power flow of the corresponding transmission line along the reference direction, i.e., it has an effect of alleviating congestion.

[0048] Therefore, this embodiment constructs positive and negative influence subgraphs based on the sign of the influence coefficient to preserve the directional information of the impact of market participants' trading behavior on transmission congestion.

[0049] Specifically, for power grid nodes Corresponding influence vector Based on the sign of the influence coefficient, it is split into a positive influence vector. and negative influence vector : (10) (11) in, Represents a node For transmission lines The positive influence coefficient, Represents a node For transmission lines The absolute value of the negative impact coefficient.

[0050] Based on the positive influence vector, the degree of positive influence correlation between nodes is calculated in the same way as in S33, and a positive influence subgraph is constructed. Based on the negative impact vector, the degree of negative impact correlation between nodes is calculated in the same way as in S33, and a negative impact subgraph is constructed. The positive impact subplot is used to characterize the increasing influence of market participants' trading behavior on the transmission line power flow along the reference direction, while the negative impact subplot is used to characterize the decreasing influence of market participants' trading behavior on the transmission line power flow along the reference direction.

[0051] The second diagram structure consists of a positive influence subgraph and a negative influence subgraph, represented as follows: (12) in, This represents the set of edges that positively influence the subgraph. This represents the set of edges in the negative influence subgraph. The positive and negative influence subgraphs share the same set of nodes. However, the edge weights are determined by the degree of positive influence and the degree of negative influence, respectively.

[0052] S35. Since the influence relationships in the second graph structure are calculated based on influence vectors, the degree of influence correlation between different nodes varies. If all influence relationships are included in the second graph structure, there will be a large number of low-contribution correlation edges in the graph structure, which will introduce non-critical correlation information during the node feature propagation process and reduce the model's ability to extract transmission congestion-related features.

[0053] Therefore, this embodiment filters out weak influence relationships based on the amplitude distribution of the influence coefficient.

[0054] Specifically, obtain the adjacency matrix of the positively influenced subgraph. Adjacency matrix of negative influence subgraph The amplitude distribution of the elements is analyzed, and the screening criteria are determined based on the amplitude distribution.

[0055] For the positive influence subgraph, the positive influence screening criteria are determined based on the magnitude distribution of the non-zero elements in its adjacency matrix: (13) For the negative influence subgraph, the negative influence screening criteria are determined based on the magnitude distribution of the non-zero elements in its adjacency matrix: (14) in, This represents a function that determines the filtering threshold based on a preset quantile threshold. The preset quantile parameters are used. Influence relationships with amplitudes less than the corresponding screening threshold are considered weak influence relationships and are eliminated.

[0056] In one implementation, the preset quantile parameter The value ranges from the 20th percentile to the 40th percentile. Those skilled in the art can adjust this value according to the actual power grid scale and forecast accuracy requirements. The specific value to be taken.

[0057] Through the above screening, the influential relationships with a greater degree of impact are retained, while the weak influential relationships with a smaller degree of impact are eliminated. This reduces redundant information in the second graph structure and prevents weak influential relationships from introducing non-critical related information during subsequent graph convolution propagation.

[0058] The adjacency matrix of the second graph structure constructed in the above manner As a continuous value matrix, it can quantify the similarity of the impact of different market participants' trading behaviors on power transmission line flow, and is not limited by whether there is a direct electrical connection between nodes. Even if two nodes are geographically far apart and there is no direct transmission line connection, as long as they have similar impact patterns on power transmission line flow, they can still have a high correlation strength in the second graph structure, thus distinguishing it from the first graph structure that only represents electrical connection relationships.

[0059] S4, merge the adjacency matrices of the first graph structure and the second graph structure to obtain the merged adjacency matrix; In this embodiment, to simultaneously characterize the physical connection relationships of the power grid and the inter-node impact relationships caused by the trading behavior of market participants, the adjacency matrix of the first graph structure is... Adjacency matrix of the second graph structure The fusion is performed to obtain the fused adjacency matrix. .

[0060] The adjacency matrix of the first graph structure is used to describe the electrical connection relationship between power grid nodes based on transmission lines, and its elements are mainly used to indicate whether there is a direct connection relationship between nodes; the adjacency matrix of the second graph structure is used to describe the power flow influence relationship between different power grid nodes based on the trading behavior of market entities, and its elements are used to indicate the strength of the influence relationship between nodes.

[0061] Since the first and second graph structures originate from different sources, the meanings and numerical distribution ranges of their adjacency matrices differ. Direct fusion would result in one graph structure's information dominating the fusion process. Therefore, this embodiment normalizes the two adjacency matrices based on their respective weight distribution characteristics.

[0062] The weight distribution feature is used to characterize the distribution of edge weights in the adjacency matrix. For the first adjacency matrix, the weight distribution feature includes the distribution of the number of node connections; for the second adjacency matrix, the weight distribution feature includes the numerical distribution of the influence on the strength of the relationship.

[0063] S41, for the first adjacency matrix Calculate the degree of each node: (15) Construct a degree matrix based on node degree. : (16) Based on the distribution of node connection relationships, the first adjacency matrix is ​​normalized: (17) in, This is the first adjacency matrix after normalization.

[0064] The above processing adjusts the information propagation capability between nodes with different numbers of connections, avoiding excessive influence of highly connected nodes on the graph convolution process.

[0065] S42, Adjacency Matrix of the Second Graph Structure The elements in the matrix represent the strength of the market behavior influence relationship between nodes, and their value range is determined by the degree of correlation of the influence coefficient. It contains positive and negative influence relationships (positive influence corresponds to exacerbating blockage, and negative influence corresponds to alleviating blockage), so it is normalized according to the distribution of edge weight magnitude in the second adjacency matrix.

[0066] Calculate the maximum absolute value of all elements in the second adjacency matrix, and then normalize the second adjacency matrix based on this maximum value: (18) in, The normalized second adjacency matrix has elements ranging from 1 to 10. , This represents the maximum absolute value of all elements in the second adjacency matrix. Through the above processing, the influence relationships in the second graph structure are mapped to a uniform scale while preserving the relative strength of these relationships.

[0067] S43, according to the preset fusion weights, fuse the normalized first adjacency matrix and the second adjacency matrix. Let the weights corresponding to the first adjacency matrix be... The weights corresponding to the second adjacency matrix are ,and The fused adjacency matrix is ​​then expressed as: (19) Furthermore, to adapt to the varying degrees of change in the trading behavior of market participants under different operating conditions, this embodiment can also dynamically adjust the fusion weights based on the degree of change in the trading behavior of market participants.

[0068] Specifically, the degree of change in trading behavior is determined based on the change in the injected power of each node within the current time period: (20) The weights of the second adjacency matrix are determined based on the degree of change in the transaction behavior. (twenty one) in, It is a monotonically increasing function. According to... Determine the weights corresponding to the first adjacency matrix.

[0069] When the degree of change in the trading behavior of market participants increases, the fusion weight corresponding to the second adjacency matrix is ​​increased, so that the fusion adjacency matrix pays more attention to the power flow impact relationship caused by market behavior; when the degree of change in the trading behavior of market participants decreases, the fusion weight corresponding to the first adjacency matrix is ​​increased, so that the fusion adjacency matrix pays more attention to the physical topology relationship of the power grid.

[0070] The fused adjacency matrix is ​​obtained through the above method. The fused adjacency matrix simultaneously includes the physical connection characteristics of the power grid and the influence characteristics of market participants' transaction behavior, which are used for subsequent node feature extraction in the graph convolutional network.

[0071] S5 integrates the adjacency matrix and the input graph of the running parameters into a convolutional network to extract node features. It then combines the features of the nodes at both ends of the line to determine the probability of line congestion risk and outputs a warning signal.

[0072] S51, based on the operating parameters obtained in step S1, determine the initial feature vector corresponding to each power grid node. The operating parameters include, but are not limited to, node active power, node reactive power, voltage amplitude, and voltage phase angle.

[0073] For any power grid node Its initial eigenvector Represented as: (twenty two) in, Indicates the active power of the node. Indicates the reactive power of the node. Indicates the node voltage amplitude. This indicates the phase angle of the node voltage.

[0074] Arrange the initial feature vectors of each node according to the node number to obtain the initial feature matrix of the nodes. .

[0075] S52 will merge the adjacency matrix With the initial feature matrix of the nodes Input graph convolutional network. The adjacency matrix is ​​fused. It simultaneously incorporates the electrical connections in the first graph structure and the influence of market participants' transaction behaviors in the second graph structure. Therefore, during graph convolution, node feature propagation considers not only the physical connections between adjacent nodes but also the influence relationships between different nodes resulting from changes in transaction behaviors.

[0076] Normalizing the fused adjacency matrix yields: (twenty three) in, To merge the adjacency matrix The corresponding degree matrix, This is the normalized fusion adjacency matrix.

[0077] Node feature update using graph convolutional networks: (twenty four) in, Indicates the first The node feature matrix after layer graph convolution. Indicates the first The learnable weight matrix of the layer, This represents the activation function (such as the ReLU function).

[0078] After multi-layer graph convolution calculations, the final node feature matrix is ​​obtained: (25) in, This represents the total number of layers in the graph convolutional network. Represents a power grid node The node characteristics.

[0079] S53, for transmission lines to be predicted Let the two ends of the power grid be respectively and From the node feature matrix Obtain the node features corresponding to the two nodes respectively. and .

[0080] Based on the differences and correlations between the characteristics of the two end nodes, the line characteristics of this transmission line are constructed. (Represented in vector form): (26) in, This represents the concatenation of features from both ends of the node, used to characterize the relationship between the two ends of the node. This represents the absolute value of the difference between the features of the two endpoints, used to characterize the difference relationship between the two endpoints. This represents the element-wise product of the features of the two endpoints, used to characterize the joint influence relationship between the two endpoints.

[0081] Line characteristics constructed in the above manner It also includes the self-state information of the nodes at both ends of the transmission line, the state difference information between the two ends, and the correlation and influence information between the two ends, providing a more complete line-level feature expression for subsequent congestion risk prediction.

[0082] S54, the obtained line characteristics Input the risk prediction model to obtain the congestion risk probability corresponding to the transmission line: (27) in, Indicates transmission line The probability of a blockage occurring. The larger the value, the higher the risk of the transmission line becoming blocked. This represents the mapping function corresponding to the risk prediction model.

[0083] In one implementation, the risk prediction model employs a fully connected prediction network, which outputs a corresponding congestion risk probability value by performing nonlinear mapping on line characteristics.

[0084] S55 will obtain the congestion risk probability of each transmission line. By comparing the results with preset risk conditions, the congestion warning results for each transmission line are determined.

[0085] Specifically, a first risk threshold is preset. Second risk threshold ,and The probability of congestion risk on a certain transmission line. When the line is deemed to have a high risk of congestion, a high-risk warning signal is generated; when When the line is deemed to have a medium-risk congestion risk, a medium-risk warning signal is generated; when If the risk of the route is deemed low, no warning will be triggered or a low-risk alert will be generated.

[0086] The final output includes a warning signal containing the transmission line identifier and its corresponding congestion risk probability and / or congestion warning level. This warning signal can be output to the power grid monitoring platform for operation and management personnel to identify congestion risks.

[0087] In this way, this embodiment can utilize the node characteristics after integrating the physical topology of the power grid and the influence of market participants' trading behavior to predict the risk of transmission congestion caused by multi-participant trading behavior, thereby improving the ability to characterize the influencing factors of market behavior in the transmission congestion early warning process.

[0088] S6. In this embodiment, after completing the prediction of transmission line congestion risk, the main market players that cause the target transmission line risk to increase are further analyzed.

[0089] Specifically, the target transmission line that generates the congestion warning signal is identified. Based on the influence coefficient determined in step S3.2 Obtain information from each power grid node. For the target transmission line The influence coefficient. Wherein, the magnitude of the influence coefficient... Represents a node The degree of influence of changes in injected power on the power flow of the target transmission line, and the sign of the influence coefficient. Indicates the direction of influence.

[0090] Based on the market entity access information corresponding to each node, the node-level influence coefficient is mapped to the market entity level. Let the market entity be... The connected power grid node is Its target transmission line degree of influence for: (28) in, For nodes For the target transmission line Influence coefficient, For market entities The change in the amount of electricity traded. A positive value indicates that changes in the trading behavior of the market participant have exacerbated the risk of congestion on the target route. A negative value indicates that changes in the trading behavior of the market participant have mitigated the risk of congestion on the target route.

[0091] When multiple market participants connect to the same grid node, the changes in the trading behavior of the multiple market participants jointly affect the change in the injected power of that node. At this time, the degree of influence of each market participant on the target line is calculated by multiplying the change in their respective trading power by the influence coefficient corresponding to that node.

[0092] Based on the degree of impact Market participants are ranked according to the absolute value of their impact, from largest to smallest. A larger absolute value indicates that the change in the trading behavior of that market participant contributes more to the congestion risk of the target transmission line. A list of entities affected by congestion is generated, containing the identifiers of the market participants ranked by impact, along with their corresponding impact values ​​and directions.

[0093] Using the above methods, the market entities causing transmission congestion can be identified based on the list of entities affected by the congestion, providing a basis for subsequent congestion management.

[0094] Example 2: A multi-entity transmission congestion early warning system for the power market, used to execute the multi-entity transmission congestion early warning method for the power market described in Example 1 above. Figure 2 This is a schematic diagram of the structure of a multi-entity transmission congestion early warning system for the power market provided in an embodiment of the present invention. Figure 2 As shown, the system includes: The data acquisition module is used to acquire power grid topology, operating parameters, and market participant transaction data; The first graph construction module is used to construct a first graph structure representing the electrical connection relationship between nodes based on the topology; The second graph construction module is used to determine the injected power change of the grid node associated with the trading behavior of market participants based on the transaction data, determine the impact coefficient on the power flow of the transmission line based on the injected power change, and construct the second graph structure based on the impact coefficient. The graph fusion module is used to fuse the adjacency matrices of the first graph structure and the second graph structure to obtain a fused adjacency matrix. The risk warning module is used to extract node features from the convolutional network of the fused adjacency matrix and the input graph of the running parameters, combine the node features at both ends of the line to determine the probability of line congestion risk, and output a warning signal.

[0095] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0096] Example 3: To further illustrate the implementation effect of the multi-entity transmission congestion early warning method for the power market provided by the present invention, this example uses the IEEE 39-node standard test system for simulation verification.

[0097] Figure 4 The topology diagram of the IEEE 39-node standard test system provided for embodiments of the present invention.

[0098] The IEEE 39-bus node test system comprises 39 bus nodes, 10 generator sets, and 46 transmission lines. In the diagram, nodes 1 to 29 are represented by a single loop, representing ordinary bus nodes; nodes 30 to 39 are represented by a double loop, representing generator nodes, corresponding to generator sets G1 to G10 respectively. Specifically, node 30 corresponds to G1, node 31 to G2, node 32 to G3, node 33 to G4, node 34 to G5, node 35 to G6, node 36 to G7, node 37 to G8, node 38 to G9, and node 39 to G10.

[0099] To further illustrate the mapping relationship between market participant transactions and grid nodes in the method of this invention, key nodes involved in the simulation example are marked in the figure. Specifically, node 32 is marked "(+120MW)", indicating that generator G3 increased its transaction volume by 120MW during that period; node 36 is marked "(+100MW)", indicating that generator G7 increased its transaction volume by 100MW; node 38 is marked "(+150MW)", indicating that generator G9 increased its transaction volume by 150MW; node 20 is marked "L5 (-200MW)", indicating that load-side generator L5 reduced its electricity demand by 200MW; and node 39 is marked "G10", indicating that generator G10 is the corresponding market participant at that node. These markings correspond one-to-one with the changes in transaction volume of market participants in Table 1.

[0100] In addition, transmission line 16-19 is highlighted in the figure with a thick red dashed line and labeled "Line 16-19". This line is the target line for triggering the high-risk warning signal in subsequent simulation verification, used to visually demonstrate the specific location where the blockage risk occurs.

[0101] The generating units in the system are considered as market participants in the electricity market, and a mapping relationship between market participants and grid nodes is established based on the grid connection nodes corresponding to each generating unit. The total system load is approximately 6150MW.

[0102] (a) Setting up market transaction scenarios During a simulated trading session in a spot market, the trading plans of some market participants changed relative to the baseline operating state.

[0103] The changes in the trading volume of each market participant are shown in Table 1.

[0104] Table 1 Changes in Electricity Transactions by Market Entities

[0105] Positive values ​​indicate an increase in injected power corresponding to an increase in the amount of electricity traded by market participants, while negative values ​​indicate a decrease in injected power corresponding to a decrease in the amount of electricity traded. L5 represents the load-side market participant, and its negative value indicates that the user has reduced their electricity demand due to spot market price signals.

[0106] In this example, the fusion weight is: , The higher weighting of the second adjacency matrix indicates that the behavior of market participants changes significantly during the current trading session, and that market behavior plays a significant role in risk prediction.

[0107] The prediction results show that the probability of blockage risk for transmission lines 16-19 is [missing information]. The load exceeded the preset high-risk threshold of 0.8, triggering a high-risk warning signal. The predicted power flow is approximately 468MW, indicating that the line is operating at near full capacity.

[0108] (ii) According to the method described in step S6, the impact subject analysis is performed on the target lines 16-19 that generate the early warning results.

[0109] Based on the impact coefficients corresponding to the trading behavior of each market participant and the change in node injected power, the contribution of each market participant to the congestion risk of line 16-19 is calculated, and the results are shown in Table 2.

[0110] Table 2. List of entities affected by the blockage on lines 16-19

[0111] As shown in Table 2, G3, G9, and G7 are the main market players that cause the increased risk of congestion on lines 16-19.

[0112] Among them, node 32 corresponding to G3 has no direct topological association with line 16-19, but due to the impact of power flow caused by changes in market transaction behavior, it still has a significant impact on line 16-19. This shows that the second graph structure constructed based on the influence relationship of market behavior can supplement the cross-regional influence relationship that the traditional physical topology graph structure cannot represent.

[0113] (III) Comparative Verification To verify the effectiveness of the second diagram structure and fusion mechanism in the method of the present invention, a comparison method is set up.

[0114] The comparison method uses the GCN prediction method based solely on the physical topology of the power grid, i.e., only the first graph structure is input, without considering the influence of market participants' trading behavior. Under the same training data, test data, and network hyperparameters, the prediction results of the two methods are shown in Table 3.

[0115] Table 3 Comparison of the accuracy of different methods in identifying blockage risks

[0116] The comparison results show that the method of this invention, by constructing a second graph structure representing the influence of market participants' trading behaviors and integrating it with the power grid's physical topology, enables the graph convolutional network to simultaneously acquire power grid physical connection information and power flow impact information caused by market trading behaviors. This improves the ability to identify transmission congestion risks caused by multi-participant trading behaviors. Under the same test conditions, the congestion risk identification accuracy of the method of this invention reaches 96.5%, which is a significant improvement compared to the GCN method that only uses the physical topology graph structure. Figure 5 This is a comparison chart showing the accuracy of blocking risk identification between the embodiments of the present invention and the comparative methods.

[0117] It should be noted that the above simulation examples are only used to illustrate the implementation process and technical effects of the method of the present invention, and do not constitute a limitation on the scope of protection of the present invention.

[0118] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0119] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0122] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early warning of transmission congestion in a multi-entity electricity market, characterized in that, Includes the following steps: Acquire power grid topology, operating parameters, and market participant transaction data; Construct a first graph structure representing the electrical connection relationships between nodes based on the topology; Based on the transaction data, determine the injected power change of the grid node associated with the transaction behavior of market participants, determine the impact coefficient on the power flow of the transmission line based on the injected power change, and construct the second graph structure based on the impact coefficient. The adjacency matrices of the first graph structure and the second graph structure are merged to obtain the merged adjacency matrix; The convolutional network that integrates the adjacency matrix and the input graph of the running parameters extracts node features, combines the features of nodes at both ends of the line to determine the probability of line congestion risk, and outputs a warning signal.

2. The method according to claim 1, characterized in that, The determination of the influence coefficient on the power flow of the transmission line based on the change in injected power includes: The influence coefficient of the injected power change on the power flow of each transmission line is determined based on the power transfer distribution factor.

3. The method according to claim 2, characterized in that, The process of fusing the adjacency matrices of the first graph structure and the second graph structure to obtain a fused adjacency matrix includes: Extract the weight distribution features of the adjacency matrix of the first graph structure and the adjacency matrix of the second graph structure respectively; Based on the weight distribution characteristics, the adjacency matrices of the first graph structure and the second graph structure are normalized respectively to obtain the normalized first adjacency matrix and the second adjacency matrix. Based on the preset fusion weights, the normalized first adjacency matrix and the second adjacency matrix are weighted and fused to obtain the fused adjacency matrix.

4. The method according to claim 2, characterized in that, The construction of the second graph structure based on the influence coefficient includes: Based on the sign of the influence coefficient, the influence coefficients are divided into a set of positive influence coefficients and a set of negative influence coefficients; A positive influence subgraph is constructed based on the set of positive influence coefficients, and a negative influence subgraph is constructed based on the set of negative influence coefficients. The positive influence subgraph and the negative influence subgraph are used to characterize the increasing and decreasing influence of market participants' trading behavior on power transmission line flow, respectively. The positive and negative influence subgraphs together form the second graph structure.

5. The method according to claim 4, characterized in that, The construction of the second graph structure based on the influence coefficient includes: Obtain the magnitude distribution of the impact coefficients corresponding to the positive and negative impact coefficient sets, and determine the screening criteria for impact relationships based on the magnitude distribution of the impact coefficients; The edges in the positive and negative influence subgraphs are filtered according to the influence relationship filtering conditions. The edges that meet the influence relationship filtering conditions are retained, and the second graph structure is formed based on the filtered positive and negative influence subgraphs.

6. The method according to claim 3, characterized in that, The step of weighted fusion of the normalized first adjacency matrix and the second adjacency matrix according to the preset fusion weight includes: The degree of change in market participants' trading behavior is determined based on the change in injection power corresponding to the market participants' transaction data within the current time period. The fusion weights corresponding to the first and second adjacency matrices are adjusted according to the degree of change in the trading behavior of market participants; Using the adjusted fusion weights, the normalized first and second adjacency matrices are weighted and fused to obtain the fused adjacency matrix.

7. The method according to claim 6, characterized in that, The process of determining the probability of line congestion risk by combining the characteristics of the nodes at both ends of the line and outputting an early warning signal includes: Obtain the node features corresponding to the nodes at both ends of the transmission line to be predicted; Based on the differences and relationships between the node characteristics of the nodes at both ends of the line, the line characteristics are constructed; By inputting the line characteristics into the risk prediction model, the probability of blockage risk of the transmission line can be obtained. The blocking warning result is determined based on the blocking risk probability and preset risk conditions, and the warning signal is output.

8. The method according to claim 7, characterized in that, Also includes: Based on the target transmission line that generated the congestion warning result, obtain the influence coefficient corresponding to the target transmission line; Based on the magnitude and direction of the influence coefficient, determine the degree of impact of each market participant's trading behavior on the congestion risk of the target transmission line; The market entities are ranked according to the degree of impact, and a list of entities affected by the blockage is generated.

9. The method according to claim 1, characterized in that, The step of determining the injected power change of the grid node associated with the trading behavior of market participants based on the transaction data includes: The change in trading volume for each market participant is determined based on the aforementioned trading data. Based on the grid access node corresponding to each market participant, the change in the traded electricity volume is mapped to the corresponding grid node; The change in injected power of the power grid node is determined based on the change in the transaction power volume corresponding to each power grid node.

10. A multi-entity transmission congestion early warning system for the electricity market, characterized in that, include: The data acquisition module is used to acquire power grid topology, operating parameters, and market participant transaction data; The first graph construction module is used to construct a first graph structure representing the electrical connection relationship between nodes based on the topology; The second graph construction module is used to determine the injected power change of the grid node associated with the trading behavior of market participants based on the transaction data, determine the impact coefficient on the power flow of the transmission line based on the injected power change, and construct the second graph structure based on the impact coefficient. The graph fusion module is used to fuse the adjacency matrices of the first graph structure and the second graph structure to obtain a fused adjacency matrix. The risk warning module is used to extract node features from the convolutional network of the fused adjacency matrix and the input graph of the running parameters, combine the node features at both ends of the line to determine the probability of line congestion risk, and output a warning signal.

Images