New Energy Grid Vulnerable Line Identification Method, System, Electronic Device and Medium

Abstract

The present invention provides a method, system, electronic device and medium for identifying vulnerable lines in a new energy power grid, belonging to the field of identifying key lines in a power grid. The method includes: obtaining the topological structure and current operation data of the new energy power grid; based on the topological structure and current operation data, generating current power grid graph data with lines and new energy units as nodes and busbars as edges; according to the current power grid graph data, based on a vulnerable line identification model, determining the current category of each line in the new energy power grid; the current category of each line is a vulnerable line or a non-vulnerable line; the vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set. By combining the topological structure and operation data of the new energy power grid, and using a power grid vulnerable line identification model based on a gated neural network to classify the lines in the new energy power grid, the identification speed and accuracy of vulnerable lines are improved.

Description

Technical Field

[0001] The present invention relates to the field of identification of key lines in power grids, and particularly to a method, system, electronic device and medium for identifying vulnerable lines in a power grid considering the impact of new energy on cascading failures. Background Art

[0002] While large-scale interconnected power grids greatly improve social benefits and the rationality of resource allocation, their security issues have become increasingly prominent. Frequent large-scale power supply interruptions worldwide are the best evidence. Research shows that most of these accidents are caused by cascading failures of some components in the system, that is, the process in which a failure of a certain component in the system triggers successive failures of other components in the network. Therefore, accurately identifying the lines that are prone to opening during the cascading failure process and whose withdrawal from operation will cause a wider range of cascading failures, that is, key vulnerable lines, is of great significance for power grid security control. Summary of the Invention

[0003] The purpose of the present invention is to provide a method, system, electronic device and medium for identifying vulnerable lines in a new energy power grid, which can improve the accuracy and efficiency of identifying vulnerable lines in the power grid.

[0004] To achieve the above purpose, the present invention provides the following solutions:

[0005] A method for identifying vulnerable lines in a new energy power grid, wherein the new energy power grid includes multiple new energy units, multiple lines and buses, and the method for identifying vulnerable lines in the new energy power grid includes:

[0006] Obtain the topological structure and current operation data of the new energy power grid; the current operation data includes the current output data of each new energy unit and the current power flow data of each line;

[0007] Based on the topological structure and the current operation data, taking lines and new energy units as nodes and buses as edges, generate current power grid graph data; the current power grid graph data includes the adjacency matrix and current eigenvector of each node; the current eigenvector of a new energy unit is the current output data of the corresponding new energy unit; the current eigenvector of a line is the current power flow data of the corresponding line;

[0008] According to the current power grid graph data, based on the vulnerable line identification model, determine the current category of each line in the new energy power grid; the current category of each line is a vulnerable line or a non-vulnerable line; the vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set; the training sample set includes multiple historical power grid graph data and the label information corresponding to each historical power grid graph data; each historical power grid graph data includes the adjacency matrix and historical eigenvector of each node; the label information includes the historical category of each line.

[0009] Optionally, the current output data of the new energy unit includes: the active power of the new energy unit, the reactive power of the new energy unit, the rated voltage of the new energy unit, and the rated capacity of the new energy unit.

[0010] Optionally, the current power flow data of the line includes: the active power on the line, the reactive power on the line, the voltage difference at the beginning and end of the line, and the phase angle difference.

[0011] Optionally, the gated graph neural network includes a propagation model, an output model, and a classification layer connected in sequence;

[0012] The method for establishing the vulnerable line identification model includes:

[0013] Obtain the topological structure and multiple sets of historical operation data of the new energy power grid;

[0014] For any set of historical operation data, based on the topological structure and the historical operation data, taking lines and new energy units as nodes and buses as edges, generate corresponding historical power grid graph data;

[0015] Establish a power grid simulation model based on the topological structure of the new energy power grid;

[0016] For any set of historical operation data, perform cascading fault simulation based on the power grid simulation model and the historical operation data, and determine the cascading fault chain and the cascading fault vector load;

[0017] According to the cascading fault chain and the cascading fault vector load, calculate the risk values of each line in the new energy power grid;

[0018] According to the risk values of each line, determine the historical categories of each line corresponding to the historical operation data;

[0019] For any historical power grid graph data, perform iterative learning on the historical feature vectors of each node in the historical power grid graph data through the propagation model to obtain the final state of each node;

[0020] For any node, determine the output feature of the node through the output model according to the final state and the historical feature vector of the node;

[0021] Classify the output features of each line through the classification layer to obtain the predicted category of each line;

[0022] According to the predicted category and the historical category of each line, based on the cross-entropy loss function, perform iterative training on the propagation model and the output model to obtain the vulnerable line identification model.

[0023] Optionally, according to the cascading fault chain and the cascading fault vector load, calculate the risk values of each line in the new energy power grid, specifically including:

[0024] For any line, according to the cascading fault chain and the cascading fault vector load, determine the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy generator tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line;

[0025] According to the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy generator tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line, calculate the risk value of the line.

[0026] Optionally, use the following formula to calculate the risk value of line i:

[0027] R(i) = H f (i)L(i)G(i)M(i);

[0028] Wherein, R(i) is the risk value of line i, H f (i) is the fault probability of line i in the cascading fault simulation, L(i) is the vector load caused by the disconnection of line i, G(i) is the total amount of new energy generator tripping caused by the disconnection of line i, and M(i) is the number of DC line outages caused by the disconnection of line i.

[0029] Optionally, the cross-entropy loss function is:

[0030]

[0031] Wherein, F loss is the loss function value, N is the total number of lines in the new energy power grid, a i is the historical category of line i, is the probability value that the predicted category of line i is a vulnerable line.

[0032] To achieve the above object, the present invention also provides the following solution:

[0033] A system for identifying vulnerable lines in a new energy power grid, the new energy power grid includes a plurality of new energy generators, multiple lines and buses, and the system for identifying vulnerable lines in the new energy power grid includes:

[0034] A current data acquisition unit, configured to acquire the topological structure and current operation data of the new energy power grid; the current operation data includes the current output data of each new energy generator and the current power flow data of each line;

[0035] A power grid diagram generation unit, connected to the current data acquisition unit, is configured to generate current power grid diagram data based on the topological structure and the current operation data, with lines and new energy units as nodes and buses as edges; the current power grid diagram data includes the adjacency matrix of each node and the current eigenvector; the current eigenvector of the new energy unit is the current output data of the corresponding new energy unit; the current eigenvector of the line is the current power flow data of the corresponding line.

[0036] A classification unit, connected to the power grid diagram generation unit, is configured to determine the current category of each line in the new energy power grid based on the current power grid diagram data and a vulnerable line identification model; the current category of each line is a vulnerable line or a non-vulnerable line; the vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set; the training sample set includes multiple historical power grid diagram data and the label information corresponding to each historical power grid diagram data; each historical power grid diagram data includes the adjacency matrix of each node and the historical eigenvector; the label information includes the historical category of each line.

[0037] To achieve the above object, the present invention also provides the following solutions:

[0038] An electronic device includes a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned method for identifying vulnerable lines in a new energy power grid.

[0039] To achieve the above object, the present invention also provides the following solutions:

[0040] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned method for identifying vulnerable lines in a new energy power grid.

[0041] According to the specific embodiments provided by the present invention, the following technical effects are disclosed: By combining the topological structure and operation data of the new energy power grid, with lines and new energy units as nodes and buses as edges, the power grid operation conditions are transformed into a graph form, and a power grid vulnerable line identification model based on a gated graph neural network is used to classify the lines in the new energy power grid to identify vulnerable lines, improving the identification speed and accuracy of vulnerable lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0043] Figure 1 This is a flow chart of the new energy power grid vulnerable line identification method of the present invention;

[0044] Figure 2 It is a schematic diagram of power grid diagram data;

[0045] Figure 3 This is a schematic diagram of the vulnerable line identification process of the renewable energy grid;

[0046] Figure 4 It is a module schematic diagram of the new energy power grid vulnerable line identification system of the present invention.

[0047] Explanation of symbols:

[0048] G1, G2-new energy units, L1, L2, L3, L4-lines, current data acquisition unit-1, power grid diagram generation unit-2, classification unit-3. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] With the rise of a new generation of artificial intelligence technology represented by deep learning technology, there are new ideas for online identification of vulnerable lines in power grids. In order to realize online identification of vulnerable lines of new energy power grids, and at the same time improve the speed and accuracy of identifying vulnerable lines of power grids, and to be applicable to the identification of key vulnerable lines in different operating scenarios of new energy power grids, the present invention provides a method, system, electronic device and medium for identifying vulnerable lines of new energy power grids, and uses deep learning technology to mine the intrinsic connection between the historical operating conditions of new energy power grids and vulnerable lines and establish an identification model, thereby improving the online identification speed and accuracy of vulnerable lines.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Embodiment 1

[0053] like Figure 1 As shown, the new energy power grid vulnerable line identification method provided in this embodiment includes:

[0054] S1: Acquire the topological structure and current operation data of the new energy power grid, wherein the current operation data includes the current output data of each new energy unit and the current flow data of each line.

[0055] In this embodiment, the current output data of the new energy unit includes: the active power of the new energy unit, the reactive power of the new energy unit, the rated voltage of the new energy unit, and the rated capacity of the new energy unit. The current power flow data of the line includes: the active power on the line, the reactive power on the line, the voltage difference and phase angle difference at the beginning and end of the line.

[0056] S2: Based on the topological structure and the current operation data, taking the lines and new energy units as nodes and the busbars as edges, generate the current power grid graph data. The current power grid graph data includes the adjacency matrix of each node and the current eigenvector. The current eigenvector of the new energy unit is the current output data of the corresponding new energy unit. The current eigenvector of the line is the current power flow data of the corresponding line. As Figure 2 shown in the schematic diagram of the power grid graph data, where G1 and G2 are new energy units, and L1, L2, L3, and L4 are lines.

[0057] S3: According to the current power grid graph data, based on the vulnerable line identification model, determine the current category of each line in the new energy power grid. The current category of each line is a vulnerable line or a non-vulnerable line. The vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set. The training sample set includes multiple historical power grid graph data and the corresponding label information for each historical power grid graph data. Each historical power grid graph data includes the adjacency matrix of each node and the historical eigenvector. The label information includes the historical category of each line.

[0058] GGNN (Gated Graph Neural Network) is a classical graph neural network model for spatial domain message passing based on GRU (Gate Recurrent Unit). It enhances the ability of GNN (Graph Neural Network) to process input sequences, enhances the long-term memory ability of the network by adding a gating device, and effectively alleviates the problem of parameter gradient disappearance during backpropagation. It can extract the current power flow state of the line and the output situation of the new energy unit when iteratively learning the node features, and can better identify the vulnerable lines corresponding to the new energy power grid.

[0059] Furthermore, the gated graph neural network includes an input layer, a GGNN layer (propagation model and output model), a classification layer, and an output layer connected in sequence. In this embodiment, the classification layer uses a Softmax classifier. The input layer inputs the power grid graph data, including the eigenvector X Input of the node and the adjacency matrix C i ={c1, c2,...}. The output layer outputs the vulnerable lines according to the category of each line.

[0060] The method for establishing the vulnerable line identification model includes:

[0061] (1) Obtain the topological structure of the new energy power grid and multiple sets of historical operation data. The historical operation data includes the historical output data of each new energy unit and the historical power flow data of each line.

[0062] The historical output data of the new energy unit includes: the active power x of the new energy unit j,p , the reactive power x of the new energy unit j,q , the rated voltage x of the new energy unit j,u , and the rated capacity x of the new energy unit j,s .

[0063] The historical power flow data of the line includes: the active power x on the line i,p , the reactive power x on the line i,q , the voltage difference between the beginning and end of the line x i,u , and the phase angle difference x i,θ .

[0064] (2) Considering the impact of new energy access on the identification of vulnerable lines in the power grid, for any set of historical operation data, based on the topological structure and the historical operation data, taking the lines and new energy units as nodes and the buses as edges, generate the corresponding historical power grid graph data.

[0065] The historical power grid graph data includes the adjacency matrix of each node and the historical eigenvector. The historical eigenvector of the new energy unit is the historical output data of the corresponding new energy unit. The historical eigenvector of the line is the historical power flow data of the corresponding line. The historical eigenvector of each node is represented by X Input . When the node is a new energy unit, X Input = {x j,p , x j,q , x j,u , x j,s}. When the node is a line, X Input = {x i,p , x i,q , x i,u , x i,θ}.

[0066] Specifically, the adjacency matrix C i = {c1, c2,...} represents the connection relationship between line node i and other nodes. If node i is connected to node 1, then c1 is 1; otherwise, it is 0.

[0067] (3) Establish a power grid simulation model based on the topological structure of the new energy power grid.

[0068] (4) For any set of historical operation data, perform cascading failure simulation based on the power grid simulation model and the historical operation data to determine the cascading failure chain and the cascading failure vector load. Specifically, build a power grid simulation model, set different operation scenarios, consider line overload-dominated cascading failures, and perform N-1 cascading failure simulation on the lines to obtain the cascading failure chain and the cascading failure load loss.

[0069] (5) Calculate the risk value of each line in the new energy power grid according to the cascading failure chain and the cascading failure vector load.

[0070] In order to more accurately analyze the output of different new energy units and the impact of AC / DC transmission on cascading failures, and identify the vulnerable lines in the new energy power grid, the line fault probability, the load loss caused, the total amount of new energy unit tripping, and the number of DC line outages need to be considered when calculating the line risk value.

[0071] Specifically, for any line, according to the cascading failure chain and the cascading failure vector load, determine the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy unit tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line.

[0072] Use the formula to calculate the fault probability H f (i) of line i in the cascading failure simulation.

[0073] Use the formula to calculate the vector load loss L(i) caused by the disconnection of line i.

[0074] Use the formula to calculate the total amount of new energy unit tripping caused by the disconnection of line i.

[0075] Among them, n i represents the number of times line i appears in the set of cascading failure chains, N f represents the total number of cascading failure chains, λ i is the number of operating switchgears, used to represent the weight of fault recovery difficulty, P loss represents the load loss value caused by the cascading failure chain containing the disconnection of line i, k loss represents the fault recovery difficulty coefficient. When the faulty equipment is a line, k loss =1; when the faulty equipment is a transformer, k loss =1.5; when the faulty equipment is a bus or a generator, k loss =3, ω j represents the weight of new energy unit j, d j represents the electrical distance of new energy unit j from the fault location, b jIndicates the disconnection sequence of the new energy unit j, P g Indicates the amount of new energy unit tripping caused by the cascading fault chain including the disconnection of line i, m i Indicates the number of new energy units tripped due to the disconnection of line i.

[0076] Calculate the risk value of the line according to the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy unit tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line.

[0077] In this embodiment, the following formula is used to calculate the risk value of line i:

[0078] R(i) = H f (i)L(i)G(i)M(i);

[0079] Among them, R(i) is the risk value of line i, H f (i) is the fault probability of line i in the cascading fault simulation, L(i) is the load loss caused by the disconnection of line i, G(i) is the total amount of new energy unit tripping caused by the disconnection of line i, and M(i) is the number of DC line outages caused by the disconnection of line i.

[0080] (6) According to the risk values of each line, determine the historical categories of each line corresponding to the historical operation data. Specifically, sort the risk values of each line in descending order, and the first 10 lines L = {l1, l2,..., l 10} are vulnerable lines, and the remaining lines are non-vulnerable lines.

[0081] In this embodiment, considering the impact of new energy on cascading faults, combined with the power grid topology structure and operation data, the power grid operation conditions are transformed into a graph form (historical power grid graph data), reflecting the output of new energy units and the power flow state of lines, as the basis for determining vulnerable lines in the new energy power grid. The 10 vulnerable lines are used as labels for the historical power grid graph data to construct a training sample set.

[0082] (7) For any historical power grid graph data, perform iterative learning on the historical feature vectors of each node in the historical power grid graph data through the propagation model to obtain the final state h v (t) of each node.

[0083] (8) For any node, determine the output feature o v of the node according to the final state and historical feature vector of the node through the output model: o v = g(h v (t), X Input ).

[0084] In this embodiment, the propagation model and the output model adopt the existing GGNN network for iterative learning, and the specific learning process will not be elaborated here.

[0085] (9) Classify the output features of each line through the classification layer to obtain the predicted category of each line. That is, input the output features of the nodes of each line obtained into Softmax classification to obtain the corresponding predicted category (vulnerable line or non-vulnerable line).

[0086] In this embodiment, a vulnerable line identification model based on GGNN is constructed. The historical power grid map data under different operating conditions of the power grid is used as the model input, and the model output is the vulnerable lines of the power grid under this operating condition (label information: including the historical categories of each line), and the output can be expressed as:

[0087] A Output ={a1,a2,...,a i ,...,a n};

[0088] Among them, A Output represents the identification result of the vulnerable line output, a i represents whether line i is a vulnerable line. If it is, it is 1, otherwise it is 0, and n is the total number of lines.

[0089] (10) Based on the cross-entropy loss function, iteratively train the propagation model and the output model according to the predicted category and historical category of each line to obtain a vulnerable line identification model.

[0090] Specifically, the cross-entropy loss function is:

[0091]

[0092] Among them, F loss is the loss function value, N is the total number of lines in the new energy power grid, a i is the historical category of line i, is the probability value that the predicted category of line i is a vulnerable line.

[0093] As Figure 3 shown is a schematic diagram of the overall process of identifying vulnerable lines in the new energy power grid of the present invention.

[0094] In this embodiment, the power grid map sample set (historical power grid map data and corresponding label information) is divided into a training set and a test set in a ratio of 8:2 and input into the vulnerable line identification model based on GGNN. The GGNN model is used to perform feature learning on the historical power grid map data, and a Softmax classifier is used to classify the nodes, where new energy units are not included in the classification. The model is trained offline to output the vulnerable lines of the power grid under different operating states, and the test set is used to verify the identification ability of the model.

[0095] The vulnerable line identification model based on GGNN constructed by the present invention can autonomously learn the input power grid map data, analyze the uncertain cascading failures that may occur in the new energy power grid under different operating states, and correspondingly output the vulnerable lines under this operating state, greatly improving the identification speed and accuracy.

[0096] Embodiment 2

[0097] In order to execute the method corresponding to the above Embodiment 1 to achieve the corresponding functions and technical effects, a new energy power grid vulnerable line identification system is provided below.

[0098] As Figure 4 shown, the new energy power grid vulnerable line identification system provided in this embodiment includes: a current data acquisition unit, a power grid map generation unit, and a classification unit.

[0099] Among them, the current data acquisition unit is used to acquire the topological structure and current operating data of the new energy power grid. The current operating data includes the current output data of each new energy unit and the current power flow data of each line.

[0100] The power grid map generation unit is connected to the current data acquisition unit. The power grid map generation unit is used to generate current power grid map data based on the topological structure and the current operating data, with lines and new energy units as nodes and busbars as edges. The current power grid map data includes the adjacency matrix of each node and the current feature vector. The current feature vector of the new energy unit is the current output data of the corresponding new energy unit. The current feature vector of the line is the current power flow data of the corresponding line.

[0101] The classification unit is connected to the power grid map generation unit. The classification unit is used to determine the current category of each line in the new energy power grid based on the current power grid map data and the vulnerable line identification model; the current category of each line is a vulnerable line or a non-vulnerable line. The vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set. The training sample set includes a plurality of historical power grid map data and the label information corresponding to each historical power grid map data. Each historical power grid map data includes the adjacency matrix of each node and the historical feature vector. The label information includes the historical category of each line.

[0102] Embodiment 3

[0103] This embodiment provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the new energy grid vulnerable line identification method of Embodiment 1.

[0104] Optionally, the above electronic device may be a server.

[0105] In addition, an embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the new energy grid vulnerable line identification method of Embodiment 1.

[0106] The present invention uses a deep learning model to mine the correspondence between the grid operating conditions and the grid vulnerable lines. The influence of new energy on the identification of vulnerable lines is considered in the model, including:

[0107] (1) Considering the situation of new energy unit tripping and DC line outage, the number of new energy unit outages and the number of DC line outages after the cascading failure are incorporated into the risk value calculation.

[0108] (2) Considering the randomness of new energy output, combining the grid topology and operation data, taking the lines and new energy units as nodes and the buses as edges, transforming the grid operating conditions into a graph form, reflecting the output situation in the node features of the new energy units, and using the graph form of the grid operating conditions as the input.

[0109] (3) Construct a vulnerable line identification model based on GGNN. The gated graph neural network in the model is used to autonomously learn the mapping relationship between the grid graph samples and the vulnerable lines, output the feature information of each node, and the Softmax classifier is used to output the vulnerable line identification result. First, the model is offline trained to learn the correspondence between the historical operating conditions and the vulnerable lines. When applied online, the grid operating conditions are directly input into the model to obtain the new energy grid vulnerable lines under the new operating conditions, realizing end-to-end identification of new energy grid vulnerable lines.

[0110] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0111] Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for identifying vulnerable lines in a new energy power grid, wherein the new energy power grid includes multiple new energy generating units, multiple lines and buses, and is characterized in that, The method for identifying vulnerable lines in the new energy power grid includes: Obtaining the topological structure and current operation data of the new energy power grid; the current operation data includes the current output data of each new energy unit and the current power flow data of each line; Based on the topological structure and the current operation data, taking lines and new energy units as nodes and buses as edges, generating current power grid graph data; the current power grid graph data includes the adjacency matrix and current eigenvectors of each node; the current eigenvector of a new energy unit is the current output data of the corresponding new energy unit; the current eigenvector of a line is the current power flow data of the corresponding line; According to the current power grid graph data, based on the vulnerable line identification model, determining the current category of each line in the new energy power grid; the current category of each line is a vulnerable line or a non-vulnerable line; the vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set; the training sample set includes multiple historical power grid graph data and the label information corresponding to each historical power grid graph data; each historical power grid graph data includes the adjacency matrix and historical eigenvectors of each node; the label information includes the historical category of each line; The gated graph neural network includes a propagation model, an output model, and a classification layer connected in sequence; The method for establishing the vulnerable line identification model includes: Obtaining the topological structure and multiple groups of historical operation data of the new energy power grid; For any group of historical operation data, based on the topological structure and the historical operation data, taking lines and new energy units as nodes and buses as edges, generating the corresponding historical power grid graph data; Establishing a power grid simulation model based on the topological structure of the new energy power grid; For any group of historical operation data, performing a cascading failure simulation based on the power grid simulation model and the historical operation data to determine the cascading failure chain and the cascading failure vector load; According to the cascading failure chain and the cascading failure vector load, calculating the risk value of each line in the new energy power grid; According to the risk value of each line, determining the historical category of each line corresponding to the historical operation data; For any historical power grid graph data, iteratively learning the historical eigenvectors of each node in the historical power grid graph data through the propagation model to obtain the final state of each node; For any node, determining the output feature of the node through the output model according to the final state and historical eigenvector of the node; Classifying the output features of each line through the classification layer to obtain the predicted category of each line; According to the predicted category and historical category of each line, based on the cross-entropy loss function, iteratively training the propagation model and the output model to obtain the vulnerable line identification model.

2. The new energy grid vulnerable line identification method according to claim 1, wherein, The current output data of the new energy unit includes: the active power of the new energy unit, the reactive power of the new energy unit, the rated voltage of the new energy unit, and the rated capacity of the new energy unit.

3. The method for identifying vulnerable lines in a new energy power grid according to claim 1, wherein The current power flow data of the line includes: the active power on the line, the reactive power on the line, the voltage difference and phase angle difference at the beginning and end of the line.

4. The method for identifying vulnerable lines in a new energy power grid according to claim 1, wherein Calculate the risk values of each line in the new energy power grid according to the cascading fault chain and the cascading fault vector load, specifically including: For any line, determine the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy generator tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line according to the cascading fault chain and the cascading fault vector load; Calculate the risk value of the line according to the fault probability of the line, the load loss caused by the disconnection of the line, the total amount of new energy generator tripping caused by the disconnection of the line, and the number of DC line outages caused by the disconnection of the line.

5. The new energy grid vulnerable line identification method according to claim 4, characterized in that Use the following formula to calculate the risk value of line i: R(i) = H f (i)L(i)G(i)M(i); where R(i) is the risk value of line i, and H f (i) is the fault probability of line i in the cascading failure simulation, L(i) is the loss of load caused by the disconnection of line i, G(i) is the total amount of new energy generator tripping caused by the disconnection of line i, and M(i) is the number of DC line outages caused by the disconnection of line i.

6. The method for identifying vulnerable lines in a new energy power grid according to claim 1, wherein The cross-entropy loss function is: Among them, F loss is the loss function value, N is the total number of lines in the new energy power grid, and a i is the historical category of line i, and is the probability value that the predicted category of line i is a vulnerable line.

7. A vulnerable line identification system for a new energy power grid, wherein the new energy power grid includes multiple new energy generating units, multiple lines and buses, and is characterized in that, The new energy power grid vulnerable line identification system includes: A current data acquisition unit, configured to acquire the topological structure and current operation data of the new energy power grid; the current operation data includes the current output data of each new energy generator and the current power flow data of each line; A power grid graph generation unit, connected to the current data acquisition unit, configured to generate current power grid graph data with lines and new energy generators as nodes and buses as edges based on the topological structure and the current operation data; the current power grid graph data includes the adjacency matrix of each node and the current eigenvector; the current eigenvector of a new energy generator is the current output data of the corresponding new energy generator; the current eigenvector of a line is the current power flow data of the corresponding line; A classification unit, connected to the power grid graph generation unit, configured to determine the current category of each line in the new energy power grid based on the current power grid graph data and a vulnerable line identification model; the current category of each line is a vulnerable line or a non-vulnerable line; the vulnerable line identification model is obtained by pre-training a gated graph neural network with a training sample set; the training sample set includes multiple historical power grid graph data and the label information corresponding to each historical power grid graph data; each historical power grid graph data includes the adjacency matrix of each node and the historical eigenvector; the label information includes the historical category of each line; The gated graph neural network includes a propagation model, an output model, and a classification layer connected in sequence; The method for establishing the vulnerable line identification model includes: Obtain the topological structure and multiple groups of historical operation data of the new energy power grid; For any group of historical operation data, generate corresponding historical power grid graph data with lines and new energy generators as nodes and buses as edges based on the topological structure and the historical operation data; Establish a power grid simulation model based on the topological structure of the new energy power grid; For any group of historical operation data, perform cascading fault simulation based on the power grid simulation model and the historical operation data to determine the cascading fault chain and the cascading fault vector load; Calculate the risk values of each line in the new energy power grid according to the cascading fault chain and the cascading fault vector load; Determine the historical category of each line corresponding to the historical operation data according to the risk value of each line; For any historical power grid diagram data, iterative learning is performed on the historical feature vectors of each node in the historical power grid diagram data through the propagation model to obtain the final state of each node; For any node, the output feature of the node is determined through the output model according to the final state and historical feature vector of the node; The output features of each line are classified through the classification layer to obtain the predicted category of each line; Based on the predicted category and historical category of each line, the propagation model and the output model are iteratively trained based on the cross-entropy loss function to obtain a vulnerable line identification model.

8. An electronic device, characterized in that, It includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the new energy power grid vulnerable line identification method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by a processor, it implements the new energy power grid vulnerable line identification method according to any one of claims 1 to 6.

Images