A method for constructing a calculation model for the severity index of power grid transient voltage

Abstract

This invention relates to a method for constructing a calculation model for the transient voltage severity index of a power grid. It constructs a static scenario set and a transient scenario set of the power grid based on simulated random conditions, and further constructs a heterogeneous graph structure data of the power grid. A fault set is constructed based on the static scenario set, and the transient voltage severity index of all nodes is calculated through time-domain simulation of the fault set. A heterogeneous graph database is constructed based on the heterogeneous graph structure data and the transient voltage severity index. The constructed heterogeneous graph attention network is trained using the heterogeneous graph database to obtain the calculation model for the transient voltage severity index of the power grid. The calculation model constructed by this invention considers the uncertainties of various static and transient parameters in the database construction and mines the attention and semantic attention of power grid nodes through the heterogeneous graph attention network, improving the accuracy of rapid calculation of the transient voltage severity index of the power grid and meeting the calculation requirements for transient voltage stability of AC / DC hybrid power grids.

Description

Technical Field

[0001] This invention relates to the field of power grid index calculation technology, specifically to a method for constructing a calculation model for the severity index of transient voltage in a power grid. Background Technology

[0002] In recent years, with the increase in installed capacity of new energy power plants, the increase in the number of power electronic devices, and the large-scale commissioning of high-voltage direct current transmission projects, the power grid has continued to expand in scale and become increasingly complex in structure, posing a huge challenge to its transient safe and stable operation. Therefore, how to provide a rapid calculation method for transient voltage stability of AC / DC hybrid power grids has become a key and difficult point in power system transient stability analysis.

[0003] Currently, transient voltage assessment methods are categorized into three types: time-domain simulation, direct methods, and artificial intelligence (AI) methods. Time-domain simulation, by constructing complex system models, achieves accurate results but suffers from drawbacks such as long processing times and high computational demands. Direct methods struggle to accurately reflect the transient characteristics of complex AC / DC hybrid power grids. Artificial intelligence, with its rapid computational capabilities, offers new approaches to power grid transient voltage assessment and calculation. However, some AI methods lack consideration for power grid topology, transient parameters, or homogenize different types of nodes in the power grid, failing to accurately distinguish their potential differences.

[0004] Therefore, existing artificial intelligence methods cannot accurately characterize AC / DC hybrid power grids and lack in-depth research on the scenarios. Furthermore, most fast calculation methods are limited to binary classification of stability, lacking further research on severity indicators. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method for constructing a calculation model for the severity index of power grid transient voltage, so as to solve the problem that the existing technical solutions cannot accurately characterize AC / DC hybrid power grids and lack in-depth research on the scenario.

[0006] According to a first aspect of the present invention, a method for constructing a calculation model for the severity index of power grid transient voltage is provided, comprising:

[0007] Step S1: Construct a static scenario set and a transient scenario set for the power grid based on simulated random conditions. The simulated random conditions include random power output conditions of new energy sources, random load change conditions, injected power of high voltage direct current, proportion of induction motors, SVG configuration scheme, and low voltage ride-through control logic.

[0008] Step S2: Calculate the reactive power reserve for each scenario in the static scenario set; select effective features from the static and transient scenario sets as node features based on node type; add directed edges and edge features to the static scenario set based on power flow direction and line admittance to construct the heterogeneous graph structure data of the power grid.

[0009] Step S3: Construct a fault set based on the static scenario set. Based on the heterogeneous graph structure data of the power grid, calculate the transient voltage severity index of all nodes by performing time-domain simulation on the fault set, and construct a heterogeneous graph database based on the transient voltage severity index.

[0010] Step S4: Define meta-paths based on power grid flow direction; construct a heterogeneous graph attention network based on meta-paths, node features, and edge features; train the heterogeneous graph attention network using the heterogeneous graph database to obtain a calculation model for the severity index of power grid transient voltage.

[0011] Preferably, a static scenario set for the power grid is constructed based on simulated random conditions, specifically as follows:

[0012] Based on simulated random conditions, the random static parameters of the corresponding nodes are obtained. A static scenario set for the power grid is constructed by modifying the initial static data of the power grid, including:

[0013] Typical historical data of new energy power plants and load nodes are obtained, and the relationship between the typical historical data and time is fitted to obtain deterministic prediction results. A disturbance term for simulating random fluctuations is added to the deterministic prediction results to obtain the random output conditions of new energy and the random change conditions of load.

[0014] A random floating factor within a first preset range is generated as the injected power of the high-voltage direct current.

[0015] Based on the calculated active and reactive power deficits in the power grid, the power grid balancing method is set as adjusting the output of synchronous generators and increasing or decreasing local reactive power compensation.

[0016] Based on the random output conditions of new energy sources, the random change conditions of load, the injected power of high voltage direct current and the grid balancing mode, the random static parameters of the corresponding nodes are obtained. By modifying the initial static data of the grid, multiple static scenarios of the grid are constructed, and the scenarios that satisfy power flow convergence in the multiple static scenarios are integrated into a static scenario set.

[0017] Preferably, a set of transient scenarios for the power grid is constructed based on simulated random conditions, specifically as follows:

[0018] Based on the simulated random conditions, the random transient parameters of the corresponding nodes are obtained. By modifying the initial transient data of the power grid, a set of transient scenarios for the power grid is constructed, including:

[0019] A random percentage within a second preset range is generated as the percentage of induction motors for each load node;

[0020] Randomly generate SVGs of different capacities within a preset range, and add the generated SVGs to any load node as an SVG configuration scheme;

[0021] Generate a third preset range of random numbers with a step size of 0.01 as the low voltage ride-through control logic for the new energy power station;

[0022] Based on the proportion of induction motors, SVG configuration scheme and low voltage ride-through control logic, the random transient parameters of the corresponding nodes are obtained. By modifying the initial transient data of the power grid, multiple transient scenarios of the power grid are constructed and integrated into a transient scenario set.

[0023] Preferably, in step S2, reactive power reserve calculation is performed, effective features are screened, and the heterogeneous grid structure data is constructed, including:

[0024] Based on a static scenario set, the voltage-to-reactive-injection VQ curve is obtained by controlling the node voltage, and the reactive power reserve for each scenario is calculated based on the VQ curve.

[0025] Based on static and transient scenario sets, voltage, phase angle, active power output, reactive power output, reactive power reserve, and low voltage ride-through control logic of new energy power stations are selected as six-dimensional point features of PV nodes.

[0026] Based on static and transient scenario sets, voltage, phase angle, active load, reactive load, reactive power reserve, local reactive power compensation capacity, SVG capacity, and induction motor ratio are selected as eight-dimensional point features of PQ nodes.

[0027] Based on a static scene set, directed edges are set according to the power flow direction, and line admittance is selected as a two-dimensional edge feature to construct a heterogeneous graph structure data of the power grid.

[0028] Preferably, in step S3, time-domain simulation is performed on the fault set in batches, and transient voltage severity indicators are calculated to construct a heterogeneous graph database, including:

[0029] Based on the static scenario set, N-1 faults with an interruption time of 0.1 seconds are set on lines with voltage levels of 220kV and above. Faults that operate in islanded mode are removed from the faults, and the remaining faults are integrated into a fault set.

[0030] Based on static scene sets, transient scene sets, and power grid heterogeneous graph structure data, time-domain simulation is performed by simulating fault batches in the fault set, and the directed edges corresponding to the power grid heterogeneous graph structure data are removed to obtain the heterogeneous graph database input.

[0031] The transient voltage severity index of all nodes is calculated using the multi-binary table method based on the input from the heterogeneous graph database.

[0032] An initial heterogeneous graph database is constructed based on the transient voltage severity index and the heterogeneous graph database input.

[0033] The data in the initial heterogeneous graph database is normalized to obtain the heterogeneous graph database.

[0034] Preferably, based on the power flow direction of the power grid, a meta-path is defined, including:

[0035] Based on the power flow direction of the power grid, three meta-paths, PV-PQ, PQ-PQ and PQ-PV, are defined to describe the power flow transmission relationship between nodes, and the three meta-paths are integrated into a meta-path set.

[0036] Preferably, the constructed heterogeneous graph attention network consists of two multilayer perceptrons of different scales, three node attention layers, one semantic-level aggregation layer, and one multilayer perceptron, used for:

[0037] By using two multilayer perceptrons of different scales, nonlinear transformations are applied to node features and edge features respectively to obtain node features and edge features of the same dimension.

[0038] Based on the metapath set and node and edge features of the same dimension, different attention values ​​are assigned to adjacent nodes of the metapath through a node-level attention mechanism with three layers of node attention, thus obtaining the set of node embedding representations corresponding to the metapath set.

[0039] Based on the meta-path set and the node embedding representation set, the semantic-level aggregation layer aggregates the different power flow propagation semantic information extracted from different meta-paths to obtain the node features after semantic-level aggregation.

[0040] The node features after semantic-level aggregation are subjected to nonlinear transformation to obtain the predicted node transient voltage severity index. A loss function is set to compare the distance between the predicted value and the true value.

[0041] Preferably, node features and edge features are nonlinearly transformed using two multilayer perceptrons of different scales to obtain node features and edge features of the same dimension, including:

[0042] The six-dimensional point features are nonlinearly transformed using a multilayer perceptron and the ReLU activation function to obtain the mapped PV features h'. pv ;

[0043] The eight-dimensional point features are nonlinearly transformed using a multilayer perceptron and the ReLU activation function to obtain the mapped PQ features h'. pq ;

[0044] By using a multilayer perceptron and the ReLU activation function to perform a nonlinear transformation on the edge features, the mapped edge features h are obtained. e '.

[0045] Preferably, based on the meta-path set and node and edge features of the same dimension, a node-level attention mechanism with three layers of node attention layers is used to assign different attention values ​​to adjacent nodes of the meta-path, resulting in a set of node embedding representations corresponding to the meta-path set, including:

[0046] Based on the metapath set and the mapped feature h' pv 、Feature h' pv and edge features h' e Build a node attention layer with 3 layers;

[0047] For each node attention layer, calculate the weight coefficients of each node to its neighboring nodes on the metapath; use the weight coefficients to aggregate the features mapped to the neighboring nodes to obtain the embedding representation based on the metapath; concatenate the embedding representations obtained from all attention heads to obtain the set of node embedding representations corresponding to the metapath set.

[0048] Preferably, based on the meta-path set and the node embedding representation set, a semantic-level aggregation layer is used to aggregate the different power flow propagation semantic information extracted from different meta-paths, resulting in node features after semantic-level aggregation, including:

[0049] The node attention layer calculates the meta-path weight coefficient for each meta-path, and the embedding representation of the meta-path is aggregated using the meta-path weight coefficient to obtain the node features after semantic-level aggregation.

[0050] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0051] It is understood that the technical solution presented in this invention can construct a static scenario set and a transient scenario set of the power grid based on simulated random conditions, and further construct heterogeneous graph structure data of the power grid; construct a fault set based on the static scenario set, calculate the transient voltage severity index of all nodes by performing time-domain simulation on the fault set, construct a heterogeneous graph database based on the heterogeneous graph structure data of the power grid and the transient voltage severity index, and train the constructed heterogeneous graph attention network using the heterogeneous graph database to obtain the calculation model of the power grid transient voltage severity index. The calculation model constructed by this invention considers the uncertainty of various static and transient parameters in the construction of the database, and improves the accuracy of rapid calculation of the power grid transient voltage severity index by mining the attention and semantic attention of power grid nodes through the heterogeneous graph attention network, thus meeting the calculation requirements of transient voltage stability of AC / DC hybrid power grids.

[0052] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0053] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0054] Figure 1 This is a schematic diagram illustrating the steps of constructing a calculation model for the severity index of power grid transient voltage, according to an exemplary embodiment.

[0055] Figure 2 This is a flowchart illustrating a method for constructing a calculation model for the severity index of power grid transient voltage, according to an exemplary embodiment. Detailed Implementation

[0056] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0057] In one embodiment, Figure 1 This is a schematic diagram illustrating the steps of constructing a calculation model for the severity index of power grid transient voltage, according to an exemplary embodiment. See also... Figure 1 This paper provides a method for constructing a calculation model for the severity index of power grid transient voltage, including:

[0058] Step S1: Construct a static scenario set and a transient scenario set for the power grid based on simulated random conditions. The simulated random conditions include random power output conditions of new energy sources, random load change conditions, injected power of high voltage direct current, proportion of induction motors, SVG configuration scheme, and low voltage ride-through control logic.

[0059] The static and transient scenarios of the power grid constructed in this embodiment are based on simulated random conditions and take into account various uncertainties.

[0060] Step S2: Calculate the reactive power reserve of each scenario in the static scenario set; select effective features from the static and transient scenario sets as node features based on node type; add directed edges and edge features to the static scenario set based on power flow direction and line admittance to construct the heterogeneous graph structure data of the power grid.

[0061] Step S3: Construct a fault set based on the static scenario set. Based on the heterogeneous graph structure data of the power grid, calculate the transient voltage severity index of all nodes by performing time-domain simulation on the fault set, and construct a heterogeneous graph database based on the transient voltage severity index.

[0062] Step S4: Define meta-paths based on power grid flow direction; construct a heterogeneous graph attention network based on meta-paths, node features, and edge features; train the heterogeneous graph attention network using the heterogeneous graph database to obtain a calculation model for the severity index of power grid transient voltage.

[0063] It is understood that the technical solution shown in this embodiment can construct a static scenario set and a transient scenario set of the power grid based on simulated random conditions, and further construct heterogeneous graph structure data of the power grid; construct a fault set based on the static scenario set, calculate the transient voltage severity index of all nodes by performing time-domain simulation on the fault set, construct a heterogeneous graph database based on the heterogeneous graph structure data of the power grid and the transient voltage severity index, and train the constructed heterogeneous graph attention network using the heterogeneous graph database to obtain the calculation model of the power grid transient voltage severity index. The calculation model constructed in this embodiment considers the uncertainty of various static and transient parameters in the construction of the database, and improves the accuracy of the rapid calculation of the power grid transient voltage severity index by mining the attention and semantic attention of power grid nodes through the heterogeneous graph attention network, thus meeting the calculation requirements of transient voltage stability of AC / DC hybrid power grids.

[0064] Figure 2 This is a flowchart illustrating a method for constructing a calculation model for the severity index of power grid transient voltage, according to an exemplary embodiment. See [link / reference]. Figure 2 The following explanation is provided.

[0065] In step S1, a static scenario set for the power grid is constructed based on simulated random conditions. Specifically,

[0066] Based on simulated random conditions, the random static parameters of the corresponding nodes are obtained. A static scenario set for the power grid is constructed by modifying the initial static data of the power grid, including:

[0067] S11. Obtain typical historical data of new energy power plants and load nodes, fit the relationship between the typical historical data and time to obtain deterministic prediction results, add a disturbance term to the deterministic prediction results to simulate random fluctuations, and obtain the random output conditions of new energy and the random change conditions of load.

[0068] In practice, linear regression is used to fit the relationship between the typical historical data and time to obtain deterministic prediction results. An added perturbation term follows a normal distribution to simulate random fluctuations.

[0069] Furthermore, to ensure that the simulation results are not negative, constraints need to be applied to the results. The specific formula is as follows:

[0070] P new (t)=max(P new,h (t)+ε(t),0)

[0071] P load (t)=max(P load,h (t)+ε(t),0)

[0072] Among them, P new (t) represents the active power of the new energy power plant; P load (t) represents the active power at the load node; P new,h (t) represents the deterministic prediction result of the active power of the energy plant; P load,h ε(t) represents the deterministic prediction result of the active power of the load node; ε(t) represents the disturbance term that conforms to a normal distribution, with a standard deviation of 5 and a mean of 0.

[0073] S12. Generate a random floating factor within a first preset range as the injection power of the high-voltage direct current.

[0074] The first preset range can be [0.8, 1.2], and a random floating factor with a generation range of [0.8, 1.2] is generated to randomly adjust the injected power of high voltage DC between 80% and 120%.

[0075] S13. Based on the calculated active and reactive power deficits of the power grid, the power grid balancing method is set to adjust the output of synchronous generators and increase or decrease local reactive power compensation.

[0076] S14. Based on the random output conditions of new energy sources, random load changes, the injected power of high-voltage direct current, and the grid balancing method, obtain the random static parameters of the corresponding nodes. By modifying the initial static data of the grid, construct the N-type grid. dat A static scene set D is formed by integrating multiple static scenes that satisfy the power flow convergence requirement. dat .

[0077] It should be noted that the transient scenario set of the power grid is constructed based on simulated random conditions, specifically,

[0078] Based on the simulated random conditions, the random transient parameters of the corresponding nodes are obtained. By modifying the initial transient data of the power grid, a set of transient scenarios for the power grid is constructed, including:

[0079] S15. Generate a random percentage within a second preset range as the percentage of induction motors for each load node.

[0080] The second preset range is [0.4, 0.6]. A random number (random proportion) in the range [0.4, 0.6] is generated by uniform distribution to simulate the proportion of induction motors at each load node.

[0081] S16. Randomly generate SVGs of different capacities within a preset quantity range, and add the generated SVGs to any load node as an SVG configuration scheme.

[0082] The preset quantity range is 1, 2, or 3, and the capacity range is 50, 100, 150, or 200 MVA. Randomly generate 1 to 3 SVGs with capacities of 50, 100, 150, or 200 MVA and add them to any load node to simulate the power grid SVG configuration scheme.

[0083] S17. Generate a random number with a third preset range step size of 0.01 as the low voltage ride-through control logic for the new energy power station.

[0084] The third preset range is [0.88, 0.92], and a random number with a step size of 0.01 is generated in the range [0.88, 0.92] to simulate the low voltage ride-through control logic of the new energy power station.

[0085] S18. Based on the proportion of induction motors, the SVG configuration scheme, and the low-voltage ride-through control logic, obtain the random transient parameters of the corresponding nodes. By modifying the initial transient data of the power grid, construct the N-type of the power grid. swi These transient scenarios are integrated into a transient scenario set D. swi .

[0086] It should be noted that in step S2, reactive power reserve calculation is performed, effective features are screened, and the heterogeneous grid structure data is constructed, including:

[0087] Step S21: Based on the static scenario set, obtain the VQ curve of voltage versus reactive power injection by controlling the node voltage, and calculate the reactive power reserve of each scenario based on the VQ curve.

[0088] In practice, based on the static scene set D dat A virtual synchronous condenser is deployed at the node, and the voltage at the node is controlled to obtain the VQ curve of voltage versus reactive power injection. The reactive power reserve for all scenarios is obtained through multiple batch calculations.

[0089] Step S22: Based on the static scenario set and the transient scenario set, select voltage, phase angle, active power output, reactive power output, reactive power reserve and low voltage ride-through control logic of new energy power stations as six-dimensional point features of PV nodes.

[0090] Step S23: Based on the static scene set and the transient scene set, select the voltage, phase angle, active load, reactive load, reactive power reserve, local reactive power compensation capacity, SVG capacity and induction motor ratio as the eight-dimensional point features of the PQ node.

[0091] Step S24: Based on the static scene set, set directed edges according to the power flow direction, select line admittance as the two-dimensional edge feature, and construct the heterogeneous graph structure data of the power grid.

[0092] It should be noted that in step S3, time-domain simulation is performed on the fault set in batches, and transient voltage severity indicators are calculated to construct a heterogeneous graph database, including:

[0093] Step S31: Based on the static scenario set, set up N-1 faults with an interruption time of 0.1 seconds on lines with voltage levels of 220kV and above. Eliminate faults that are operating in an islanded manner and integrate the remaining faults into fault set D. lsd Let N be the number of faults in the fault set. lsd .

[0094] Step S32: Based on the static scenario set, transient scenario set, and power grid heterogeneous graph structure data, time-domain simulation is performed by simulating fault batches from the fault set. Directed edges corresponding to the power grid heterogeneous graph structure data are removed to obtain the heterogeneous graph database input. The number of heterogeneous graph database inputs is N. data =N dat *N swi *N lsd That is, the number of static scenes N dat Number of transient scenarios N sw i and the number of faults N lsd The product of.

[0095] Step S33: Calculate the transient voltage severity index of all nodes using the multi-binary table method based on the heterogeneous graph database input.

[0096] The formula for calculating the severity index of transient voltage is as follows:

[0097]

[0098] Among them, Λ n,i S is an index of the severity of transient voltage at node i under sample n; n Si is the sum of the areas of the sub-regions of the binary table corresponding to node i under sample n; S0 is the critical area of ​​the corresponding binary table; t n τ represents the moment when the corresponding voltage threshold value is not met. n The moment when the corresponding voltage threshold value is met; V e This is the rated voltage; V n,i (t) represents the actual voltage trajectory, V cr Voltage threshold value of a binary meter.

[0099] Step S34: Construct an initial heterogeneous graph database based on the transient voltage severity index and the heterogeneous graph database input.

[0100] Step S35: Normalize the data in the initial heterogeneous graph database to obtain a heterogeneous graph database. Use Min-Max normalization to scale all features of the data to the range [0,1] to obtain the heterogeneous graph database.

[0101] It should be noted that in step S4, the meta-path is defined according to the power flow direction, including:

[0102] Based on the power flow direction of the power grid, three meta-paths—PV-PQ, PQ-PQ, and PQ-PV—are defined to describe the power flow transmission relationships between nodes. These three meta-paths are then integrated into a meta-path set {Φ0, Φ1, Φ2}. This meta-path set {Φ0, Φ1, Φ2} reflects the semantic information of power transmission in the power grid.

[0103] It should be noted that the constructed heterogeneous graph attention network consists of two multilayer perceptrons of different scales, three node attention layers, one semantic-level aggregation layer, and one multilayer perceptron layer, used for:

[0104] Step S41: Using two multilayer perceptrons of different scales, nonlinear transformations are applied to the node features and edge features respectively to obtain node features and edge features of the same dimension.

[0105] Specifically, a (6, 24) multilayer perceptron and the ReLU activation function are used to perform a nonlinear transformation on the six-dimensional point features (feature dimension 6) to obtain the mapped PV features h'. pv The eight-dimensional point features (feature dimension 8) are nonlinearly transformed using a multilayer perceptron of dimension (8, 24) and the ReLU activation function to obtain the mapped PQ features h'. pq The edge features (feature dimension 2) are nonlinearly transformed using a multilayer perceptron of dimension (2, 16) and the ReLU activation function to obtain the mapped edge features h. e '.

[0106] The mapping formula is as follows:

[0107] h' pv =Relu(MLP(h) pv ,θ mlp,pv ))

[0108] h' pq =Relu(MLP(h) pq ,θ mlp,pq ))

[0109] h e =Relu(MLP(h) e ,θ mlp,e ))

[0110] Among them, h' pv h' represents the mapped PV features. pq h' represents the mapped PQ features. eThe edge features are represented by the mapped edge features, MLP stands for Multilayer Perceptron, ReLU stands for activation function, and h represents the edge features after mapping. pv h represents the initial features of the PV node. pq h represents the initial features of the PQ node. e Denotes the initial features of the edge, θ mlp,pv θ mlp,pq and θ mlp,e This represents the learnable parameters in deep learning.

[0111] Step S42: Based on the metapath set and the node features (six-dimensional and eight-dimensional point features) and edge features of the same dimension, different attention values ​​are assigned to adjacent nodes of the metapath through the node-level attention mechanism of the three-layer node attention layer, so as to obtain the node embedding representation set corresponding to the metapath set.

[0112] Specifically, based on the metapath set {Φ0,Φ1,Φ2} and the mapped feature h' pv 、Feature h' pv and edge features h' e A node attention layer with three layers is constructed. For each node attention layer, the weight coefficient of each node i on the metapath Φ relative to its neighbor node j is calculated. The features mapped from the neighbor nodes are aggregated using the weight coefficients to obtain the embedding representation based on the metapath. The embedding representations obtained from all attention heads are concatenated and passed through the three node attention layers to obtain the set of node embedding representations corresponding to the metapath set {Φ0, Φ1, Φ2}.

[0113] The specific formula is as follows:

[0114]

[0115] in, This represents the weight coefficient of node i on the meta-path Φ relative to its neighbor node j. The embedding feature of node i after aggregation and concatenation on the meta-path Φ is represented by ReLU, σ represents the softmax activation function, and h' i h' j and h' k This represents the features mapped from nodes i, j, and k. This represents the node attention vector of the metapath Φ. Let K represent the set of all nodes d1 on the metapath Φ of node i, and K represent the number of attention heads.

[0116] Step S43: Based on the meta-path set and the node embedding representation set, the semantic-level aggregation layer aggregates the different power flow propagation semantic information extracted from different meta-paths to obtain the node features after semantic-level aggregation.

[0117] Specifically, the meta-path weight coefficient for each meta-path is calculated through the node attention layer. By aggregating the embedding representations of meta-paths using meta-path weight coefficients, we obtain the node features Z after semantic-level aggregation. sem .

[0118] The formula is as follows:

[0119]

[0120] in, Atten represents the weight coefficient of each meta-path. sem This represents a semantic-level attention layer. Z represents a set of nodes embedded in each other. sem This represents the node features after semantic-level aggregation.

[0121] Step S44: Perform nonlinear transformation on the node features after semantic-level aggregation to obtain the predicted node transient voltage severity index, and set a loss function to compare the distance between the predicted value and the true value.

[0122] Specifically, it refers to: node features Z based on semantic-level aggregation. sem A multilayer perceptron of dimension (64, 1) and the ReLU activation function are used for nonlinear transformation to obtain the predicted severity index of node transient voltage. A loss function is set to compare the distance between the predicted value and the true value. The specific formula is as follows:

[0123]

[0124] in, The term represents the severity index of the predicted transient voltage at node i, L represents the mean square error loss function, MLP represents the multilayer perceptron, and Z represents the mean square error loss function. sem,i Λ represents the node features after semantic-level aggregation. i represents the true value of the transient voltage severity index at node i, and N represents the total number of nodes in the heterogeneous graph data.

[0125] In step S4, the heterogeneous graph attention network is trained, including: inputting the PV, PQ and edge features of the power grid, outputting the power grid transient voltage severity index, completing the training of the neural network, and obtaining the power grid transient voltage severity index calculation model.

[0126] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0127] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0128] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0129] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0130] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0131] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0132] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0133] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0134] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a calculation model for the severity index of power grid transient voltage, characterized in that, include: Step S1: Constructing a static scenario set and a transient scenario set for the power grid based on simulated random conditions. Specifically, this involves obtaining the random static parameters of the corresponding nodes based on the simulated random conditions, constructing a static scenario set for the power grid by modifying the initial static data of the power grid, and obtaining the random transient parameters of the corresponding nodes based on the simulated random conditions, constructing a transient scenario set for the power grid by modifying the initial transient data of the power grid. The simulated random conditions include random power output conditions of new energy sources, random load change conditions, injected power of high-voltage direct current, proportion of induction motors, SVG configuration scheme, and low-voltage ride-through control logic. Step S2: Calculate the reactive power reserve for each scenario in the static scenario set; select effective features from the static and transient scenario sets as node features based on node type; add directed edges and edge features to the static scenario set based on power flow direction and line admittance to construct the heterogeneous graph structure data of the power grid. Step S3: Construct a fault set based on the static scenario set. Based on the heterogeneous graph structure data of the power grid, calculate the transient voltage severity index of all nodes by performing time-domain simulation on the fault set, and construct a heterogeneous graph database based on the transient voltage severity index. Step S4: Define meta-paths based on power grid flow direction; construct a heterogeneous graph attention network based on meta-paths, node features, and edge features; train the heterogeneous graph attention network using the heterogeneous graph database to obtain a calculation model for the severity index of power grid transient voltage.

2. The method for constructing the power grid transient voltage severity index calculation model according to claim 1, characterized in that, Based on simulated random conditions, the random static parameters of the corresponding nodes are obtained. A static scenario set for the power grid is constructed by modifying the initial static data of the power grid, including: Typical historical data of new energy power plants and load nodes are obtained, and the relationship between the typical historical data and time is fitted to obtain deterministic prediction results. A disturbance term for simulating random fluctuations is added to the deterministic prediction results to obtain the random output conditions of new energy and the random change conditions of load. A random floating factor within a first preset range is generated as the injected power of the high-voltage direct current. Based on the calculated active and reactive power deficits in the power grid, the power grid balancing method is set as adjusting the output of synchronous generators and increasing or decreasing local reactive power compensation. Based on the random output conditions of new energy sources, the random change conditions of load, the injected power of high voltage direct current and the grid balancing mode, the random static parameters of the corresponding nodes are obtained. By modifying the initial static data of the grid, multiple static scenarios of the grid are constructed, and the scenarios that satisfy power flow convergence in the multiple static scenarios are integrated into a static scenario set.

3. The method for constructing the power grid transient voltage severity index calculation model according to claim 2, characterized in that, Based on the simulated random conditions, the random transient parameters of the corresponding nodes are obtained. By modifying the initial transient data of the power grid, a set of transient scenarios for the power grid is constructed, including: A random percentage within a second preset range is generated as the percentage of induction motors for each load node; Randomly generate SVGs of different capacities within a preset range, and add the generated SVGs to any load node as an SVG configuration scheme; Generate a third preset range of random numbers with a step size of 0.01 as the low voltage ride-through control logic for the new energy power station; Based on the proportion of induction motors, SVG configuration scheme and low voltage ride-through control logic, the random transient parameters of the corresponding nodes are obtained. By modifying the initial transient data of the power grid, multiple transient scenarios of the power grid are constructed and integrated into a transient scenario set.

4. The method for constructing the power grid transient voltage severity index calculation model according to claim 1, characterized in that, In step S2, reactive power reserve is calculated, effective features are selected, and the heterogeneous grid structure data is constructed, including: Based on a static scenario set, the voltage-to-reactive-injection VQ curve is obtained by controlling the node voltage, and the reactive power reserve for each scenario is calculated based on the VQ curve. Based on static and transient scenario sets, voltage, phase angle, active power output, reactive power output, reactive power reserve, and low voltage ride-through control logic of new energy power stations are selected as six-dimensional point features of PV nodes. Based on static and transient scenario sets, voltage, phase angle, active load, reactive load, reactive power reserve, local reactive power compensation capacity, SVG capacity, and induction motor ratio are selected as eight-dimensional point features of PQ nodes. Based on a static scene set, directed edges are set according to the power flow direction, and line admittance is selected as a two-dimensional edge feature to construct a heterogeneous graph structure data of the power grid.

5. The method for constructing the power grid transient voltage severity index calculation model according to claim 1, characterized in that, In step S3, time-domain simulations are performed on the fault set in batches, and transient voltage severity indices are calculated to construct a heterogeneous graph database, including: Based on the static scenario set, N-1 faults with an interruption time of 0.1 seconds are set on lines with voltage levels of 220kV and above. Faults that operate in islanded mode are removed from the faults, and the remaining faults are integrated into a fault set. Based on static scene sets, transient scene sets, and power grid heterogeneous graph structure data, time-domain simulation is performed by simulating fault batches in the fault set, and the directed edges corresponding to the power grid heterogeneous graph structure data are removed to obtain the heterogeneous graph database input. The transient voltage severity index of all nodes is calculated using the multi-binary table method based on the input from the heterogeneous graph database. An initial heterogeneous graph database is constructed based on the transient voltage severity index and the heterogeneous graph database input. The data in the initial heterogeneous graph database is normalized to obtain the heterogeneous graph database.

6. The method for constructing the power grid transient voltage severity index calculation model according to claim 4, characterized in that, Based on the power flow direction of the power grid, define the meta-path, including: Based on the power flow direction of the power grid, three meta-paths, PV-PQ, PQ-PQ and PQ-PV, are defined to describe the power flow transmission relationship between nodes, and the three meta-paths are integrated into a meta-path set.

7. The method for constructing the power grid transient voltage severity index calculation model according to claim 6, characterized in that, The constructed heterogeneous graph attention network consists of two multilayer perceptrons of different scales, three node attention layers, one semantic-level aggregation layer, and one multilayer perceptron layer, used for: By using two multilayer perceptrons of different scales, nonlinear transformations are applied to node features and edge features respectively to obtain node features and edge features of the same dimension. Based on the metapath set and node and edge features of the same dimension, different attention values ​​are assigned to adjacent nodes of the metapath through a node-level attention mechanism with three layers of node attention, thus obtaining the set of node embedding representations corresponding to the metapath set. Based on the meta-path set and the node embedding representation set, the semantic-level aggregation layer aggregates the different power flow propagation semantic information extracted from different meta-paths to obtain the node features after semantic-level aggregation. The node features after semantic-level aggregation are subjected to nonlinear transformation to obtain the predicted node transient voltage severity index. A loss function is set to compare the distance between the predicted value and the true value.

8. The method for constructing the power grid transient voltage severity index calculation model according to claim 7, characterized in that, By using two multilayer perceptrons at different scales, nonlinear transformations are applied to node features and edge features to obtain node features and edge features of the same dimension, including: The six-dimensional point features are nonlinearly transformed using a multilayer perceptron and the ReLU activation function to obtain the mapped PV features. ; The eight-dimensional point features are nonlinearly transformed using a multilayer perceptron and the ReLU activation function to obtain the mapped PQ features. ; By using a multilayer perceptron and the ReLU activation function to perform a nonlinear transformation on the edge features, the mapped edge features are obtained. .

9. The method for constructing the power grid transient voltage severity index calculation model according to claim 8, characterized in that, Based on the metapath set and node and edge features of the same dimension, a node-level attention mechanism with three layers of node attention is used to assign different attention values ​​to adjacent nodes of the metapath, resulting in a set of node embedding representations corresponding to the metapath set, including: Based on the metapath set and the mapped features ,feature Sum of edge features Build a node attention layer with 3 layers; For each node attention layer, calculate the weight coefficients of each node to its neighboring nodes on the metapath; use the weight coefficients to aggregate the features mapped to the neighboring nodes to obtain the embedding representation based on the metapath; concatenate the embedding representations obtained from all attention heads to obtain the set of node embedding representations corresponding to the metapath set.

10. The method for constructing the power grid transient voltage severity index calculation model according to claim 9, characterized in that, Based on the meta-path set and the node embedding representation set, a semantic-level aggregation layer is used to aggregate the different power flow propagation semantic information extracted from different meta-paths, resulting in node features after semantic-level aggregation, including: The node attention layer calculates the meta-path weight coefficient for each meta-path, and the embedding representation of the meta-path is aggregated using the meta-path weight coefficient to obtain the node features after semantic-level aggregation.

Images