High-proportion new energy sending-end power grid transient voltage rise severity index evaluation method based on hyperbolic neural network

Abstract

The invention relates to a hyperbolic neural network-based high-proportion new energy sending-end power grid transient voltage rise severity index evaluation method. The method comprises the steps of constructing a power grid steady-state sample set considering operation mode uncertainty; a typical disturbance fault set including direct current locking, commutation failure and the like is set, transient voltage rise severity indexes of samples are calculated in batches through time domain simulation, and the indexes serve as supervision labels to construct a power grid graph database; constructing a hyperbolic neural network structure with negative curvature level representation characteristics, and performing supervised training; and inputting ground state power flow information of a sending-end power grid based on the trained hyperbolic neural network model to realize rapid evaluation of the transient voltage rise severity of the system. According to the method, based on the hyperbolic neural network, the high-proportion new energy sending-end power grid with a typical claw-shaped structure is efficiently represented by utilizing the learning capability of the negative curvature space of the hyperbolic neural network on an exponential expansion structure, and the evaluation accuracy is effectively improved while the transient voltage rise severity index is calculated in real time.

Description

Technical Field

[0001] This invention relates to a method for assessing the severity of transient voltage in power grids, and more particularly to a method for assessing the severity of transient voltage rise in power grids with a high proportion of renewable energy transmission based on hyperbolic neural networks. Background Technology

[0002] With a high proportion of renewable energy integrated into the grid, the transient voltage stability problem of the sending-end power grid is becoming increasingly prominent. Affected by the randomness of photovoltaic and wind power output and the dynamic characteristics of hydropower units, the power system is highly susceptible to transient instability phenomena such as voltage drops, oscillations, and unrecoverable events when subjected to large disturbances such as DC blocking, commutation failure, and line short circuits. Traditional simulation-based methods for analyzing the severity of transient voltage rise are computationally complex and lack real-time performance, making it difficult to meet the needs for quantitative stability assessment and online early warning after a high proportion of renewable energy integration.

[0003] In recent years, with the deep integration of artificial intelligence and power systems, data-driven transient voltage rise severity index assessment methods have gradually become a research hotspot. Deep neural networks, graph convolutional neural networks and other models can, to a certain extent, explore voltage dynamic characteristics and topological dependencies, but still have the following problems: (1) Existing methods are mainly based on Euclidean space modeling, which makes it difficult to accurately characterize the complex hierarchical and non-Euclidean geometric structure of the sending-end power grid; (2) The models are not sensitive enough to topological disturbances and have limited ability to represent discontinuous topological changes such as DC blocking and branch switching; (3) When faced with disturbances in multiple operating modes and multiple scenarios, the prediction accuracy and generalization ability of steady-state information-driven models decrease significantly. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method for evaluating the severity of transient voltage rise in a high-proportion renewable energy power grid based on a hyperbolic neural network. By introducing a hyperbolic geometric structure in a negative curvature space, a high-dimensional embedding expression of the characteristics of power grid nodes and branches is achieved, which efficiently characterizes a high-proportion renewable energy power grid with a typical claw-like structure. While meeting the requirements of real-time calculation, it improves the accuracy and robustness of the transient voltage rise severity prediction.

[0005] Technical solution: The method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks, as described in this invention, includes the following steps:

[0006] Step 1: Construct a steady-state sample set of the power grid that takes into account the uncertainties of its operation mode;

[0007] Step 2: Based on the steady-state sample set, set a typical disturbance fault set, perform time-domain simulation on each fault sample in the fault set, and calculate the transient voltage rise severity index; use the transient voltage rise severity index as a supervision label to construct a high-proportion renewable energy power grid graph database containing node characteristics and topological relationships;

[0008] Step 3: Based on the graph database, construct a hyperbolic neural network model with negative curvature hierarchical representation characteristics, and use the supervision labels to perform supervised training of the hyperbolic neural network;

[0009] Step 4: Based on the trained hyperbolic neural network model, input the ground-state power flow information and operating parameters of the sending-end power grid, perform real-time calculation of the transient voltage rise severity index, and output the power system transient voltage rise severity index and risk identification results.

[0010] Preferably, in step 1, for the initial operating mode of the power system, the random fluctuation range of photovoltaic and wind power is set, the reactive power support effect of the hydropower unit excitation system is considered, and the operating status of synchronous condensers and dynamic reactive power compensation devices is set to construct the power grid steady-state sample set considering the uncertainty of the operating mode.

[0011] Preferably, the construction method specifically includes:

[0012] Step 1-1: Based on the baseline operation mode of the sending-end power grid, select photovoltaic power stations and wind farm nodes, and set the photovoltaic and wind power outputs to fluctuate randomly within the range of 80% to 120% of the rated power, respectively;

[0013] Step 1-2: For the initial operation mode of the sending-end power grid, select the hydropower unit node, set the excitation system parameters and voltage amplitude adjustment range, and randomly generate the bus voltage amplitude of each generator and synchronous condenser within the range of [0.95, 1.05].

[0014] Steps 1-3: For the initial operation mode of the sending-end power grid, considering the operating status and capacity constraints of the synchronous condenser and static synchronous compensation device, set the reactive power output range and adjustment sensitivity of the condenser and device to form reactive power support and voltage regulation samples under multiple operating modes.

[0015] Steps 1-4: Based on the random parameter combination results from Steps 1-1 to 1-3, power flow calculations are used to generate node voltage, branch power flow, and active and reactive power distribution data, constructing a steady-state sample set of the power grid containing N different operating modes. .

[0016] Preferably, the method for calculating the severity index of transient pressure rise in step 2 is as follows:

[0017] Step 2-1: Set the initial operating mode of the sending-end power grid. A set of anticipated faults These include DC blocking, commutation failure, line short circuit, water-optical islanding, and critical branch disconnection operation conditions.

[0018] Step 2-2: Based on the steady-state sample set and anticipated fault set By using time-domain simulation to obtain the transient voltage response trajectory of nodes in batches, the dynamic voltage change process of the power system under disturbance can be obtained.

[0019] Steps 2-3: Based on the transient voltage response trajectory, and using the binary meter transient stability criterion, calculate the transient voltage rise severity index for each node. The calculation formula is as follows:

[0020] ,

[0021] In the formula, For the sample in the fault node The severity index of transient pressure rise; This is the rated voltage. For the actual voltage trajectory, Voltage threshold value of a binary meter; The moment when the corresponding voltage threshold value is not met; The moment when the corresponding voltage threshold value is met; The critical area of ​​the binary table; For the sample in the fault node Voltage below threshold The voltage drop area.

[0022] Preferably, the method for constructing the high-proportion renewable energy power grid map database in step 2 is as follows:

[0023] Steps 2-4: Based on the steady-state sample set ,Will As a supervisory label, the voltage of the feature node is selected. Node amplitude Active power and reactive power As input features, and based on grid connection relationships, a Euclidean space high-proportion renewable energy sending-end grid map database is constructed. ,in, For a set of nodes, For branch road collection, Node running characteristics As an indicator of the severity of transient pressure rise, For sample number, This is the fault number.

[0024] Preferably, the hyperbolic neural network model construction method in step 3 is as follows:

[0025] Step 3-1: Use a hyperbolic mapping layer Möbius linear layer Batch normalization layer With hyperbolic activation function Each layer forms the input layer of the hyperbolic neural network model:

[0026] ,

[0027] in, , , ,

[0028] In the formula, For node operation characteristics, Here is a hyperbolic mapping function used to map Euclidean space to a hyperbolic space with negative curvature. The curvature coefficient; Denotes the Möbius linear transform layer, where , These are learnable parameters; For batch normalization layer, , These are the batch mean and variance, respectively. To prevent division by zero of small constants; , These are learnable scaling and translation parameters; For hyperbolic space activation functions;

[0029] Step 3-2: Use hyperbolic graph convolution and activation function Constructing the hyperbolic graph convolutional layer of the hyperbolic neural network model:

[0030] ,

[0031] ,

[0032] In the formula, For hyperbolic graph convolution, For the first The node embedding features of the layer input, For the first Layer weight matrix; , They are nodes , The degree; For nodes The set of adjacent nodes; and These are the Möbius multiplication and addition operators, respectively;

[0033] Step 3-3: By fully connected layer Batch normalization layer With hyperbolic activation function The output layer of the hyperbolic neural network model is constructed in the following order:

[0034] ,

[0035] The predicted value of the transient pressure rise severity index output by the hyperbolic neural network model is expressed as follows:

[0036] .

[0037] Preferably, step 4 includes the following sub-steps:

[0038] Step 4-1: Based on the real-time monitoring data and power flow calculation results of the sending-end power grid, construct the input high-proportion renewable energy sending-end power grid diagram structure. ;

[0039] Step 4-2: Input the power grid diagram structure into the trained hyperbolic neural network model. The hyperbolic neural network model performs forward calculations based on the model parameters to obtain the predicted values ​​of the transient voltage rise severity index for each node. ;

[0040] Step 4-3: Predict the value based on the aforementioned indicator. Implement targeted reactive voltage control and prevention adjustment strategies;

[0041] Step 4-4: Generate a severity assessment report on transient voltage rise in the sending-end power grid, outputting node stability, risk level matrix, and reactive power regulation recommendations to achieve online voltage stability monitoring and proactive defense control for power grids with a high proportion of renewable energy transmission.

[0042] Preferably, step 4-1 includes: obtaining the power system's base-state voltage amplitude, active and reactive power distribution, node voltage angles, and equipment operating status parameters based on real-time monitoring data and power flow calculation results of the sending-end power grid, combined with the anticipated fault set. Construct the input power grid diagram structure ,,in For a set of nodes, For branch road collection, It is a real-time feature matrix, including node voltage, reactive power output, load level, and the operating status of synchronous condensers and STATCOM equipment.

[0043] Preferably, the predicted value of the indicator The calculation formula is:

[0044] ,

[0045] In the formula, For nodes The predicted value of the indicator, For the trained hyperbolic neural network model function, This is the optimal parameter set obtained from model training.

[0046] Preferably, step 4-3 includes: predicting the value based on the indicator. The stability of node voltage is graded and assessed, and based on the sensitivity coefficient of node voltage to reactive power, targeted reactive voltage control and preventive regulation strategies are implemented. If the node stability falls below a set threshold... Based on the sensitivity distribution results, the excitation of the synchronous condenser and the output of the STATCOM are adjusted to achieve early prevention and control of voltage support for high-risk nodes and system voltage stability.

[0047] Beneficial effects: The present invention has the following advantages: Based on steady-state operating data, it can combine multiple disturbance scenarios and graph-level geometric features to achieve rapid calculation and risk identification of transient voltage rise severity indicators, so as to meet the actual needs of safe and stable power grid operation under the conditions of large-scale transmission of new energy. Attached Figure Description

[0048] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0049] The technical solution of the present invention will be described in detail below with reference to the embodiments and accompanying drawings.

[0050] like Figure 1 As shown, a method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks includes the following steps:

[0051] Step 1: For the initial operating mode of the power system, set random fluctuations in photovoltaic, wind power, and load, set the voltage amplitudes of generators and synchronous condensers, and construct a steady-state sample set of the power grid considering the uncertainties of the operating mode:

[0052] Step 1-1: Based on the baseline operation mode of the sending-end power grid, select photovoltaic power stations and wind farm nodes, and set the photovoltaic and wind power outputs to fluctuate randomly within the range of 80% to 120% of the rated power, in order to simulate the uncertainty of new energy output;

[0053] Step 1-2: For the initial operation mode of the sending-end power grid, select the hydropower unit node, set the excitation system parameters and voltage amplitude adjustment range, and randomly generate the bus voltage amplitude of each generator and synchronous condenser within the range of [0.95, 1.05] to reflect the reactive power support and voltage regulation role of the hydropower unit.

[0054] Steps 1-3: For the initial operation mode of the sending-end power grid, considering the operating status and capacity constraints of the synchronous condenser and static synchronous compensator (STATCOM), set the reactive power output range and regulation sensitivity of the condenser and device to form reactive power support and voltage regulation samples under multiple operating modes, so as to reflect the impact of fast voltage support equipment on the steady-state characteristics of the power system.

[0055] Steps 1-4: Based on the random parameter combination results from Steps 1-1 to 1-3, power flow calculations are used to generate node voltage, branch power flow, and active and reactive power distribution data, constructing a steady-state sample set of the power grid containing N different operating modes. This provides a basis for input features in subsequent analysis of the severity of transient pressure rise indicators.

[0056] Step 2: Based on the steady-state power grid sample set from Step 1, set up a set of anticipated faults, and calculate the severity index of transient voltage rise in the samples in batches through time-domain simulation. Using this as a label, construct a high-proportion renewable energy-to-the-end power grid diagram database.

[0057] Step 2-1: Set the initial operating mode of the sending-end power grid. A set of anticipated faults The operating conditions include DC blocking, commutation failure, line short circuit, water-solar islanding, and disconnection of key branches, in order to cover the main types of transient voltage disturbances that may occur in the sending-end power grid.

[0058] Step 2-2: Based on the steady-state sample set from Step 1 and anticipated fault set By using time-domain simulation to obtain the transient voltage response trajectory of nodes in batches, the dynamic voltage change process of the power system under disturbance can be obtained.

[0059] Step 2-3: Based on the transient voltage response trajectory in Step 2-2, and using the binary meter transient stability criterion, calculate the transient voltage rise severity index for each node. For the fault node The calculation formula is as follows:

[0060] (1)

[0061] In the formula, For the sample in the fault node The severity index of transient pressure rise; This is the rated voltage. For the actual voltage trajectory, Voltage threshold value of a binary meter; The moment when the corresponding voltage threshold value is not met; The moment when the corresponding voltage threshold value is met; It is the critical area of ​​the binary table; Is the sample in the fault? node Voltage below threshold The voltage drop area;

[0062] Steps 2-4: Based on the steady-state sample set from Step 1 , calculate As a supervisory label, the voltage of the feature node is selected. Node amplitude Active power and reactive power As input features, and based on grid connection relationships, a Euclidean space high-proportion renewable energy sending-end grid map database is constructed. ,in For a set of nodes, For branch road collection, For node operation characteristics, As an indicator of the severity of transient pressure rise, For sample number, Number the fault;

[0063] Step 3: Based on the steady-state sample set from Step 1 and the high-proportion renewable energy power grid graph database from Step 2, construct a hyperbolic neural network containing an input layer, a hyperbolic graph convolutional layer, and an output layer. Train the model using label-supervised training, including the following steps:

[0064] Step 3-1: Use a hyperbolic mapping layer Möbius linear layer Batch normalization layer With hyperbolic activation function Each layer is used to construct the input layer. The calculation formula is as follows:

[0065] (2)

[0066] in:

[0067] (3)

[0068] (4)

[0069] (5)

[0070] In the formula, For node operation characteristics, Here is a hyperbolic mapping function used to map Euclidean space to a hyperbolic space with negative curvature. The curvature coefficient; Denotes the Möbius linear transform layer, where , These are learnable parameters; For batch normalization layer, , These are the batch mean and variance, respectively. To prevent division by zero of small constants; , These are learnable scaling and translation parameters; For hyperbolic space activation functions, such as Poincaré ReLU;

[0071] Step 3-2: Use hyperbolic graph convolution and activation function The hyperbolic graph convolutional layer is constructed using the following formula:

[0072] (6)

[0073] (7)

[0074] In the formula, For hyperbolic graph convolution, For the first The node embedding features of the layer input, For the first Layer weight matrix; , They are nodes , The degree; For nodes The set of adjacent nodes; and These are the Möbius multiplication and addition operators, respectively;

[0075] Step 3-3: By fully connected layer Batch normalization layer With hyperbolic activation function The output layer is constructed in the following order, and the calculation formula for the output layer is as follows:

[0076] (8)

[0077] The predicted value of the transient pressure rise severity index output by the hyperbolic neural network model is:

[0078] (9)

[0079] Step 4: Based on the hyperbolic neural network trained in Step 3, input the power grid ground-state power flow information and perform real-time transient voltage rise severity index assessment calculation, including the following steps:

[0080] Step 4-1: Based on the real-time monitoring data and power flow calculation results of the sending-end power grid, obtain the power system's base-state voltage amplitude, active and reactive power distribution, node voltage angles, and equipment operating status parameters, combined with the anticipated fault set. Construct the input power grid diagram structure ,in For a set of nodes, For branch road collection, It is a real-time feature matrix, including node voltage, reactive power output, load level, and operating status of equipment such as synchronous condensers and STATCOM;

[0081] Step 4-2: Input the graph structure constructed in Step 4-1 into the hyperbolic neural network model trained in Step 3, based on the model parameters. Forward calculations are performed to obtain the predicted values ​​of the transient pressure rise severity index for each node. This allows for real-time estimation of the severity of transient voltage rise in the power grid. The calculation formula is as follows:

[0082] (10)

[0083] In the formula, For the sample The predicted value of the indicator, This is the function of the trained hyperbolic neural network model. The optimal parameter set obtained from model training;

[0084] Step 4-3: Based on the index prediction results obtained in Step 4-2, classify and determine the node voltage stability, and combine this with the node voltage sensitivity coefficient to reactive power. Implement targeted reactive power control and preventative regulation strategies. If node stability falls below a set threshold... Based on the sensitivity distribution results, the excitation of the synchronous condenser and the output of the STATCOM are adjusted to achieve early prevention and control of voltage support for high-risk nodes and system voltage stability.

[0085] Step 4-4: Based on the results of Step 4-3, generate an assessment report on the severity of transient voltage rise in the sending-end power grid, and output the results including node stability, risk level matrix and reactive power regulation suggestions, so as to realize online voltage stability monitoring and active defense control of the sending-end power grid with a high proportion of new energy.

Claims

1. A method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks, characterized in that... Includes the following steps: Step 1: Construct a steady-state sample set of the power grid that takes into account the uncertainties of its operation mode; Step 2: Based on the steady-state sample set, set a typical disturbance fault set, calculate the transient voltage rise severity index for each fault sample in the fault set; use the transient voltage rise severity index as a supervision label to construct a high-proportion renewable energy power grid graph database containing node characteristics and topological relationships; Step 3: Based on the graph database, construct a hyperbolic neural network model with negative curvature hierarchical representation characteristics, and perform supervised training using the supervision labels; Step 4: Based on the trained hyperbolic neural network model, input the ground-state power flow information and operating parameters of the sending-end power grid, perform real-time calculation of the transient voltage rise severity index, and output the power system transient voltage rise severity index and risk identification results.

2. The method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 1, characterized in that, In step 1, for the initial operation mode of the power system, the random fluctuation range of photovoltaic and wind power is set, the reactive power support effect of the hydropower unit excitation system is considered, and the operation status of synchronous condensers and dynamic reactive power compensation devices is set to construct the power grid steady-state sample set considering the uncertainty of the operation mode.

3. The method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 2, characterized in that, The construction method specifically includes: Step 1-1: Based on the baseline operation mode of the sending-end power grid, select photovoltaic power stations and wind farm nodes, and set the photovoltaic and wind power outputs to fluctuate randomly within the range of 80% to 120% of the rated power, respectively; Step 1-2: For the initial operation mode of the sending-end power grid, select the hydropower unit node, set the excitation system parameters and voltage amplitude adjustment range, and randomly generate the bus voltage amplitude of each generator and synchronous condenser within the range of [0.95, 1.05]. Steps 1-3: For the initial operation mode of the sending-end power grid, considering the operating status and capacity constraints of the synchronous condenser and static synchronous compensation device, set the reactive power output range and adjustment sensitivity of the condenser and device to form reactive power support and voltage regulation samples under multiple operating modes. Steps 1-4: Based on the random parameter combination results from Steps 1-1 to 1-3, power flow calculation is used to generate node voltage, branch power flow and active and reactive power distribution data, and a power grid steady-state sample set containing several different operating modes is constructed.

4. The method for assessing the severity of transient voltage rise in high-proportion renewable energy transmission grids based on hyperbolic neural networks according to claim 1, characterized in that, The method for calculating the severity index of transient pressure rise in step 2 is as follows: Step 2-1: Set the initial operating mode of the sending-end power grid. A set of anticipated faults, including DC blockage, commutation failure, line short circuit, water-optical islanding, and critical branch disconnection operation conditions; Step 2-2: Based on the steady-state sample set and the expected fault set, the transient voltage response trajectory of the nodes is obtained in batches through time-domain simulation to obtain the dynamic voltage change process of the power system under disturbance; Steps 2-3: Based on the transient voltage response trajectory, and using the binary meter transient stability criterion, calculate the transient voltage rise severity index for each node. The calculation formula is as follows: ， In the formula, For the sample in the fault node The severity index of transient pressure rise; This is the rated voltage. For the actual voltage trajectory, Voltage threshold value of a binary meter; The moment when the corresponding voltage threshold value is not met; The moment when the corresponding voltage threshold value is met; The critical area of ​​the binary table; For the sample in the fault node Voltage below threshold The voltage drop area.

5. The method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 1, characterized in that, Step 2 involves constructing the high-proportion renewable energy transmission-end power grid diagram database as follows: Steps 2-4: Based on the steady-state sample set, determine the severity index of transient pressure rise. As a supervisory label, the voltage of the feature node is selected. Node amplitude Active power and reactive power As input features, and based on grid connection relationships, a Euclidean space high-proportion renewable energy sending-end grid map database is constructed. ,in, For a set of nodes, For branch road collection, Node running characteristics As an indicator of the severity of transient pressure rise, For sample number, This is the fault number.

6. The method for assessing the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 1, characterized in that, The hyperbolic neural network model construction method described in step 3 is as follows: Step 3-1: Use a hyperbolic mapping layer Möbius linear layer Batch normalization layer With hyperbolic activation function Each layer forms the input layer of the hyperbolic neural network model: ， in, , , , In the formula, For node operation characteristics, For hyperbolic mapping functions, The curvature coefficient; This represents a Möbius linear transform layer. , These are learnable parameters; For batch normalization layer, , These are the batch mean and variance, respectively. To prevent division by zero of small constants; , These are learnable scaling and translation parameters; For hyperbolic space activation functions; Step 3-2: Use hyperbolic graph convolution and activation function Constructing the hyperbolic graph convolutional layer of the hyperbolic neural network model: ， ， In the formula, For the first The node embedding features of the layer input, For the first Layer weight matrix; , They are nodes , The degree; For nodes The set of adjacent nodes; and These are the Möbius multiplication and addition operators, respectively; Step 3-3: By fully connected layer Batch normalization layer With hyperbolic activation function The output layer of the hyperbolic neural network model is constructed in the following order: ， The predicted value of the transient pressure rise severity index output by the hyperbolic neural network model is expressed as follows: 。 7. The method for assessing the severity of transient voltage rise in high-proportion renewable energy transmission grids based on hyperbolic neural networks according to claim 1, characterized in that, Step 4 includes the following sub-steps: Step 4-1: Based on the real-time monitoring data and power flow calculation results of the sending-end power grid, construct the input high-proportion renewable energy sending-end power grid diagram structure. ; Step 4-2: Input the power grid diagram structure into the trained hyperbolic neural network model. The hyperbolic neural network model performs forward calculations based on the model parameters to obtain the predicted values ​​of the transient voltage rise severity index for each node. ; Step 4-3: Predict the value based on the aforementioned indicator. Implement targeted reactive voltage control and prevention adjustment strategies; Step 4-4: Generate a severity assessment report on transient voltage rise in the sending-end power grid, outputting node stability, risk level matrix, and reactive power regulation recommendations to achieve online voltage stability monitoring and proactive defense control for power grids with a high proportion of renewable energy transmission.

8. The method for evaluating the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 7, characterized in that, Step 4-1 includes: obtaining the power system's base-state voltage amplitude, active and reactive power distribution, node voltage angles, and equipment operating status parameters based on real-time monitoring data and power flow calculation results from the sending-end power grid; and constructing the input power grid diagram structure by combining the anticipated fault set. ,,in For a set of nodes, For branch road collection, It is a real-time feature matrix, including node voltage, reactive power output, load level, and the operating status of synchronous condensers and STATCOM equipment.

9. The method for evaluating the severity of transient voltage rise in a high-proportion renewable energy power grid based on a hyperbolic neural network according to claim 7, characterized in that, The predicted value of the indicator The calculation formula is: ， In the formula, For nodes The predicted value of the indicator, For the trained hyperbolic neural network model function, This is the optimal parameter set obtained from model training.

10. The method for evaluating the severity of transient voltage rise in high-proportion renewable energy power grids based on hyperbolic neural networks according to claim 7, characterized in that, Step 4-3 includes: predicting the value based on the indicator. The stability of node voltage is graded and assessed, and based on the sensitivity coefficient of node voltage to reactive power, targeted reactive voltage control and preventive regulation strategies are implemented. If the node stability falls below a set threshold... Based on the sensitivity distribution results, the excitation of the synchronous condenser and the output of the STATCOM are adjusted to achieve early prevention and control of voltage support for high-risk nodes and system voltage stability.

Images