A power grid frequency stability determination method and system, electronic device and medium

Abstract

The application discloses a kind of power grid frequency stability determination method, system, electronic equipment and medium, it is related to electric power system field, combine plan and the time series frequency data of monitoring site, the transient time series animation generated can accurately and comprehensively reflect the actual operating condition of target power grid from two angles of time domain and space, the three-dimensional convolutional neural network model constructed in advance can directly determine the frequency stability of target power grid according to transient time series animation;With the space-time information that is easy to obtain, replace the complicated device modeling process, greatly reduce the workload, three-dimensional convolutional neural network model can extract the time series characteristics of target power grid frequency response and the spatial characteristics of combined geographic information simultaneously by transient time series animation, by constructing transient time series animation visual input information, improve the transparency of three-dimensional convolutional neural network model, ensure that the key information of input data is not missed, improve the reliability and accuracy of final frequency stability evaluation result.

Description

Technical Field

[0001] This invention relates to the field of power systems, and in particular to a method, system, electronic device, and medium for determining the frequency stability of a power grid. Background Technology

[0002] To address the long-distance gap between energy supply and demand, high-voltage direct current (HVDC) and high-voltage alternating current (HVAC) transmission technologies, characterized by long distances, high voltage, and large transmission capacity, have emerged. HVDC transmission technology effectively promotes the sustainable development of the power industry and improves resource optimization, facilitating the meeting of the growing load demands of the receiving-end power grid. However, while meeting the energy exchange needs of both sending and receiving grids, this technology also presents significant risks and challenges. In large-scale receiving-end systems with both ultra-high-voltage AC and ultra-high-voltage direct current (UHVDC) feeds, the transient frequency stability of the receiving-end power grid is severely threatened if a serious fault occurs in the load center of the receiving region, or if local active power sources cannot meet the active power demands of highly concentrated dynamic loads. Therefore, how to sense and assess the frequency of the power system so that grid operators can understand and predict regional grid frequency levels, intervene in advance, and prevent frequency instability in larger areas is an urgent problem to be solved.

[0003] Existing frequency dynamic analysis and evaluation methods mainly fall into three categories: the first is time-domain simulation based on numerical analysis; the second is equivalent modeling based on mathematical analysis; and the third is data-driven intelligent methods. The first method requires constructing a full-system model from the individual component models of the power system based on their topological relationships. Then, it calculates the frequency of each time-domain power system component step by step based on this full-system model. If the full-system model is inaccurate, the accuracy of the numerical results obtained from the simulation cannot accurately reflect the actual situation. Furthermore, large power grid systems contain numerous components, resulting in a very large computational load, requiring accurate models and powerful computing capabilities. The process of constructing the full-system model is prone to errors and is highly complex. The second method can simplify the modeling process to some extent, but given the large number of components in a large power grid system, if a large number of components are simplified using equivalent models, the final calculation results will be inaccurate. The simulation cannot closely approximate the actual application process, cannot accurately reflect the actual situation of the power system, and is difficult to adapt to practical needs. The third method mostly uses real-time phasor data from each node measured by a wide-area measurement system as sample input. This real-time phasor data is implemented in the form of two-dimensional or three-dimensional matrices, which cannot intuitively reflect the physical meaning of the input samples, thus exacerbating the black-box nature of the evaluation process and resulting in low reliability of the calculation results. Furthermore, all three methods in the existing technology neglect the impact of the spatial characteristics of the power system on frequency, failing to comprehensively evaluate the frequency stability of the power system. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, electronic device, and medium for determining the frequency stability of a power grid. It utilizes readily available spatiotemporal information to replace the complex device modeling process, greatly reducing workload. The three-dimensional convolutional neural network model can simultaneously extract the temporal characteristics of the target power grid's frequency response and the spatial characteristics combined with geographic information through transient time-series animation. By constructing transient time-series animation to visualize the input information of the three-dimensional convolutional neural network model, the transparency of the model is improved. The three-dimensional convolutional neural network model can ensure that key information in the input data is not omitted, thus improving the reliability and accuracy of the final frequency stability assessment results.

[0005] To address the aforementioned technical problems, this invention provides a method for determining the frequency stability of a power grid, comprising:

[0006] A plan view of the target power grid is constructed based on the geographic information of the target power grid;

[0007] A transient time-series animation of the target power grid is constructed based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid;

[0008] The frequency stability of the target power grid is determined using a pre-built three-dimensional convolutional neural network model and transient temporal animation of the target power grid.

[0009] Optionally, constructing a plan view of the target power grid based on its geographical information includes:

[0010] Geographic information of the target power grid is obtained, and a rectangular image is constructed based on the geographic information, wherein the length of the rectangular image is greater than the maximum longitude span and the width is greater than the maximum latitude span of the target power grid;

[0011] The rectangular image is evenly divided into several grids;

[0012] The grid containing the monitoring stations of the target power grid is defined as the measured grid, and the grid that does not contain the monitoring stations is defined as the virtual grid, resulting in a grid map with longitude as the horizontal axis and latitude as the vertical axis. The grid map is a plan view of the target power grid.

[0013] Optionally, the step of constructing a transient time-series animation of the target power grid based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid includes:

[0014] Acquire time-series frequency data from the monitoring stations of the target power grid;

[0015] The time-series frequency data of each grid in the grid diagram are determined based on the time-series frequency data of the monitoring stations.

[0016] The time-series frequency data of each grid are framed in chronological order to obtain the transient time-series animation of the target power grid.

[0017] Optionally, determining the time-series frequency data of each grid in the grid diagram based on the time-series frequency data of the monitoring stations includes:

[0018] The time-series frequency data of the virtual grid is determined by using a spatial interpolation algorithm, based on the time-series frequency data of the monitoring station and the electrical distance between the measured grid and the virtual grid, wherein the information source of the measured grid is the time-series frequency data of the monitoring station.

[0019] Optionally, before framing the time-series frequency data of each grid in chronological order, the method further includes:

[0020] The time-series frequency data of each grid in the grid diagram is mapped to a color spectrum, which includes several types of color information, and the color information is related to the magnitude of the time-series frequency data.

[0021] The temporal frequency data of each grid in the grid diagram are replaced with the color information in the color spectrum;

[0022] Correspondingly, the step of framing the time-series frequency data of each grid in chronological order includes:

[0023] The color information of each grid is framed in chronological order.

[0024] Optionally, before determining the frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and the transient temporal animation of the target power grid, the method further includes:

[0025] The typical operating mode of the target power grid was simulated using a time-domain simulation model, and several simulation results of the target power grid under different fault disturbances were obtained;

[0026] Record the time-series frequency data of the monitoring stations of the target power grid corresponding to the simulation results and the frequency stability of the target power grid;

[0027] Determine the transient time-series animation of the target power grid corresponding to the simulation results;

[0028] A three-dimensional convolutional neural network model is constructed based on the correspondence between the transient time-series animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results.

[0029] Optionally, the step of constructing a three-dimensional convolutional neural network model based on the correspondence between the transient time-series animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results includes:

[0030] The transient temporal animation of the target power grid is used as the input layer of a three-dimensional convolutional neural network model;

[0031] A hard kernel layer is constructed for a three-dimensional convolutional neural network model, wherein the hard kernel layer extracts image feature information from consecutive frames of transient temporal animation of the target power grid corresponding to the simulation results;

[0032] The convolutional layer of the three-dimensional convolutional neural network model is constructed based on the convolutional methods of three-dimensional convolution and two-dimensional convolution. The convolutional layer uses the convolutional methods of three-dimensional convolution and two-dimensional convolution to extract feature vectors from the image feature information.

[0033] The convolutional layer is connected to the output layer using a fully connected layer, and the output layer is constructed using an activation function. The output layer maps the feature vector to a probability value corresponding to the frequency stability of the target power grid based on the activation function.

[0034] To address the aforementioned technical problems, the present invention also provides a power grid frequency stability determination system, comprising:

[0035] A plan view construction unit is used to construct a plan view of the target power grid based on the geographic information of the target power grid;

[0036] The timing animation construction unit is used to construct a transient timing animation of the target power grid based on the timing frequency data of the monitoring stations of the target power grid and the plan view of the target power grid;

[0037] A stability assessment unit is used to determine the frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and transient temporal animation of the target power grid.

[0038] To address the aforementioned technical problems, the present invention also provides an electronic device, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor for implementing the steps of the power grid frequency stability determination method as described above.

[0041] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the power grid frequency stability determination method as described above.

[0042] This invention provides a method for determining the frequency stability of a power grid. It constructs a planar map of the target power grid and combines this map with time-series frequency data from monitoring stations to generate a transient time-series animation of the target power grid. This animation accurately and comprehensively reflects the actual operating status of the target power grid from both temporal and spatial perspectives. Then, a pre-constructed three-dimensional convolutional neural network (3D Convolutional Neural Network) model and the transient time-series animation are used to determine the frequency stability of the target power grid. The 3D Convolutional Neural Network model can determine the corresponding frequency stability of the target power grid based on the input transient time-series animation, directly assessing the frequency stability of the target power grid. By using readily available spatiotemporal information to replace the complex device modeling process, the workload is greatly reduced. The 3D Convolutional Neural Network model can simultaneously extract the temporal characteristics of the target power grid's frequency response and the spatial characteristics combined with geographical information through the transient time-series animation. Visualizing the input information of the 3D Convolutional Neural Network model through the construction of the transient time-series animation improves the transparency of the model. The 3D Convolutional Neural Network model ensures that key information from the input data is not omitted, improving the reliability and accuracy of the final frequency stability assessment results.

[0043] The present invention also provides a power grid frequency stability determination system, electronic device, and computer-readable storage medium, which have the same beneficial effects as the power grid frequency stability determination method described above. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the prior art and embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating a method for determining the frequency stability of a power grid provided by the present invention;

[0046] Figure 2 A flowchart illustrating another method for determining the frequency stability of a power grid provided by the present invention;

[0047] Figure 3 A schematic diagram of a plan view of a target power grid provided by the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of a three-dimensional convolutional neural network model provided by the present invention;

[0049] Figure 5 A schematic diagram of a power grid frequency stability determination system provided by the present invention;

[0050] Figure 6 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0051] The core of this invention is to provide a method, system, electronic device, and medium for determining the frequency stability of a power grid. It utilizes readily available spatiotemporal information to replace the complex device modeling process, greatly reducing workload. The three-dimensional convolutional neural network model can simultaneously extract the temporal characteristics of the target power grid's frequency response and the spatial characteristics combined with geographic information through transient time-series animation. By constructing transient time-series animation to visualize the input information of the three-dimensional convolutional neural network model, the transparency of the model is improved. The three-dimensional convolutional neural network model can ensure that key information in the input data is not omitted, thus improving the reliability and accuracy of the final frequency stability assessment results.

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Please refer to Figure 1 , Figure 1 A flowchart illustrating a method for determining the frequency stability of a power grid provided by this invention; please refer to... Figure 2 , Figure 2 This is a flowchart illustrating another method for determining the frequency stability of a power grid provided by the present invention. To solve the above-mentioned technical problems, the present invention provides a method for determining the frequency stability of a power grid, comprising:

[0054] S11: Construct a plan view of the target power grid based on its geographic information;

[0055] Considering that power systems, i.e., power grids, typically occupy a large area during the design process, and that the frequency characteristics of the grid at different locations may vary due to differences in geographical information, a plan view of the target power grid can be constructed first based on its geographical information. This plan view reflects the spatial characteristics of the target power grid, thus enabling the determination of the frequency characteristics at various locations within the target power grid. It should be noted that the target power grid in this application refers to a regional power grid, and the specific method of constructing the plan view of the target power grid is not specifically limited herein.

[0056] S12: Construct a transient time-series animation of the target power grid based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid;

[0057] It is easy to understand that frequency data at certain locations within the power grid can be detected in real time by setting up monitoring stations. Therefore, the timing frequency data of some locations within the target power grid can be directly obtained using the monitoring stations of the target power grid. The timing frequency data from the monitoring stations of the target power grid can then be combined with a planar map of the target power grid to construct a transient timing animation of the target power grid. This allows the transient timing animation to simultaneously reflect the temporal and spatial characteristics of the target power grid. This application does not impose any special limitations on the specific implementation method of the transient timing animation. The transient timing animation can be an interpolated frequency animation within 0.5 seconds after the fault occurs, i.e., 50 frames of interpolated frequency images after the fault occurs. The generation of the transient timing animation of the target power grid can be implemented in a Python environment. The monitoring stations can be directly implemented using the frequency detection locations of the target power grid itself, or monitoring stations can be set up in advance according to the actual operating conditions of the target power grid. This application does not impose any special limitations on the number of monitoring stations or their specific implementation methods.

[0058] S13: Determine the frequency stability of the target power grid using a pre-built 3D convolutional neural network model and transient temporal animation of the target power grid.

[0059] Specifically, before applying the power grid frequency stability determination method provided by this invention to the target power grid, a three-dimensional convolutional neural network model corresponding to the target power grid is first constructed. This model includes the correspondence between transient temporal animations and frequency stability assessment results. The model can directly output the corresponding frequency stability assessment results based on the input transient temporal animations, thereby achieving the determination of the frequency stability of the target power grid. Frequency stability is generally divided into two stability cases: frequency stability and frequency instability. Frequency stability means that after a fault disturbance occurs, the frequency values ​​of all nodes in the power system are not lower than 49.5Hz, and the power grid can remain stable after the fault disturbance without much control intervention. Conversely, if the frequency values ​​are lower than 49.5Hz, the power grid is considered frequency unstable, and in this case, the power grid cannot maintain stable operation after a fault disturbance, requiring additional stability control measures.

[0060] It is understandable that the time-domain simulation sample set P of the target power grid can be obtained through the time-series frequency data of the monitoring stations of the target power grid in step S12. This time-domain simulation sample set P is then input in batches into a pre-built three-dimensional convolutional neural network model for learning and training until the model converges. The resulting converged model can then be used as the final frequency stability assessment model. When it is necessary to assess the frequency stability of the target power grid, an online frequency response animation is constructed within the target power grid. Online measurement information of the frequency data from the monitoring stations of the target power grid is collected in real time. Based on the planar map of the target power grid constructed in step S11, spatial interpolation algorithms and other methods are used to generate the online frequency distribution animation of the target power grid, i.e., the transient time-series animation of the target power grid. When the target power grid encounters a large disturbance during operation, a time-series frequency animation is constructed using the frequency distribution images within T time sections of the power grid after the disturbance as input. The frequency distribution images are a time-series animation formed by combining the planar map and the time-series frequency data. The frequency distribution image at a certain moment at a discontinuity is used as a frame. The frequency distribution images of several time sections are combined to form a transient time-series animation of the target power grid. A pre-built three-dimensional convolutional neural network model is used as a frequency stability assessment model to evaluate and determine the frequency stability of the power grid. Finally, the three-dimensional convolutional neural network model outputs the evaluation result S' of the frequency stability state of the power grid. If the evaluation result S' = 0, it indicates that the power grid system is about to encounter the threat of frequency instability and frequency stability control measures need to be taken. If the evaluation result S' = 1, it indicates that the system frequency can still remain stable and no control measures are required.

[0061] Furthermore, after determining the frequency stability of the target power grid, the frequency stability state of the power grid can be determined by the subsequent operating state of the power grid. It can then be further determined whether the frequency stability evaluation result of the final 3D convolutional neural network model is consistent with the frequency stability state reflected by the actual operating state. If they are consistent, it indicates that the evaluation result of the 3D convolutional neural network model has high accuracy and can be widely applied. If they are inconsistent, it indicates that the 3D convolutional neural network model still has some errors. In this case, the frequency stability state reflected by the actual operating state can be fed back to the 3D convolutional neural network model as a training sample to further iterate the 3D convolutional neural network model until the evaluation result of the 3D convolutional neural network model has high accuracy, so that the 3D convolutional neural network model can be widely used to evaluate the frequency stability of the power grid in the future.

[0062] The frequency stability determination method for power grids provided by this invention is a power grid frequency stability assessment method based on a three-dimensional convolutional neural network model, belonging to the field of power system stability analysis and assessment. Based on time-series frequency data measured by a time-domain measurement system, it integrates actual geographic information to generate a three-dimensional transient time-series animation. This transient time-series animation, reflecting frequency changes, is used as a dataset. Transient frequency stability is assessed based on the three-dimensional convolutional neural network model. The transient time-series animation simultaneously extracts the temporal characteristics of the system frequency response and the spatial characteristics combined with geographic information. A spatiotemporal information machine learning model is used to implement the overall power grid frequency stability assessment method, fully extracting key regional transient frequency instability characteristics and patterns. The three-dimensional convolutional neural network model uses multiple three-dimensional convolutional layers to ensure that key information in the input data is not missed, thus improving the overall reliability of the frequency stability assessment results. Visualizing the input information through transient time-series animation enhances the transparency of machine learning. Utilizing readily available spatiotemporal information replaces the complex device modeling process, greatly reducing workload.

[0063] This invention first constructs a time-domain architecture model of the target power grid in transient time-domain simulation software. Then, it performs transient time-domain simulation of the power grid within the time-domain architecture model, using typical operating modes and fault disturbances, to obtain the time-series frequency response information of all monitoring stations and the transient frequency stability state of the entire power grid system, forming an initial sample set. Next, it constructs a planar map of the target power grid by building a two-dimensional uniform grid image on the xy-plane. The irregular geographical boundaries of the power grid are drawn using the latitude and longitude coordinates of the target power grid as the boundaries of the planar map. The locations of the monitoring stations are marked on the planar map based on their latitude and longitude coordinates. The grid containing the monitoring stations is used as the measured grid, and other grids in the planar map are used as virtual grids. Using the obtained time-series frequency data from the monitoring stations as the information source, the two-dimensional uniform grid image is processed... The virtual grid in the image is interpolated and filled, and the values ​​of the two-dimensional virtual grid outside the boundary of the planar image are set to empty, thereby generating a transient frequency distribution map of the target power grid. The transient frequency distribution maps of multiple frames are framed according to time to generate a transient time-series animation of the target power grid. The transient time-series animation and the corresponding transient frequency stability state of the power grid constitute the final frequency stability assessment sample set containing the time-series frequency animation. The frequency stability assessment sample set containing the time-series frequency animation is used as training data to construct a three-dimensional convolutional neural network model, making it a binary classification model with transient time-series animation as input and frequency stability assessment result as output. The three-dimensional convolutional neural network model is used to learn and train the frequency stability assessment sample set containing the time-series frequency animation until the model converges. Finally, the three-dimensional convolutional neural network model is used as the frequency stability assessment model of the target power grid. A transient timing animation generation module is constructed in the power grid system, using time-series frequency data from monitoring stations as the information source. When the power grid encounters fault disturbances during online operation, the measured transient timing animation is acquired in real time and input into a three-dimensional convolutional neural network model. The three-dimensional convolutional neural network model outputs the frequency stability assessment results of the system in real time, providing a reference for whether the current power grid system will experience frequency instability, thereby realizing real-time frequency visualization analysis and online assessment of frequency stability of the power grid.

[0064] This invention provides a method for determining the frequency stability of a power grid. It constructs a planar map of the target power grid and combines this map with time-series frequency data from monitoring stations to generate a transient time-series animation of the target power grid. This animation accurately and comprehensively reflects the actual operating status of the target power grid from both temporal and spatial perspectives. Then, a pre-constructed three-dimensional convolutional neural network (3D Convolutional Neural Network) model and the transient time-series animation are used to determine the frequency stability of the target power grid. The 3D Convolutional Neural Network model can determine the corresponding frequency stability of the target power grid based on the input transient time-series animation, directly assessing the frequency stability of the target power grid. By using readily available spatiotemporal information to replace the complex device modeling process, the workload is greatly reduced. The 3D Convolutional Neural Network model can simultaneously extract the temporal characteristics of the target power grid's frequency response and the spatial characteristics combined with geographical information through the transient time-series animation. Visualizing the input information of the 3D Convolutional Neural Network model through the construction of the transient time-series animation improves the transparency of the model. The 3D Convolutional Neural Network model ensures that key information from the input data is not omitted, improving the reliability and accuracy of the final frequency stability assessment results.

[0065] Based on the above embodiments:

[0066] Please refer to Figure 3 , Figure 3 A schematic diagram of a plan view of a target power grid provided by the present invention; as an optional embodiment, constructing a plan view of the target power grid based on the geographical information of the target power grid includes:

[0067] Obtain the geographic information of the target power grid and construct a rectangular image based on the geographic information. The length of the rectangular image is greater than the maximum longitude span and the width is greater than the maximum latitude span of the target power grid.

[0068] Divide the rectangular image evenly into several grids;

[0069] The grid containing the monitoring stations of the target power grid is defined as the measured grid, and the grid that does not contain the monitoring stations is defined as the virtual grid. This results in a grid map with longitude as the horizontal axis and latitude as the vertical axis. The grid map is a plan view of the target power grid.

[0070] It's easy to understand that a planar map of the target power grid can be created by constructing a two-dimensional grid map of the target power grid. Geographically, by combining the boundary information of the target power grid's geographical area, an irregular polygonal transmission and distribution network diagram bounded by latitude and longitude coordinates is constructed. Using the x-axis as longitude and the y-axis as latitude, the latitude and longitude coordinates corresponding to monitoring stations with measured data are marked on this plane; that is, the locations of the monitoring stations are marked on the planar map. Figure 3The coordinate points shown are used to construct a rectangular image in the xy plane. The length of the rectangular image exceeds the maximum longitude span of the target power grid geographical area, and the width of the rectangular image exceeds the maximum latitude span of the target power grid geographical area. The rectangular image is uniformly divided into l×l uniform rectangular grids. The grid where the coordinates of several monitoring stations are located is divided into the measured grid. The remaining grids in the planar map that do not contain monitoring stations are divided into virtual grids. Then the source of the frequency data in the measured grid can be directly determined based on the frequency detection situation of the monitoring stations.

[0071] Specifically, a two-dimensional grid map can effectively realize the construction process of the target power grid's plan view. The grid map corresponds to the geographical latitude and longitude, which can accurately reflect the location information and spatial characteristics of the target power grid, further improving the construction process of the target power grid's plan view and facilitating the simple implementation of the method for determining the frequency stability of the entire power grid.

[0072] As an optional embodiment, a transient time-series animation of the target power grid is constructed based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid, including:

[0073] Acquire time-series frequency data from monitoring stations of the target power grid;

[0074] The time-series frequency data of each grid in the grid diagram are determined based on the time-series frequency data of the monitoring stations;

[0075] The time-series frequency data of each grid are framed in chronological order to obtain the transient time-series animation of the target power grid.

[0076] Understandably, after obtaining the time-series frequency data of the monitoring stations of the target power grid, the time-series frequency data of the measured grid can be directly obtained. The time-series frequency data of the virtual grid can also be calculated based on the time-series frequency data of the measured grid, thereby determining the time-series frequency data of each grid in the planar map. At this time, the three-dimensional frequency distribution map combining the time-series frequency data and the planar map can be represented by a transient time-series animation. Each moment or each time section will correspond to a planar frequency distribution map. After framing the two-dimensional planar frequency distribution maps of each moment in chronological order, a three-dimensional transient time-series animation can be formed, dynamically reflecting the frequency distribution of the target power grid in combination with spatial characteristics within a time period.

[0077] Specifically, because transient timing animation incorporates a planar map, it can comprehensively observe the transient frequency instability dynamic characteristics of the power grid from a spatial perspective. It also incorporates time-series frequency data, allowing the transient timing animation to fully integrate the regional node geographic information of the power grid and the corresponding time-series frequency data of the nodes. In geographic space, a grid map is drawn by combining regional boundary information, and measured frequency data of the measured grid is incorporated to obtain an intuitive 2.5-dimensional frequency animation frame, so as to finally generate a transient timing animation that comprehensively reflects the frequency situation of the target power grid.

[0078] As an optional embodiment, determining the time-series frequency data of each grid in the grid diagram based on the time-series frequency data of the monitoring stations includes:

[0079] The time-series frequency data of the virtual grid is determined by using a spatial interpolation algorithm, based on the time-series frequency data of the monitoring stations and the electrical distance between the measured grid and the virtual grid. The information source of the measured grid is the time-series frequency data of the monitoring stations.

[0080] Understandably, interpolation algorithms can be used to calculate and fill in the virtual grids that lack measured frequency data by utilizing the measured frequency values ​​in all measured grids. This can be achieved using spatial interpolation algorithms or other types of interpolation calculation methods. Spatial interpolation algorithms use the actual data of multiple measured grids on the image to extrapolate the frequency values ​​in the surrounding empty virtual grids that lack measured frequency data, and the extrapolated frequency values ​​are related to the distance between the virtual grids and the measured grids.

[0081] Specifically, in order to integrate the spatial information of the target power grid into the transient frequency response information, which is also the time-series frequency data, the time-domain simulation sample X obtained after performing multiple time-domain simulations on the target power grid is used. i The data in (f1f2...f m ) t As the information source for the measured grid, the time-domain simulation sample X i The data in this diagram are frequency data detected by monitoring stations of the target power grid. Then, the frequency values ​​of all other virtual grids in the planar diagram are estimated by interpolation based on the electrical distance information between the virtual grid and the measured grid. At the same frequency level, the electrical distance information can be equivalently represented by Euclidean distance in the xy-plane. Taking time t as an example, the frequency interpolation f of the virtual grid O(x,y) is... t The calculation process for (x,y) is as follows: Where m is the total number of measured grid cells, f it Let D be the measured frequency value of the corresponding node in the i-th measured grid at time t. Oi D represents the Euclidean distance between the virtual grid O and the i-th measured grid. OjThe Euclidean distance between the virtual grid O and the j-th measured grid can be considered as the distance weighting coefficient in the interpolation calculation. i ,y i Let (x) be the latitude and longitude coordinates of the i-th measured grid in an orthogonal rectangular coordinate system. j ,y j Let f be the latitude and longitude coordinates of the j-th measured grid in an orthogonal rectangular coordinate system. The virtual frequency value f of all virtual grids at time t is filled in using the interpolation method described above. t After completing the frequency filling of the virtual grid, the geographical boundary of the target power grid is used as the dividing line. The grid values ​​outside the boundary are set to empty, and only the frequency values ​​on the planar map inside the boundary are retained, thus obtaining a frequency distribution map reflecting the various locations of the target power grid.

[0082] Specifically, the frequency data in the virtual grid can be effectively calculated using spatial interpolation algorithms, thereby determining the frequency information corresponding to each location in the target power grid, ensuring the integrity of the frequency information in the transient time-series animation, and facilitating a comprehensive reflection and analysis of the frequency situation of the entire target power grid.

[0083] As an optional embodiment, before framing the time-series frequency data of each grid in chronological order, the method further includes:

[0084] The time-series frequency data of each grid in the grid diagram is mapped to a color spectrum, which includes several color information, and the color information is related to the magnitude of the time-series frequency data.

[0085] Replace the temporal frequency data of each grid in the grid diagram with color information from the color spectrum;

[0086] Correspondingly, the time-series frequency data of each grid are framed in chronological order, including:

[0087] The color information of each grid is framed in chronological order.

[0088] It's easy to understand that, for easier differentiation and observation, the frequency values ​​of the grid within the entire planar diagram can be mapped onto a color spectrum to obtain a color-based frequency distribution map at time t. Different color spectra can be set according to the magnitude of the frequency data to distinguish different frequency data. For example, the color spectrum corresponding to the lowest frequency value can be set to blue, the color spectrum corresponding to the middle frequency value to yellow, and the color spectrum corresponding to the highest frequency value to red. Framing the frequency distribution maps of T time segments in chronological order constitutes a transient time-series animation sample of the target power grid. This transient time-series animation is denoted as g. g, along with the corresponding system frequency steady state S obtained and recorded from time-domain simulation, constitutes a new system time-domain simulation sample P. i Pi ={(g i ,S i )}, 1≤i≤N, N P i The final time-domain simulation sample set P is constructed for training the 3D convolutional neural network model. This application does not impose specific limitations on the specific type and implementation of the color spectrum, as long as the color spectrum can reflect the high and low levels of frequency values ​​at the same time and constitute animation samples. Frequency values ​​can also be distinguished by setting the color of the spectrum corresponding to frequency intervals. The color information of the spectrum corresponding to lower frequency data below 49.5Hz is set to black, the color information of the spectrum corresponding to frequency data between 49.5Hz and 50.5Hz is set to white, and the color information of the spectrum corresponding to higher frequency data above 50.5Hz is set to red. The correspondence between frequency data and color spectrum can be set according to the frequency fluctuation of the target power grid, and this application does not impose specific limitations here.

[0089] Specifically, color maps can be used to more intuitively represent the frequency fluctuations of each grid in the target power grid, making transient time-series animations easier to observe. The color changes can directly and effectively reflect the frequency fluctuations at each location in the target power grid, further improving the generation process of transient time-series animations.

[0090] As an optional embodiment, before determining the frequency stability of the target power grid using a pre-built 3D convolutional neural network model and transient temporal animation of the target power grid, the method further includes:

[0091] The typical operating mode of the target power grid was simulated using a time-domain simulation model, and several simulation results of the target power grid under different fault disturbances were obtained.

[0092] Record the time-series frequency data of the monitoring stations of the target power grid and the frequency stability of the target power grid corresponding to the simulation results;

[0093] Determine the transient time-series animation of the target power grid corresponding to the simulation results;

[0094] A three-dimensional convolutional neural network model is constructed based on the correspondence between the transient time-series animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results.

[0095] It is not difficult to understand that we can first determine the monitoring stations in the target power grid used for frequency stability assessment, and denote the number of monitoring stations as m. In the electromechanical transient time-domain simulation software, we build a time-domain simulation model of the target power grid based on its grid structure. In the time-domain simulation model of the target power grid, we perform N time-domain simulations on the typical operating modes and possible fault disturbances of the target power grid, where N = number of typical operating modes * number of fault disturbances. We obtain the results of N time-domain simulations. After each time-domain simulation, we collect the time-series frequency data of m monitoring substations after the disturbance. The time-series frequency data of each monitoring station can be unified into time-series frequency data f collected at T time sections with a sampling interval of d. At the same time, we also need to record the frequency stability state S of the power grid after each time-domain simulation, with S = 0 indicating frequency instability and S = 1 indicating system frequency stability. The time-series frequency data f of the monitoring stations obtained after each time-domain simulation and the corresponding frequency stability state S together constitute a time-domain simulation sample X. i X i ={{(f1f2...f m ) t i ,S i}, 1≤t≤T}, 1≤i≤N, N time-domain simulation samples X i The time-domain simulation sample set X constitutes the monitoring station.

[0096] It should be noted that this application does not impose specific limitations on the construction of the time-domain simulation module or the specific implementation method of the time-domain simulation process. The time-domain simulation calculation process can be performed using PSASP software, and the parameter values ​​in the simulation process can be determined according to the actual situation of the target power grid. For example, d can be 0.01s and T can be 50. Typical operating modes mainly refer to the typical operating modes of the power grid system, such as the operating modes of the power grid in winter and summer. Power grid systems at different times and locations will have their own typical operating modes. Possible fault disturbances may include: UHVDC blocking, UHVAC disconnection, generator tripping, or sudden increases in regional load, all of which will interfere with the frequency stability of the power grid system.

[0097] Specifically, the time-domain simulation method is used to simulate some fault conditions that may occur in the real power grid and interfere with the frequency stability of the power system. Through these simulation processes, partial node information of the power grid system is obtained so that the three-dimensional convolutional neural network model, a machine learning model, can learn from the simulation samples. This process can be repeated many times to generate a large amount of data for training the three-dimensional convolutional neural network model, so as to ensure the accurate application of the three-dimensional convolutional neural network model in the future.

[0098] Please refer to Figure 4 , Figure 4The diagram below illustrates the structure of a three-dimensional convolutional neural network model provided by this invention. As an optional embodiment, a three-dimensional convolutional neural network model is constructed based on the correspondence between the transient temporal animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results, including:

[0099] The transient temporal animation of the target power grid is used as the input layer of a three-dimensional convolutional neural network model;

[0100] A hard kernel layer is constructed for a three-dimensional convolutional neural network model. The hard kernel layer extracts image feature information from consecutive frames of transient time-series animation of the target power grid corresponding to the simulation results.

[0101] The convolutional layer of the three-dimensional convolutional neural network model is constructed based on the convolutional methods of three-dimensional convolution and two-dimensional convolution. The convolutional layer uses the convolutional methods of three-dimensional convolution and two-dimensional convolution to extract feature information from image features and obtain feature vectors.

[0102] The convolutional layer is connected to the output layer using a fully connected layer, and the output layer is constructed using an activation function. The output layer maps the feature vector to a probability value corresponding to the frequency stability of the target power grid based on the activation function.

[0103] It's easy to understand that the structure of a 3D convolutional neural network model can contain a total of nine layers. The structural order from input to output is: input layer, hard kernel layer, 3D convolutional layer, average pooling layer, 3D convolutional layer, average pooling layer, 2D convolutional layer, fully connected layer, and output layer. The input layer is responsible for data input; in this 3D convolutional neural network model, the input is the transient temporal animation g of the target power grid, which is a temporal image composed of l×l grids across T frames. The hard kernel layer is responsible for processing grayscale, x-axis gradient, y-axis gradient, x-axis optical flow, and y-axis optical flow on consecutive frames. Feature extraction of temporal animation: A three-dimensional convolutional learning module is formed by connecting a three-dimensional convolutional layer and an average pooling layer to further extract the feature information of the temporal animation. After two or more three-dimensional convolutions, the temporal dimension of the sample has been compressed to the point that it is impossible to perform three-dimensional convolution again. A two-dimensional convolutional layer can be used for final feature extraction. After feature extraction, a fully connected layer is used to connect the finally extracted feature vector to the output layer. An activation function is set for the output layer. The activation function is used to make the final three-dimensional convolutional neural network model output the prediction result of the frequency stability state S of the target power grid.

[0104] Understandably, the hard core layer performs initial image processing, while subsequent 3D convolutional layers, average pooling layers, and 2D convolutional layers perform deeper feature extraction. The resulting feature vector is a string of data within the model obtained after multiple feature extractions for a given input sample. This vector represents the inherent feature information of this transient temporal animation and can be provided to the output layer for judgment. The input to the 3D convolutional neural network model is the transient temporal animation g, and the output is S (frequency stable state, 0 or 1). The sample set P of the 3D convolutional neural network model contains pairs of gi and Si. After performing time-domain simulation for each set fault, a gi and a stable state Si are generated. This sample set is input into the 3D convolutional neural network model in pairs for learning and training. In subsequent applications, once a fault occurs, by generating and inputting the transient temporal animation of the target power grid into the 3D convolutional neural network model, the frequency stability assessment result of whether the target power grid can maintain the frequency stability of the system after a certain fault disturbance can be directly output.

[0105] It should be noted that the activation function is a function that runs on the neurons of an artificial neural network, responsible for mapping the neuron's input to its output. It can be implemented using the sigmoid function, which maps a series of feature vectors to a probability value to determine whether a sample belongs to 0 (instability) or 1 (stability). This application does not impose specific restrictions on the specific construction and implementation process of the 3D convolutional neural network model; it can be built and implemented within the TensorFlow framework.

[0106] This invention employs continuous frequency animation frames formed by fully integrating spatiotemporal information as input to a 3D convolutional neural network model. Combining 3D and 2D convolution, it accurately extracts preliminary features from the animation frames, including grayscale, gradient, and optical flow, thereby forming feature vectors to capture the transient frequency stability information of the target power grid. This constructs a complete, reliable, and intuitive transient frequency stability assessment method. The data samples are based on a grid network spatial information segmented according to the geospatial division of the power grid, incorporating actual frequency measurement data. A visualized 3D transient frequency animation serves as the standard container to achieve wide-area spatiotemporal information fusion, expanding the information content of the dataset samples. The adopted 3D convolutional neural network model can simultaneously consider the temporal and spatial dimensions of the wide-area spatiotemporal information data, effectively extracting dynamic features and better understanding the relationships and trends between data. It performs excellently in capturing time-series information, thus improving the accuracy of frequency stability assessment results. Based on the 3D convolutional neural network model, high-order features are repeatedly extracted from the input transient time-series animation, accurately and quickly forming feature vectors. Based on these feature vectors, the frequency stability of the system after disturbances is quickly output, effectively achieving the assessment of the frequency stability of the target power grid.

[0107] Please refer to Figure 5 , Figure 5 The present invention provides a schematic diagram of a power grid frequency stability determination system; to solve the above-mentioned technical problems, the present invention also provides a power grid frequency stability determination system, comprising:

[0108] The plan view construction unit 11 is used to construct a plan view of the target power grid based on the geographic information of the target power grid;

[0109] The timing animation construction unit 12 is used to construct the transient timing animation of the target power grid based on the timing frequency data of the monitoring stations of the target power grid and the plan view of the target power grid;

[0110] The stability assessment unit 13 is used to determine the frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and transient time-series animation of the target power grid.

[0111] For an introduction to the frequency stability determination system for a power grid provided by this invention, please refer to the embodiments of the power grid frequency stability determination method described above. This invention will not be repeated here.

[0112] Please refer to Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device provided by the present invention. To solve the above-mentioned technical problems, the present invention also provides an electronic device, comprising:

[0113] Memory 21 is used to store computer programs;

[0114] Processor 22 is used to implement the steps of the power grid frequency stability determination method as described above.

[0115] The processor 22 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 22 may be implemented using at least one hardware form selected from DSP (Digital Signal Processor), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 22 may also include a main processor and a coprocessor. The main processor, also known as the central processing unit, is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 22 may integrate a GPU (graphics processing unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 22 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0116] The memory 21 may include one or more computer-readable storage media, which may be non-transitory. The memory 21 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 21 is used to store at least the following computer program, which, after being loaded and executed by the processor 22, is capable of implementing the relevant steps of the power grid frequency stability determination method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 21 may also include an operating system and data, and the storage method may be temporary or permanent storage. The operating system may include Windows, Unix, Linux, etc. The data may include, but is not limited to, data related to the power grid frequency stability determination method.

[0117] In some embodiments, the electronic device may further include a display screen, input / output interfaces, communication interfaces, a power supply, and a communication bus.

[0118] It will be understood by those skilled in the art that Figure 6 The structures shown do not constitute a limitation on electronic devices and may include more or fewer components than those shown.

[0119] For an introduction to the electronic device provided by this invention, please refer to the embodiments of the above-described method for determining the frequency stability of the power grid; the present invention will not be described in detail here.

[0120] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the aforementioned method for determining the frequency stability of a power grid.

[0121] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. Specifically, the computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, and portable hard drives, or any type of media or device suitable for storing instructions or data, etc., and this application does not make any special limitations here.

[0122] For an introduction to the computer-readable storage medium provided by the present invention, please refer to the embodiments of the above-described method for determining the frequency stability of the power grid; the present invention will not be repeated here.

[0123] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0124] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0125] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of determining frequency stability of an electrical grid, characterized by, include: A plan view of the target power grid is constructed based on the geographic information of the target power grid; A transient time-series animation of the target power grid is constructed based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid; The frequency stability of the target power grid is determined by using a pre-built three-dimensional convolutional neural network model and the transient temporal animation of the target power grid; the three-dimensional convolutional neural network model directly outputs the corresponding frequency stability evaluation result based on the input transient temporal animation. Before determining the frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and the transient temporal animation of the target power grid, the method further includes: The typical operating mode of the target power grid was simulated using a time-domain simulation model, and several simulation results of the target power grid under different fault disturbances were obtained; Record the time-series frequency data of the monitoring stations of the target power grid corresponding to the simulation results and the frequency stability of the target power grid; Determine the transient time-series animation of the target power grid corresponding to the simulation results; A three-dimensional convolutional neural network model is constructed based on the correspondence between the transient time-series animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results. The construction of a three-dimensional convolutional neural network model based on the correspondence between the transient temporal animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results includes: The transient temporal animation of the target power grid is used as the input layer of a three-dimensional convolutional neural network model; A hard kernel layer is constructed for a three-dimensional convolutional neural network model, wherein the hard kernel layer extracts image feature information from consecutive frames of transient temporal animation of the target power grid corresponding to the simulation results; The convolutional layer of the three-dimensional convolutional neural network model is constructed based on convolutional methods of three-dimensional convolution and two-dimensional convolution. The convolutional layer uses convolutional methods of three-dimensional convolution and two-dimensional convolution to extract features from the image feature information to obtain feature vectors. After the image feature information undergoes multiple three-dimensional convolutions, it is further extracted using a two-dimensional convolutional layer. The convolutional layer is connected to the output layer using a fully connected layer, and the output layer is constructed using an activation function. The output layer maps the feature vector to a probability value corresponding to the frequency stability of the target power grid based on the activation function. Based on the probability value, it outputs 0 or 1, where 0 represents instability of the target power grid and 1 represents stability of the target power grid.

2. The method of determining the frequency stability of an electrical grid according to claim 1, wherein, The construction of the plan view of the target power grid based on the geographic information of the target power grid includes: Geographic information of the target power grid is obtained, and a rectangular image is constructed based on the geographic information, wherein the length of the rectangular image is greater than the maximum longitude span and the width is greater than the maximum latitude span of the target power grid; The rectangular image is evenly divided into several grids; The grid containing the monitoring stations of the target power grid is defined as the measured grid, and the grid that does not contain the monitoring stations is defined as the virtual grid, resulting in a grid map with longitude as the horizontal axis and latitude as the vertical axis. The grid map is a plan view of the target power grid.

3. The method of determining the frequency stability of an electrical grid according to claim 2, wherein, The step of constructing a transient time-series animation of the target power grid based on the time-series frequency data of the monitoring stations of the target power grid and the plan view of the target power grid includes: Acquire time-series frequency data from the monitoring stations of the target power grid; The time-series frequency data of each grid in the grid diagram are determined based on the time-series frequency data of the monitoring stations. The time-series frequency data of each grid are framed in chronological order to obtain the transient time-series animation of the target power grid.

4. The method of determining the frequency stability of an electrical grid according to claim 3, wherein, The step of determining the time-series frequency data of each grid in the grid diagram based on the time-series frequency data of the monitoring stations includes: The time-series frequency data of the virtual grid is determined by using a spatial interpolation algorithm, based on the time-series frequency data of the monitoring station and the electrical distance between the measured grid and the virtual grid, wherein the information source of the measured grid is the time-series frequency data of the monitoring station.

5. The method of determining the frequency stability of an electrical grid according to claim 3, wherein, Before framing the time-series frequency data of each grid in chronological order, the method further includes: The time-series frequency data of each grid in the grid diagram is mapped to a color spectrum, which includes several types of color information, and the color information is related to the magnitude of the time-series frequency data. The temporal frequency data of each grid in the grid diagram are replaced with the color information in the color spectrum; Correspondingly, the step of framing the time-series frequency data of each grid in chronological order includes: The color information of each grid is framed in chronological order.

6. A system for determining the frequency stability of an electrical grid, characterized by include: A plan view construction unit is used to construct a plan view of the target power grid based on the geographic information of the target power grid; The timing animation construction unit is used to construct a transient timing animation of the target power grid based on the timing frequency data of the monitoring stations of the target power grid and the plan view of the target power grid; The stability assessment unit is used to determine the current frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and the transient temporal animation of the target power grid; the three-dimensional convolutional neural network model directly outputs the corresponding frequency stability assessment result based on the input transient temporal animation; Before determining the frequency stability of the target power grid using a pre-built three-dimensional convolutional neural network model and transient temporal animation of the target power grid, the power grid frequency stability determination system is further used for: The typical operating mode of the target power grid was simulated using a time-domain simulation model, and several simulation results of the target power grid under different fault disturbances were obtained; Record the time-series frequency data of the monitoring stations of the target power grid corresponding to the simulation results and the frequency stability of the target power grid; Determine the transient time-series animation of the target power grid corresponding to the simulation results; A three-dimensional convolutional neural network model is constructed based on the correspondence between the transient time-series animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results. The construction of a three-dimensional convolutional neural network model based on the correspondence between the transient temporal animation of the target power grid corresponding to the simulation results and the frequency stability of the target power grid corresponding to the simulation results includes: The transient temporal animation of the target power grid is used as the input layer of a three-dimensional convolutional neural network model; A hard kernel layer is constructed for a three-dimensional convolutional neural network model, wherein the hard kernel layer extracts image feature information from consecutive frames of transient temporal animation of the target power grid corresponding to the simulation results; The convolutional layer of the three-dimensional convolutional neural network model is constructed based on convolutional methods of three-dimensional convolution and two-dimensional convolution. The convolutional layer uses convolutional methods of three-dimensional convolution and two-dimensional convolution to extract features from the image feature information to obtain feature vectors. After the image feature information undergoes multiple three-dimensional convolutions, it is further extracted using a two-dimensional convolutional layer. The convolutional layer is connected to the output layer using a fully connected layer, and the output layer is constructed using an activation function. The output layer maps the feature vector to a probability value corresponding to the frequency stability of the target power grid based on the activation function. Based on the probability value, it outputs 0 or 1, where 0 represents instability of the target power grid and 1 represents stability of the target power grid.

7. An electronic device, comprising: include: Memory, used to store computer programs; A processor for implementing the steps of the method for determining the frequency stability of a power grid as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the power grid frequency stability determination method as described in any one of claims 1 to 5.

Images