A transient voltage stability assessment method based on unbalanced residual information learning

Abstract

The present invention discloses a transient voltage stability assessment method based on unbalanced residual information learning, comprising the following steps: S1. building a target power system simulation model corresponding to the target power system; S2. performing transient voltage time-domain simulation on the target power system simulation model, and constructing a transient time-domain simulation sample set; S3. performing sample category weight calculation on the transient time-domain simulation sample set, and constructing a Focal‑Loss loss function for processing sample category imbalance; S4. constructing a residual spatiotemporal graph convolutional neural network model; S5. training the residual spatiotemporal graph convolutional neural network model until a preset performance evaluation index threshold of spatiotemporal feature learning is met, and obtaining a trained residual spatiotemporal graph convolutional neural network model; S6. performing transient voltage stability assessment on the power system through the trained residual spatiotemporal graph convolutional neural network model.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly relates to a transient voltage stability assessment method based on unbalanced residual information learning. Background Art

[0002] With the large-scale integration of new energy into the grid through direct current, the main power network will form a pattern of multiple conventional / flexible direct current complex hybrid connections. Considering the above factors comprehensively, these trends will make the network structure and operating characteristics of traditional power systems increasingly complex. When the system suffers from large transient disturbances, the fast response characteristics and complex dynamic interactions of DC converter stations and a large number of dynamic loads are likely to cause insufficient transient voltage support. In a large-scale power grid with high penetration of new energy, large-scale complex hybrid AC / DC connections, and massive heterogeneous load access, if the transient voltage stability cannot be reliably perceived in real time and control measures cannot be taken in time, voltage instability or even collapse accidents are likely to occur, posing a serious threat to the safe consumption of new energy and the stable supply of electric power.

[0003] Due to the high-dimensional, time-varying, and strongly non-linear characteristics of the actual power system transient voltage stability problem itself, the research on its assessment methods has always been a hot topic of concern. In recent years, many domestic and foreign scholars have proposed power grid transient voltage stability assessment methods from the perspectives of model-driven or data-driven. Among them, the model-driven method has the problems of complex model construction and large computational amount, and it is difficult to meet the real-time assessment requirements of complex power grid structures and changing operating conditions. Although the existing data-driven methods can process large-scale data, most of them only rely on single-time series data or isolated time series information of different monitoring nodes, lacking the excavation of the spatial and temporal coupling correlation in the time series evolution process of each monitoring node of the power grid after a fault occurs, resulting in difficulty in comprehensively capturing the inherent spatio-temporal characteristics of power grid transient voltage instability. In addition, the probability of instability in the actual power system is very small, resulting in the number of stable samples being greater than the number of unstable samples. If this severely unbalanced data in terms of category numbers is put into the data-driven method for learning, it will surely lead to insufficient attention to unstable samples, thus affecting the transient voltage stability assessment performance of the data-driven method. Summary of the Invention

[0004] In view of this, the present invention provides a transient voltage stability assessment method based on unbalanced residual information learning, which is used to at least solve the problems of limitations in the technology for excavating transient voltage stability category imbalance and spatio-temporal characteristics during the transient stability assessment of power systems in the prior art.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A transient voltage stability assessment method based on unbalanced residual information learning, comprising the following steps:

[0007] S1. Build a target power system simulation model corresponding to the target power system;

[0008] S2. Conduct a transient voltage time-domain simulation on the target power system simulation model to obtain M time-domain simulation results, and construct a spatio-temporal diagram model corresponding to the target power system simulation model under each simulation condition respectively according to the spatio-temporal diagram construction method G , and at the same time, for M time-domain simulation results, respectively calibrate the voltage stability or instability of the system state S , and based on M calibration results and the spatio-temporal diagram model G construct a transient time-domain simulation sample set, denoted as ;

[0009] S3. Construct a Focal-Loss loss function, and obtain a Focal-Loss loss function for training the residual spatio-temporal graph convolutional neural network model to learn the sample class imbalance by calculating the stable and unstable class weight factors of the transient time-domain simulation sample set and setting an adjustment factor for sample classification;

[0010] S4. Construct a residual spatio-temporal graph convolutional neural network model, and conduct spatio-temporal feature learning on the transient time-domain simulation sample set through the residual spatio-temporal graph convolutional neural network model to obtain the prediction result of the transient voltage state;

[0011] S5. Judge whether the prediction result is correct according to the actual transient voltage state result of the target power system. If it is incorrect, conduct the next round of training on the residual spatio-temporal graph convolutional neural network model until the performance evaluation index threshold of the preset spatio-temporal feature learning is satisfied, and obtain the trained residual spatio-temporal graph convolutional neural network model;

[0012] S6. Conduct power system transient voltage stability assessment through the trained residual spatio-temporal graph convolutional neural network model.

[0013] Preferably, the specific content of the spatio-temporal diagram construction method in S2 includes:

[0014] According to the spatio-temporal diagram construction method, construct the graph model corresponding to the physical topology structure of the target power system under the k th time-domain simulation sample observation time window length of , and obtain the final spatio-temporal diagram model G ; ;

[0015] Represent the th monitoring node in the target power system under the observation time window as the vertex in the graph model , denoted asX And , N Let \(N\) be the number of monitoring nodes. The connection relationship between the \(i\)-th monitoring node and the \(j\)-th monitoring node is used as an edge in the graph model ,and all such edges are represented as a topological adjacency matrix, denoted as And ,where the length of the \(k\)-th sample observation time window is A And ,where the graph model corresponding to the physical topology of the target power system at time k is ; ;

[0016] At any monitoring node ,if there is a connection, it is denoted as ,otherwise it is denoted as ;

[0017] Finally, a final spatio-temporal graph model is constructed to characterize the dynamic characteristics of the transient response data of the target power system in the spatio-temporal dimension.

[0018] Preferably, the specific content of S2 includes:

[0019] Perform a transient voltage time-domain simulation on the target power system simulation model, and combine the system's historical operation mode and fault settings to simulate the voltage stability of the power grid under various typical operating conditions, obtaining a total number of times of time-domain simulation results;

[0020] After each time-domain simulation ends, record the physical adjacency matrix of the target power system and the time-series evolution characteristics of the monitoring nodes corresponding to the time window length of after the fault occurs, including the voltage amplitude ,active power and reactive power at the time window length of ,and calibrate the system state of the simulation result according to the transient voltage stability engineering criterion adopted by the actual power system, where indicates that the power system state result of this time-domain transient simulation is voltage instability, and

[0021] Based on M calibration results and the spatio-temporal graph model G construct an input sample set denoted as ,where 。

[0022] Preferably, M It is the product of the typical operating level numbers of the generator, new energy output, and load consumption, the proportion of transient operating dynamic load, and the number of typical fault disturbances.

[0023] Preferably, the Focal-Loss loss function in S3 specifically includes: calculating the stable and unstable class weight factors of the transient time-domain simulation sample set and setting the adjustment factor for sample classification ;

[0024] Class weight factor It is set according to the quantity ratio of positive samples and negative samples:

[0025] Regarding the unstable samples in the transient time-domain simulation sample set as positive samples and the stable samples in the transient time-domain simulation sample set as negative samples;

[0026] Assume the number of positive samples is c , the number of negative samples is d , and the total number of samples is M = c + d , then for the weight ratio of unstable samples , and for the weight ratio of stable samples ;

[0027] Adjustment factor It is used to control the degree of loss weight reduction for sample classification and simultaneously enhance the attention of the residual spatio-temporal graph convolutional network model to difficult-to-evaluate samples:

[0028] ;

[0029] Based on the settings of the class weight factor and the adjustment factor , the Focal-Loss loss function is:

[0030] ;

[0031] Among them, is the prediction probability of the residual spatio-temporal graph convolutional network model for the sample. For unstable samples , and for stable samples .

[0032] Preferably, the residual spatio-temporal graph convolutional neural network model in S4 specifically includes: a residual spatio-temporal graph convolutional module and a fully connected classification module;

[0033] The residual spatio-temporal graph convolution module includes a spatial graph convolution layer, a temporal convolution layer, and a residual convolution layer;

[0034] The spatial graph convolution layer is used to obtain any spatio-temporal graph input sample in the input sample set and perform spatial feature mining on the power system topological adjacency matrix and realize the feature aggregation of the temporal information of neighbor nodes on the power system topological graph to the central node, and output spatial features ;

[0035] The temporal convolution layer is used to further perform convolution operations on to realize the spatio-temporal feature extraction of transient voltage response data, and output temporal features ;

[0036] The residual convolution layer is used to perform convolution operations on and output residual features and keep the same dimension as the spatio-temporal graph output features ;

[0037] The fully connected classification module is used to fuse the residual features and spatio-temporal features and reduce the dimension of the fused features through a linear layer, further extract high-dimensional features for classifying transient voltage stability and instability, and convert the high-dimensional features into a probability distribution through the softmax activation function to complete the mapping of the short-term transient voltage stability state space, realizing the evaluation of system transient voltage stability or instability.

[0038] Preferably, the specific content of the spatial graph convolution layer outputting spatial features includes:

[0039] Perform degree matrix conversion on the target power system topological adjacency matrix , and the conversion method is:

[0040] ;

[0041] In the formula, is the element in the th row and th column of the degree matrix;

[0042] Based on the identity matrix and the degree matrix , further perform spatial feature mining on the power system topological structure matrix using the Laplacian graph transformation formula, where the Laplacian graph transformation formula is:

[0043] ;

[0044] Among them, is the identity matrix; the degree matrix is used to represent the degree of each vertex in the graph model, and its degree is the number of connections between each node and other nodes in the power grid adjacency matrix; for the target power system topology adjacency matrix perform spatial feature mining, based on the solution formula to obtain the orthonormal eigenvectors and the spatial eigenvalue matrix of the target power system topology adjacency matrix , and use the graph node information aggregation formula to realize the feature aggregation of the time series information of neighbor nodes on the power system topology graph to the central node. Among them, the graph node information aggregation formula is:

[0045] ;

[0046] In the formula, is the graph convolution kernel, is the time series evolution feature of the time window length of the graph matrix node observation.

[0047] Preferably, the specific content of the output time series feature of the time series convolution layer includes:

[0048] ;

[0049] Among them, ReLU is a non-linear activation function, is the convolution kernel parameter, is the convolution operation.

[0050] Preferably, the fully connected classification module reduces the dimension of the fused features through a linear layer, further extracts high-dimensional features for classifying transient voltage stability and instability, and converts the high-dimensional features into a probability distribution through the fully connected classification formula, completes the mapping of the short-term transient voltage stability state space, and realizes the evaluation of the system transient voltage stability state. Among them, the fully connected classification formula is:

[0051] .

[0052] Preferably, the specific content of S5 includes:

[0053] Select the accuracy rate and the misjudgment rate as the performance evaluation indicators for the spatio-temporal feature learning of the residual spatio-temporal graph convolutional neural network model. Among them, the accuracy rate Acc is used to reflect the overall performance of the residual spatio-temporal graph convolutional neural network model, and the misjudgment rate MAR reflects the probability that the unstable samples are misjudged as stable by the residual spatio-temporal graph convolutional neural network model. The higher the accuracy rate and the lower the misjudgment rate, the better the evaluation performance of the residual spatio-temporal graph convolutional neural network model;

[0054] Record the total number of all transient time-domain simulation samples evaluated as correct by the residual spatio-temporal graph convolutional neural network model in this round of training and the total number of unstable samples evaluated as unstable as well as the total number of unstable samples evaluated as stable and calculate the transient stability evaluation accuracy and misjudgment rate of the residual spatio-temporal graph convolutional neural network model. The calculation method is as follows:

[0055] ;

[0056] If the accuracy and misjudgment rate of all transient time-domain simulation sample sets discriminated by the residual spatio-temporal graph convolutional network model reach the preset performance evaluation index threshold, a trained residual spatio-temporal graph convolutional neural network model is obtained; if not, the spatio-temporal feature mining training of the residual spatio-temporal graph convolutional neural network model is performed again.

[0057] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a transient voltage stability evaluation method based on unbalanced residual information learning, which has the following beneficial effects:

[0058] Combining the spatial topology matrix of power system physical information with the temporal characteristics of the transient voltage response of each monitoring node after a fault, the dynamic spatio-temporal evolution law of system transient voltage instability is mined from two aspects of data and power grid structure modeling itself. Different from the traditional analysis methods based on single temporal data or power system topology, the present invention combines power system physical topology and electrical measurement data, effectively taking into account the spatio-temporal characteristics of transient voltage after a fault. Aiming at the problem of spatio-temporal feature extraction smoothness caused by the increase in the number of spatio-temporal graph convolutional network layers, the present invention introduces residual convolution to retain the original features, and better realizes the mapping relationship between spatio-temporal feature extraction and transient voltage stability state, improving the accurate prediction of transient voltage instability. At the same time, the existing methods are insufficient in dealing with the problem of sample imbalance. The present invention introduces the Focal-Loss loss function to dynamically adjust the weights of positive and negative samples, making the model pay more attention to minority class samples (transient voltage instability samples), thereby improving the evaluation accuracy of transient voltage instability. The transient voltage stability evaluation method based on unbalanced residual information learning of the present invention can not only improve the prediction accuracy of transient voltage stability evaluation, but also has strong adaptability, and can provide accurate transient voltage stability early warning under complex power grid structures and various fault scenarios. In addition, with the popularization of power grid monitoring equipment, the model can also monitor the power grid state in real time, provide accurate basis for power grid dispatching and control decisions, and provide strong support for the safe operation and control strategy optimization of the power grid. Description of the Drawings

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0060] Figure 1 It is a flowchart of a transient voltage stability assessment method based on unbalanced residual information learning provided by the present invention;

[0061] Figure 2 It is a structural diagram of a residual spatio-temporal graph convolutional neural network model provided by an embodiment of the present invention. Specific Embodiments

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0063] The present invention provides a transient voltage stability assessment method based on unbalanced residual information learning, as Figure 1 shown, including the following steps:

[0064] S1. Build a target power system simulation model corresponding to the target power system;

[0065] S2. Perform transient voltage time-domain simulation on the target power system simulation model to obtain M time-domain simulation results, and respectively construct a spatio-temporal graph model corresponding to the target power system simulation model under each simulation condition according to the spatio-temporal graph construction method G , and at the same time, for M time-domain simulation results, respectively calibrate the system state S as voltage stable or unstable. Based on M calibration results and the spatio-temporal graph model G construct a transient time-domain simulation sample set, denoted as ;

[0066] S3. Construct a Focal-Loss loss function, and by calculating the stable and unstable class weight factors of the transient time-domain simulation sample set and setting an adjustment factor for sample classification, obtain a Focal-Loss loss function for training the residual spatio-temporal graph convolutional neural network model to learn sample class imbalance;

[0067] S4. Construct a residual spatio-temporal graph convolutional neural network model, and perform spatio-temporal feature learning on the transient time-domain simulation sample set through the residual spatio-temporal graph convolutional neural network model to obtain the prediction result of the transient voltage state;

[0068] S5. Judge whether the prediction result is correct according to the actual transient voltage state result of the target power system. If it is incorrect, perform the next round of training on the residual spatio-temporal graph convolutional neural network model until the performance evaluation index threshold of the preset spatio-temporal feature learning is satisfied, and obtain the trained residual spatio-temporal graph convolutional neural network model;

[0069] S6. Perform transient voltage stability assessment of the power system through the trained residual spatio-temporal graph convolutional neural network model.

[0070] It should be noted that:

[0071] The DSP-Studio electromechanical transient time-domain simulation software is used to build a model of the target power system to comprehensively capture the spatio-temporal characteristics of the transient voltage instability evolution of each monitoring node in the system;

[0072] The residual spatio-temporal graph convolutional neural network model adopts the pytorch framework of the deep neural network model in python.

[0073] To further implement the above technical solution, the specific content of the spatio-temporal graph construction method in S2 includes:

[0074] According to the spatio-temporal graph construction method, construct the k th time-domain simulation sample observation time window with a length of the graph model corresponding to the physical topology of the target power system G , and obtain the final spatio-temporal graph model ;

[0075] Each monitoring node in the target power system under the observation time window is represented as a vertex in the graph model, denoted as X and , N is the number of monitoring nodes, and the connection relationship between each monitoring node is represented as a topological adjacency matrix, denoted as A and , then the k th sample observation time window has a length of the graph model corresponding to the physical topology of the target power system ;

[0076] Obtain node time series data at any monitoring node , and the edge is used to reflect the nodes and node The physical connection relationship between them, if there is a connection, it is recorded as , otherwise recorded as ;

[0077] Finally build the final space-time graph model To characterize the dynamic characteristics of the transient response data of the target power system in the time and space dimensions.

[0078] It should be noted that:

[0079] The monitoring nodes in the system include: generators, substations and loads, and the node time series data include: voltage amplitude, active power and reactive power.

[0080] In order to further implement the above technical solution, the monitoring nodes in the system include: generator, substation and load, and the node time series data include: voltage amplitude, active power and reactive power.

[0081] In order to further implement the above technical solution, the specific contents of S2 include:

[0082] Perform transient voltage time domain simulation on the target power system simulation model, and simulate the voltage stability of the power grid under various typical operating conditions by combining the system historical operation mode and fault settings. The total number is Time domain simulation results;

[0083] After each time domain simulation, record the physical adjacency matrix of the target power system and the length of the observation time window on the corresponding monitoring node after the fault occurs is Temporal evolution characteristics of , including Voltage amplitude under observation time window length 、Active power and reactive power , and the system status of this simulation result is analyzed according to the transient voltage stability engineering criterion adopted by the actual power system calibration, where indicates that the power system state result of this time domain transient simulation is voltage instability, Indicates that the simulated power system state result is voltage stability;

[0084] Based on M Calibration results and spatiotemporal graph model G Construct the input sample set as , among which,

[0085] .

[0086] In order to further implement the above technical solutions, M ​It is the product of the typical operating level numbers of the generator, new energy output, and load consumption, the proportion of transient operating dynamic load, and the typical fault disturbance number.

[0087] To further implement the above technical solution, the Focal-Loss loss function in S3 specifically includes: calculating the stability and instability category weight factors of the transient time-domain simulation sample set and setting an adjustment factor for sample classification ;

[0088] Category weight factor It is set according to the quantity ratio of positive samples and negative samples:

[0089] Regarding the instability samples in the transient time-domain simulation sample set as positive samples and the stability samples in the transient time-domain simulation sample set as negative samples;

[0090] Assume the number of positive samples is c , the number of negative samples is d , and the total number of samples is , then for the weight ratio of instability samples , and for the weight ratio of stability samples ;

[0091] Adjustment factor It is used to control the degree of loss weight reduction for sample classification and simultaneously enhance the attention of the residual spatio-temporal graph convolutional network model to difficult-to-evaluate samples:

[0092] ;

[0093] Based on the settings of the category weight factor and the adjustment factor , the Focal-Loss loss function is:

[0094] ;

[0095] Among them, is the predicted probability of the sample by the residual spatio-temporal graph convolutional network model. For instability samples , and for stability samples .

[0096] To further implement the above technical solution, the residual spatio-temporal graph convolutional neural network model in S4 specifically includes: a residual spatio-temporal graph convolutional module and a fully connected classification module;

[0097] The residual spatio-temporal graph convolutional module includes a spatial graph convolutional layer, a temporal convolutional layer, and a residual convolutional layer;

[0098] The spatial graph convolutional layer is used to obtain any spatio-temporal graph input sample in the input sample set and perform operations on the power system topological adjacency matrix Perform spatial feature mining, and achieve the feature aggregation of the time series information of neighbor nodes on the power system topology map to the central node, and output spatial features ;

[0099] Temporal convolutional layer, used to further perform convolutional operations in the time dimension to achieve the spatio-temporal feature extraction of transient voltage response data, and output temporal features ;

[0100] Residual convolutional layer, used to perform convolutional operations and output residual features and maintain the same dimension as the output features of the spatio-temporal graph ;

[0101] Fully connected classification module, used to fuse residual features and spatio-temporal features, and reduce the dimension of the fused features through a linear layer, further extract high-dimensional features for classifying transient voltage stability and instability, and convert the high-dimensional features into a probability distribution through the softmax activation function to complete the mapping of the short-term transient voltage stability state space, and achieve the evaluation of system transient voltage stability or instability.

[0102] To further implement the above technical solution, the specific content of the spatial graph convolutional layer outputting spatial features includes:

[0103] Perform degree matrix conversion on the target power system topology adjacency matrix , and the conversion method is:

[0104] ;

[0105] In the formula, is the element in the th row and th column of the degree matrix;

[0106] Based on the identity matrix and the degree matrix , further perform spatial feature mining on the power system topology structure matrix using the Laplacian graph conversion formula, where the Laplacian graph conversion formula is:

[0107] ;

[0108] Among them, the identity matrix is a special square matrix, all elements on its main diagonal are 1, and all other elements are 0. The degree matrix It is also a special square matrix used to represent the degree of each vertex in a graph, where this degree is the number of connections of each node on the power grid adjacency matrix to other nodes. For the target power system topological adjacency matrix perform spatial feature mining, based on the solution formula to obtain the standard orthogonal eigenvectors and the spatial eigenvalue matrix of the target power system topological adjacency matrix , and use the graph node information aggregation formula to achieve the feature aggregation of the temporal information of neighbor nodes on the power system topological graph to the central node. Among them, the graph node information aggregation formula is:

[0109] ;

[0110] In the formula, is the graph convolution kernel, is the length of the temporal evolution feature of the graph matrix node observation time window .

[0111] To further implement the above technical solution, the specific content of the temporal convolution layer outputting the temporal feature includes:

[0112] ;

[0113] Among them, ReLU is a non-linear activation function, is the convolution kernel parameter, is the convolution operation.

[0114] To further implement the above technical solution, the fully connected classification module reduces the dimension of the fused features through a linear layer, further extracts the high-dimensional features for classifying transient voltage stability and instability, and converts the high-dimensional features into a probability distribution through the fully connected classification formula, completes the mapping of the short-term transient voltage stability state space, and realizes the evaluation of the system transient voltage stability state. Among them, the fully connected classification formula is:

[0115] .

[0116] It should be noted that:

[0117] As the number of layers of the spatio-temporal graph network increases, the eigenvectors of different nodes will become more and more similar. This phenomenon is called over-smoothing, which will significantly reduce the stable evaluation performance of the model. To enhance the ability to capture the spatio-temporal dependence relationship of the transient process, the residual convolution layer adds the output of the previous layer to the output of the current layer.

[0118] After being processed by the residual spatio-temporal graph convolutional network, the output information retains the temporal dependence of the transient voltage responses of each node and incorporates the spatial dependence of the power system topology, enabling a more comprehensive characterization of the spatio-temporal features of the transient response data.

[0119] The residual spatio-temporal graph convolutional network model constructed in this embodiment is as Figure 2 shown.

[0120] To further implement the above technical solution, the specific content of S5 includes:

[0121] Select the accuracy rate and the misjudgment rate as the performance evaluation indicators for the spatio-temporal feature learning of the residual spatio-temporal graph convolutional neural network model. Among them, the accuracy rate Acc is used to reflect the overall performance of the residual spatio-temporal graph convolutional neural network model, and the misjudgment rate MAR reflects the probability that the unstable samples are misjudged as stable by the residual spatio-temporal graph convolutional neural network model. The higher the accuracy rate and the lower the misjudgment rate, the better the evaluation performance of the residual spatio-temporal graph convolutional neural network model;

[0122] Denote the total number of all transient time-domain simulation samples evaluated as correct by the residual spatio-temporal graph convolutional neural network model in this round of training and the total number of unstable samples evaluated as unstable as well as the total number of unstable samples evaluated as stable , and calculate the transient stability evaluation accuracy rate and misjudgment rate of the residual spatio-temporal graph convolutional neural network model. The calculation method is:

[0123] ;

[0124] If the accuracy rate and misjudgment rate of all transient time-domain simulation sample sets discriminated by the residual spatio-temporal graph convolutional network model reach the preset performance evaluation index threshold (in this embodiment ), then the trained residual spatio-temporal graph convolutional neural network model is obtained; if not (in this embodiment ), then the residual spatio-temporal graph convolutional neural network model is re-trained for spatio-temporal feature mining.

[0125] It should be noted that:

[0126] In this embodiment, the trained residual spatio-temporal graph convolutional neural network model is used for real-time system transient voltage state monitoring, extracting the physical topology matrix of the power system after a fault and the time-series response data of the observation time window length of each monitoring node on the matrix and inputting it into the trained residual spatio-temporal graph convolutional neural network model obtained in S5. Based on the discriminated stable state to evaluate the transient voltage stable state of the corresponding system. If the discriminated stable state , it indicates that after the power system has experienced a certain transient fault, there will be a risk of voltage instability in a short period of time, and even voltage collapse may occur. Voltage instability may cause serious consequences, including damage to power equipment, large-scale power outages or system collapse. Therefore, an alarm needs to be sent to the grid control center and targeted control measures need to be prepared to restore the stable operation of the power system. If the stable state is judged , it indicates that the transient voltage of the power system is stable, there are no serious voltage fluctuations in the system, and the voltage will remain within the safe range within a certain period of time, but real-time monitoring needs to be maintained.

[0127] The above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A transient voltage stability assessment method based on unbalanced residual information learning, characterized in that: The following steps are involved: S1. Build a target power system simulation model corresponding to the target power system; S2. Perform transient voltage time domain simulation on the target power system simulation model to obtain M The time-domain simulation results are obtained, and the time-space graph model corresponding to the target power system simulation model under each simulation condition is constructed according to the time-space graph construction method. G , while targeting M The time domain simulation results are respectively used to analyze the system state S Perform voltage stability or instability calibration based on M Calibration results and spatiotemporal graph models G Construct a transient time domain simulation sample set, denoted as ; The specific contents of S2 include: The transient voltage time domain simulation is performed on the target power system simulation model, and the voltage stability of the power grid under various typical operating conditions is simulated by combining the system historical operation mode and fault settings. The total number is The time domain simulation results of the times; After each time domain simulation, record the physical adjacency matrix of the target power system And the length of the observation time window on the corresponding monitoring node after the fault occurs is The temporal evolution characteristics of ,include The voltage amplitude under the observation time window length , Active Power and reactive power , and the system status is analyzed based on the transient voltage stability engineering criterion used in the actual power system. Calibration, where Indicates that the power system state result of this time domain transient simulation is voltage instability. It indicates that the result of the simulated power system state is voltage stability; based on M Calibration results and spatiotemporal graph models G Construct the input sample set as ,in, ； S3. Construct the Focal-Loss loss function by calculating the weight factors of the stable and unstable categories of the transient time domain simulation sample set and set the adjustment factor for sample classification Obtain the Focal-Loss loss function for training the residual spatiotemporal graph convolutional neural network model to learn the imbalanced sample categories; S4. Construct a residual spatiotemporal graph convolutional neural network model, and use the residual spatiotemporal graph convolutional neural network model to learn the spatiotemporal features of the transient time domain simulation sample set to obtain the prediction result of the transient voltage state; S5. Determine whether the prediction result is correct according to the actual transient voltage state result of the target power system. If it is not correct, perform the next round of training on the residual spatiotemporal graph convolutional neural network model until the preset performance evaluation index threshold of spatiotemporal feature learning is met, and obtain the trained residual spatiotemporal graph convolutional neural network model; S6. The transient voltage stability of the power system is evaluated by using the trained residual spatiotemporal graph convolutional neural network model.

2. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 1 is characterized in that: The specific contents of the space-time graph construction method in S2 include: According to the construction method of space-time graph, construct the k The observation time window length of a time domain simulation sample is The graphical model corresponding to the physical topology of the target power system G , and obtain the final space-time graph model ; The first The monitoring nodes are represented as vertices in the graph model. , recorded as X and , N is the number of monitoring nodes, The monitoring node and The connection relationship between the monitoring nodes is used as the edge in the graph model , collect all It is represented as a topological adjacency matrix, denoted as A and , among which k The length of the sample observation time window is The graphical model corresponding to the physical topology of the target power system ; At any monitoring node Get the node time series data on the top, if there is a connection, record it as , otherwise it is recorded as ; Finally, the final spatiotemporal graph model is constructed To characterize the dynamic characteristics of the transient response data of the target power system in the time and space dimensions.

3. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 1 is characterized in that: M It is the product of the typical operating level number of generators, new energy output and load consumption, the proportion of transient operating dynamic load and the typical fault disturbance number.

4. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 1 is characterized in that: The Focal-Loss loss function in S3 specifically includes: Calculating the weight factors of stable and unstable categories of transient time domain simulation sample sets and set the adjustment factor for sample classification ; Category weight factor Set according to the ratio of positive samples to negative samples: The unstable samples of the transient time-domain simulation sample set are taken as positive samples, and the stable samples of the transient time-domain simulation sample set are taken as negative samples; Assume that the number of positive samples is c , the number of negative samples is d The total number of samples is M = c + d , then the weight ratio of unstable samples is , for the weight ratio of stable samples ; Adjustment Factor Used to control the degree of loss weight reduction for sample classification, while enhancing the residual spatiotemporal graph convolutional network model's attention to difficult-to-evaluate samples: ； Based on category weight factor and adjustment factors The Focal-Loss loss function is: ； in, is the predicted probability of the residual spatiotemporal graph convolutional network model for the sample. For unstable samples , for a stable sample .

5. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 1 is characterized in that: The residual spatiotemporal graph convolutional neural network model in S4 specifically includes: a residual spatiotemporal graph convolution module and a fully connected classification module; The residual spatiotemporal graph convolution module includes a spatial graph convolution layer, a temporal convolution layer, and a residual convolution layer; The spatial graph convolution layer is used to obtain any spatiotemporal graph input sample in the input sample set and perform topological adjacency matrix analysis on the power system. Perform spatial feature mining and realize the feature aggregation of neighbor node time series information to the central node on the power system topology map, and output spatial features ; Temporal convolution layer, used to further Perform convolution operation to extract the spatiotemporal features of transient voltage response data and output the timing features ; Residual convolution layer is used to Perform convolution operation and output residual features And keep the output features of the spatiotemporal graph Dimensionality is consistent; Fully connected classification module for residual features and spatiotemporal characteristics The fusion is performed and the dimensionality of the fused features is reduced through a linear layer to further extract high-dimensional features for classifying transient voltage stability and instability. The high-dimensional features are converted into probability distributions through a softmax activation function to complete the mapping of the short-term transient voltage stability state space and realize the transient voltage stability or instability assessment of the system.

6. A transient voltage stability assessment method based on unbalanced residual information learning according to claim 5, characterized in that: The spatial graph convolution layer outputs spatial features The specific contents include: The adjacency matrix of the target power system topology Degree Matrix Conversion, the conversion method is: ； In the formula, is the degree matrix Line Elements of a column; Based on the identity matrix Sum degree matrix , and further use the Laplace graph transformation formula to transform the power system topology matrix Perform spatial feature mining, where the Laplace graph transformation formula is: ； in, is the identity matrix; degree matrix It is used to represent the degree of each vertex in the graph model. The degree is the number of connections between each node and other nodes on the power grid adjacency matrix. Perform spatial feature mining based on the solution formula Get the target power system topology adjacency matrix The orthonormal eigenvectors of and the spatial eigenvalue matrix , the graph node information aggregation formula is used to realize the feature aggregation of the neighbor node time series information on the power system topology graph to the central node, where the graph node information aggregation formula is: ； In the formula, is the graph convolution kernel, is the length of the time window for observing nodes in the graph matrix The temporal evolution characteristics of .

7. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 1 is characterized in that: The temporal convolution layer outputs temporal features The specific contents include: ； Among them, ReLU is a nonlinear activation function. is the convolution kernel parameter, It is the convolution operation.

8. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 5, characterized in that: The fully connected classification module reduces the dimension of the fused features through the linear layer, further extracts high-dimensional features for classifying transient voltage stability and instability, and converts the high-dimensional features into probability distribution through the fully connected classification formula. , the mapping of the short-term transient voltage stability state space is completed, and the transient voltage stability state evaluation of the system is realized. The full connection classification formula is: 。 9. The transient voltage stability assessment method based on unbalanced residual information learning according to claim 5, characterized in that: The specific contents of S5 include: The accuracy and misclassification rate are selected as the performance evaluation indicators of the residual spatiotemporal graph convolutional neural network model for spatiotemporal feature learning. The accuracy rate Acc is used to reflect the overall performance of the residual spatiotemporal graph convolutional neural network model, and the misclassification rate MAR reflects the probability that an unstable sample is misclassified as stable by the residual spatiotemporal graph convolutional neural network model. The higher the accuracy rate and the lower the misclassification rate, the better the evaluation performance of the residual spatiotemporal graph convolutional neural network model. The total number of correct transient time-domain simulation samples evaluated by the residual spatiotemporal graph convolutional neural network model in this round of training is , the total number of unstable samples is evaluated as unstable And the total number of unstable samples evaluated as stable , and calculate the transient stability assessment accuracy and misjudgment rate of the residual spatiotemporal graph convolutional neural network model. The calculation method is: ； If the accuracy and misjudgment rate of all transient time-domain simulation sample sets using the residual space-time graph convolutional network model reach the preset performance evaluation index threshold, a trained residual space-time graph convolutional neural network model is obtained; if not, the residual space-time graph convolutional neural network model is re-trained for space-time feature mining.

Images