Power grid chain failure risk scene recognition method, system, device, and medium

Abstract

A power grid chain failure risk scene recognition method, system, device, and medium includes: acquiring power grid historical operation data (S101); performing chain failure simulation on the power grid historical operation data to obtain active power flow entropy (S102); visually displaying the power grid historical operation data and the active power flow entropy to obtain a multi-row heat map (S103); dividing the multi-row heat map into a training set and a test set (S104); training a neural network using the training set to obtain a power grid chain failure risk scene recognition model (S105); and testing the power grid chain failure risk scene recognition model using the test set to obtain a risk scene recognition result (S106).

Description

[Technical Field]

[0001] This application claims priority from a Chinese patent application bearing application number 202310939393.1, filed with the China Patent Office on July 28, 2023, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to the field of power risk scenes, for example, to a power grid chain failure risk scene recognition method, system, device and medium. [Background technology]

[0003] With the application of flexible AC / DC transmission technology in current power grids and the access of new loads such as electric vehicles, the grid structure is becoming increasingly complex, and some low-probability events in the power grid may trigger cascading failures with disastrous consequences. In order to improve the reliability and stability of the power grid, avoid chain reactions and large-scale power outages, reduce restoration costs and time, and ensure people's lives and safety, timely predict and identify potential risks so that corresponding measures can be taken to reduce the occurrence of cascading failures and ensure the normal operation of the power grid. Therefore, accurately understanding the current operating status of the power grid is of great significance to ensuring the normal operation of the power grid.

[0004] With the emergence of a new generation of artificial intelligence technology, represented by deep learning technology, new ideas have emerged for realizing online identification of power grid chain failure risk scenes. By using deep learning technology to uncover the inherent connection between historical operating conditions and risk scenes and constructing a recognition model, the speed of online recognition of power grid chain failure risk scenes can be improved, which is extremely important in practical applications. Summary of the Invention

[0005] The present application provides a power grid chain failure risk scene recognition method, system, device and medium that can accurately realize risk recognition of the current operation scene of the power grid.

[0006] This application is obtaining historical power grid operating data; performing a cascading fault simulation on the power grid historical operation data to obtain an active power flow entropy; Visually displaying the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map; Partitioning the multi-row heatmap into a training set and a test set; Using the training set to train a neural network to obtain a power grid chain failure risk scene recognition model; and testing the power grid chain failure risk scene recognition model using a test set to obtain a risk scene recognition result.

[0007] Preferably, performing a cascading fault simulation on the power grid historical operation data to obtain the active power flow entropy includes: constructing a power grid simulation model based on the power grid historical operation data, setting a plurality of system operation scenes, and performing a cascading fault simulation on the lines in the power grid to obtain a cascading fault chain and a cascading fault load factor; determining an active power flow entropy based on the cascading fault chain and the cascading fault load factor.

[0008] Preferably, the power grid history operation data includes a power grid chain failure risk scene, and visually displaying the power grid history operation data and the active power flow entropy to obtain a multi-row heat map includes: Using a load factor under a current operating condition of the power grid chain failure risk scene, draw a first row of a multi-row heat map corresponding to the power grid chain failure risk scene; plotting a second row of a multi-row heat map corresponding to the power grid cascading failure risk scene using the capacities of the lines in the power grid cascading failure risk scene; Using the node feature vector of the power grid cascading failure risk scene, draw a third row of a multi-row heat map corresponding to the power grid cascading failure risk scene; Using the active power flow entropy, plot a fourth row of a multi-row heat map corresponding to the power grid cascading fault risk scene; and plotting a fifth row of the multi-row heat map corresponding to the power grid cascading failure risk scene using the proportion of the historical cascading failure occurrence of the power grid cascading failure risk scene to the total failure occurrence.

[0009] Preferably, using the training set to train the neural network to obtain the power grid chain failure risk scene recognition model specifically includes: The method includes using multi-row heat maps of multiple types of power grid operating conditions in the training set as inputs to a Convolutional Neural Network (CNN) model, using risk scenes corresponding to various operating conditions in the training set as outputs of the CNN model, and using a cross-entropy loss function as a loss function to train the CNN model to obtain a power grid chain failure risk scene recognition model.

[0010] Preferably, after using a test set to test the power grid chain failure risk scene recognition model and obtaining a risk scene recognition result, The method further includes constructing a salient risk scene set using a clustering algorithm based on the risk scene recognition result.

[0011] Preferably, the active power flow entropy includes an active power flow entropy of an overloaded line and an active power flow entropy of a non-overloaded line; The expression for the active power flow entropy of the non-overloaded line is:

number

number

[0012] This application is an acquisition module configured to acquire historical power grid operating data; a cascading fault simulation module configured to perform a cascading fault simulation on the power grid historical operation data to obtain an active power flow entropy; an imaging display module configured to visualize and display the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map; a partitioning module configured to partition the multi-row heatmap into a training set and a test set; a training module configured to train a neural network using a training set to obtain a power grid chain fault risk scene recognition model; The present invention further provides a power grid chain failure risk scene recognition system, comprising: a test module configured to test the power grid chain failure risk scene recognition model using a test set to obtain a risk scene recognition result.

[0013] Preferably, the cascading fault simulation module comprises: a construction and simulation unit configured to construct a power grid simulation model based on the power grid historical operation data, set a plurality of system operation scenes, perform cascading fault simulation on lines in the power grid, and obtain a cascading fault chain and a cascading fault load factor; an active power flow entropy determination unit configured to determine an active power flow entropy based on the cascading fault chain and the cascading fault load factor.

[0014] This application is at least one processor; a storage device storing at least one program; There is further provided an electronic device in which, when the at least one program is executed by the at least one processor, the at least one processor implements the method as described above.

[0015] The present application further provides a computer storage medium having stored thereon a computer program which, when executed by a processor, implements such a method. [Brief explanation of the drawings]

[0016] The drawings that need to be used in the examples are introduced below.

[0017] [Figure 1] 1 is a schematic diagram of a power grid chain failure risk scene recognition method according to an embodiment of the present application; [Figure 2] 1 is a multi-row heat map corresponding to various power grid cascading fault risk scenarios according to an embodiment of the present application; [Figure 3]1 is a structural diagram of a power grid chain failure risk scene recognition model according to an embodiment of the present application; [Figure 4] 1 is a flowchart of a method for recognizing a power grid chain failure risk scene according to an embodiment of the present application; [Figure 5] 1 is a structural block diagram of a power grid chain failure risk scene recognition system according to an embodiment of the present application; FIG. DETAILED DESCRIPTION OF THE INVENTION

[0018] The following describes the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application, and it is clear that the described embodiments are not necessarily all of the embodiments but only a part of the embodiments of the present application. All other embodiments that can be obtained by those skilled in the art based on the embodiments of the present application without any creative work fall within the scope of protection of the present application.

[0019] The present application provides a power grid chain failure risk scene recognition method, system, device and medium that can accurately realize risk recognition of the current operation scene of the power grid.

[0020] As shown in FIGS. 1 and 4, the power grid chain failure risk scene recognition method according to the present application includes:

[0021] In step 101, historical power grid operating data is obtained.

[0022] In step 102, a cascading fault simulation is performed on the power grid historical operation data to obtain the active power flow entropy.

[0023] Step 102 includes: constructing a power grid simulation model based on the power grid historical operation data, setting different system operation scenarios, performing cascading fault simulation on lines in the power grid, obtaining cascading fault chains and cascading fault load factors; and determining an active power flow entropy based on the cascading fault chains and the cascading fault load factors.

[0024] Historical power grid operation data is collected, cascading fault simulation is performed, active power flow entropy is defined as a scene recognition index for power grid cascading faults, a power grid simulation model is constructed, different system operation scenes are set, N-1 cascading fault simulation is performed on the line considering cascading faults led by line overload, and the cascading fault chain and cascading fault load factor are obtained.

[0025] Entropy indicates the uniformity of spatial distribution of things and is used to express the uncertainty of a system. Power flow entropy is a measure of the balance of power flow among multiple branch circuits in a system, and is quantified by the distribution of the load factors of the branch circuits. The arithmetic progression of load factors Z = [Z1, Z2, , Z n ] and the power flow entropy F of the system is

number

[0026] The number of branch circuits is the number of lines in the power grid.

[0027] In order to more specifically reflect the impact of system indicators on the power grid, the active power flow entropy is set in conjunction with the characteristics of the active power flow. As can be seen from major blackouts, even small disturbances can cause lines with high load factors to be overloaded and cut off, leading to cascading faults. In addition, the load factor of a line is positively correlated with the probability of its outage. Therefore, when the load factors of all lines are less than 1, the active power flow entropy of non-overloaded lines in the system is:

number

number

[0028] In step 103, the power grid historical operation data and the active power flow entropy are visualized to obtain a multi-line heat map, where the power grid historical operation data includes a power grid chain failure risk scene.

[0029] Step 103 is Using a load factor in a current operating state of the power grid cascading failure risk scene in the power grid historical operation data, plotting a first row of a multi-row heat map corresponding to the power grid cascading failure risk scene; plotting a second row of the multi-row heat map corresponding to the power grid cascading failure risk scene using the capacity of the line in the power grid historical operation data; Using the node feature vector of the power grid cascading failure risk scene in the power grid historical operation data, plot a third row of the multi-row heat map corresponding to the power grid cascading failure risk scene; Using the active power flow entropy, plot a fourth row of a multi-row heat map corresponding to the power grid cascading fault risk scene; and plotting a fifth row of the multi-row heat map corresponding to the power grid cascading failure risk scene using the proportion of historical cascading failure occurrences of the power grid cascading failure risk scene in the power grid historical operating data to total failure occurrences.

[0030] The graphical representation of various power grid chain failure risk scenes and the construction of a power grid chain failure risk scene recognition model based on CNN include: using multi-row heat maps to graphically represent the various power grid chain failure risk scenes and generating multi-row heat maps corresponding to the various power grid chain failure risk scenes; inputting the multi-row heat maps corresponding to the various power grid chain failure risk scenes into a pre-trained machine learning model, respectively, to obtain classification results output by the machine learning model for the various power grid chain failure risk scenes. As shown in Figure 3, the basic structure of CNN consists of an input layer, a convolutional layer, a pooling layer (also called a sampling layer), a fully connected layer, and an output layer.

[0031] The CNN-based power grid chain failure risk scene recognition model consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The input layer receives multi-row heat map data.

[0032] Referring to FIG. 2, for each of the power grid chain failure risk scenarios, the load factor of the current operating status of the power grid chain failure risk scenario is used to plot the first row of the multi-row heat map corresponding to the power grid chain failure risk scenario, and the load factor is Z k Shown in.

[0033] The capacity of the line in the power grid chain failure risk scene is used to draw the second row of the multi-row heat map corresponding to the power grid chain failure risk scene, and the capacity of the line is C m Shown in.

[0034] The node feature vector of the power grid chain failure risk scene is used to draw the third row of the multi-row heat map corresponding to the power grid chain failure risk scene. In this example, the node feature vector is denoted as X Input Shown in.

[0035] The active power flow entropy is used to plot the fourth row of the multi-row heat map corresponding to the power grid chain failure risk scene, and the active power flow entropy is calculated as H and Hm Shown in.

[0036] The proportion of the historical chain failure occurrence of the power grid chain failure risk scene to the total failure occurrence is used to draw the fifth row of the multi-row heat map corresponding to the power grid chain failure risk scene, and the proportion of the historical chain failure occurrence to the total failure occurrence is Q m Up to this point, a multi-row heat map corresponding to the power grid chain failure risk scene is generated.

[0037] In step 104, the multi-row heatmap is partitioned into a training set and a test set.

[0038] In step 105, the training set is used to train the neural network to obtain a power grid chain fault risk scene recognition model.

[0039] Step 105 includes: using the multi-row heat maps of different operating conditions of the power grid in the training set as inputs of a CNN model; using the power grid chain failure risk scenes corresponding to the operating conditions in the training set as outputs of the CNN model; and using a cross-entropy loss function as a loss function to train the CNN model to obtain a power grid chain failure risk scene recognition model.

[0040] A power grid chain failure risk scene recognition model based on CNN is constructed. When multi-row heat map samples under different operating conditions of the power grid are used as input to the model, the output of the model is the power grid chain failure risk scene under the operating conditions.

number

[0041] The loss function F in the model loss adopts the cross-entropy loss function, whose formula is:

number

[0042] The constructed CNN-based power grid chain fault risk scene recognition model can analyze uncertain chain faults that may occur in the power grid under different operating conditions, and correspondingly output whether the operating condition is a risk scene.

[0043] In step 106, the power grid chain failure risk scene recognition model is tested using a test set to obtain a risk scene recognition result.

[0044] In practical applications, after testing the power grid chain failure risk scene recognition model using a test set and obtaining risk scene recognition results, the method further includes constructing a salient risk scene set using a clustering algorithm based on the risk scene recognition results.

[0045] The multi-row heat maps are divided into a training set and a test set and input into a recognition model. The model is trained offline and a clustering method is used to construct a set of salient risk scenes. A power grid cascading fault risk scene corresponding to the active power flow entropy is output. The multi-row heat maps are divided into a training set and a test set. The samples are input into a CNN-based power grid cascading fault risk scene recognition model. Feature learning is performed on the multi-row heat maps using the CNN model. A clustering method is used to construct a set of salient risk scenes. The model is trained and outputs whether various operating states are cascading fault risk scenes. The model's recognition ability is verified using the test set. When a multi-row heat map for one operating state is input, the model can recognize whether the scene corresponding to that operating state is a risk scene and output a value of 0 or 1, where 0 indicates no risk and 1 indicates yes.

[0046] The present application obtains power grid historical operation data, performs a cascading fault simulation on the power grid historical operation data, obtains active power flow entropy, visualizes the power grid historical operation data and the active power flow entropy, obtains a multi-line heat map, divides the multi-line heat map into a training set and a test set, trains a neural network using the training set to obtain a power grid cascading fault risk scene recognition model, and tests the power grid cascading fault risk scene recognition model using the test set to obtain a risk scene recognition result. In online applications, the active power flow entropy can be directly input to the power grid cascading fault risk scene recognition model to obtain a risk scene recognition result, thereby accurately realizing risk recognition of the current operating scene of the power grid.

[0047] As shown in FIG. an acquisition module 10 configured to acquire historical power grid operating data; a cascading fault simulation module 20 configured to perform a cascading fault simulation on the power grid historical operating data to obtain an active power flow entropy; an imaging display module 30 configured to visualize and display the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map; a segmentation module 40 configured to segment the multi-row heatmap into a training set and a test set; a training module 50 configured to train a neural network using a training set to obtain a power grid chain fault risk scene recognition model; The present invention further provides a power grid chain failure risk scene recognition system, comprising: a test module 60 configured to test the power grid chain failure risk scene recognition model using a test set to obtain a risk scene recognition result.

[0048] In one alternative embodiment, the cascading fault simulation module 20: a construction and simulation unit configured to construct a power grid simulation model based on the power grid historical operation data, set different system operation scenarios, perform cascading fault simulation on lines in the power grid, and obtain cascading fault chains and cascading fault load factors; and an active power flow entropy determination unit configured to determine an active power flow entropy based on the cascading fault chain and the cascading fault load factor.

[0049] The present application further provides an electronic device comprising one or more processors and a storage device in which one or more programs are stored, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement a method described in an embodiment of the present application.

[0050] The present application further provides a computer storage medium having stored thereon a computer program that, when executed by a processor, implements the method described in the embodiments of the present application.

[0051] This paper considers the active power flow of a line and defines the active power flow entropy, which reflects the risk of a cascading failure in the current scenario. The active power flow entropy is used as an important indicator for identifying power grid cascading failure risk scenarios. A deep learning model is applied to extract the correspondence between the active power flow entropy of a line and power grid cascading failure risk scenarios. A multi-row heat map corresponding to the power grid cascading failure risk scenario is drawn by linking the power grid load factor, line capacity, and line node. Based on the multi-row heat map, a power grid cascading failure risk scenario recognition model based on a convolutional neural network (CNN) is constructed. The neural network in the model is used to autonomously learn the mapping relationship between the multi-row heat map and the risk scenario, and a softmax classifier is used to output the risk scenario recognition results. This method first uses an offline training model to learn the correspondence between the active power flow entropy of historical lines and the risk scenarios of chain failures in the power grid, and obtains a set of salient risk scenarios. When applied online, the active power flow entropy is directly input into the model to realize risk recognition of the current operating scenario of the power grid.

Claims

1. obtaining historical power grid operating data; performing a cascading fault simulation on the power grid historical operation data to obtain an active power flow entropy; Visually displaying the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map; Partitioning the multi-row heatmap into a training set and a test set; Using the training set to train a neural network to obtain a power grid chain failure risk scene recognition model; Testing the power grid chain failure risk scene recognition model using a test set to obtain a risk scene recognition result; Power grid chain failure risk scene recognition method.

2. performing a cascading fault simulation on the power grid historical operation data to obtain active power flow entropy, constructing a power grid simulation model based on the power grid historical operation data, setting a plurality of system operation scenes, and performing a cascading fault simulation on the lines in the power grid to obtain a cascading fault chain and a cascading fault load factor; determining an active power flow entropy based on the cascading fault chain and the cascading fault load factor; The method of claim 1.

3. The power grid historical operation data includes a power grid chain failure risk scenario; Visually displaying the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map, Draw a first row of a multi-row heat map corresponding to the power grid chain failure risk scene using a load factor under a current operating condition of the power grid chain failure risk scene; plotting a second row of a multi-row heat map corresponding to the power grid cascading failure risk scene using the capacities of the lines in the power grid cascading failure risk scene; Using the node feature vector of the power grid cascading failure risk scene, draw a third row of a multi-row heat map corresponding to the power grid cascading failure risk scene; Using the active power flow entropy, plot a fourth row of a multi-row heat map corresponding to the power grid cascading fault risk scene; and plotting a fifth row of the multi-row heat map corresponding to the power grid cascading failure risk scene using a proportion of the historical cascading failure occurrence of the power grid cascading failure risk scene to the total failure occurrence. The method of claim 1.

4. Using the training set to train a neural network and obtain a power grid chain failure risk scene recognition model, The multi-row heat maps of the power grid in the training set for various operating conditions are used as inputs to a convolutional neural network (CNN) model, and the risk scenes corresponding to the various operating conditions in the training set are used as outputs of the CNN model. A cross-entropy loss function is used as a loss function to train the CNN model, thereby obtaining a power grid chain failure risk scene recognition model. The method of claim 1.

5. After using the test set to test the power grid chain failure risk scene recognition model and obtain the risk scene recognition result, Further included is constructing a salient risk scene set using a clustering algorithm based on the risk scene recognition result. The method of claim 1.

6. The active power flow entropy includes an active power flow entropy of an overloaded line and an active power flow entropy of a non-overloaded line; The expression for the active power flow entropy of the non-overloaded line is: [Equation 1] and H is the active power flow entropy, R is a constant, and Z ｋ is the load factor of the kth branch circuit, and Z ｋ＋１ is the load factor of the k+1-th branch circuit, n is the total number of branch circuits, k is the k-th branch circuit, and P ｋ is the load factor (Z ｋ , Z ｋ＋１ ] is the ratio of the number of non-overloaded branch circuits located in the The expression for the active power flow entropy of the overloaded line is: [Equation 2] and H ｍ is the active power flow entropy of the overloaded line m, h is the basic entropy value of the overloaded line, and (Z ｗ , Z ｗ＋１ ] is the load factor section in which line m is located, and n ｗ is the number of branch circuits located in the load factor section where line m is located, and C ｍ is the capacitance of line m, and C ｍａｘ is the maximum line capacity in the system, and F ｍａｘ is the maximum allowable load factor of the line, P ｗ is the load factor (Z ｗ , Z ｗ＋１ ] is the percentage of the number of overloaded branch circuits located in the The method of claim 1.

7. an acquisition module configured to acquire historical power grid operating data; a cascading fault simulation module configured to perform a cascading fault simulation on the power grid historical operation data to obtain an active power flow entropy; an imaging display module configured to visualize and display the power grid historical operating data and the active power flow entropy to obtain a multi-row heat map; a partitioning module configured to partition the multi-row heatmap into a training set and a test set; a training module configured to train a neural network using a training set to obtain a power grid chain fault risk scene recognition model; a test module configured to test the power grid chain failure risk scene recognition model using a test set to obtain a risk scene recognition result; Power grid chain failure risk scene recognition system.

8. The cascading fault simulation module includes: a construction and simulation unit configured to construct a power grid simulation model based on the power grid historical operation data, set a plurality of system operation scenes, perform cascading fault simulation on lines in the power grid, and obtain a cascading fault chain and a cascading fault load factor; an active power flow entropy determination unit configured to determine an active power flow entropy based on the cascading fault chain and the cascading fault load factor; The system of claim 7.

9. at least one processor; a storage device in which at least one program is stored; The at least one program, when executed by the at least one processor, causes the at least one processor to implement the method of any one of claims 1 to 6. electronic equipment.

10. A computer program is stored which, when executed by a processor, implements the method of any one of claims 1 to 6. Computer storage media.

Images