A method and system for reconstructing missing data of power system based on graph neural network

Abstract

The present invention discloses a method and system for reconstructing missing data of an electric power system based on a graph neural network. The present invention constructs the missing electric power system operation data at the current moment into a matrix and inputs it into a trained graph neural network to obtain the complete electric power system operation data reconstructed at the current moment; the missing electric power system operation data in the matrix is ​​represented by a fixed value far away from the normal value; the graph neural network is obtained by training based on the historical operation data collected by the electric power system. The present invention uses a graph neural network to learn the complex spatiotemporal correlations between measurement data from historical data, so as to accurately and efficiently reconstruct the missing electric power system data. The present invention takes into account the reconstruction of missing data of the electric power system, which is of great significance to ensuring the safe and stable operation of the electric power system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and relates to a method for reconstructing missing data of a power system when data loss, data delay, and bad data are injected into the power system due to a network attack, and in particular to a method and system for reconstructing missing data of a power system based on a graph neural network. Background Art

[0002] With the proposal of the dual carbon goals and the increasing proportion of new energy such as wind and solar power connected to the grid in my country, the intelligent perception and dispatching of power systems is an inevitable trend to ensure the safe and stable operation of power systems. The accuracy and real-time performance of data collected by power data acquisition and monitoring systems (SCADA) and phasor measurement units (PMU) are of great significance to power system state estimation, stability analysis and operation optimization. However, due to external factors such as the stability of the communication system and the stability of the uninterruptible power supply, power data may be disturbed or fail in the collection, measurement, transmission, conversion and other links, resulting in abnormal problems such as data loss and data delay in the power system, further affecting the normal operation of the intelligent dispatching center. Therefore, the reconstruction of missing data in the power system is crucial to ensure the safe and stable operation of the power system. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method for reconstructing missing data of an electric power system based on a graph neural network, which utilizes the graph neural network to learn the complex spatiotemporal correlations between measurement data from historical data, thereby accurately and efficiently reconstructing the missing electric power system data.

[0004] To this end, the present invention adopts the following technical solution:

[0005] A method for reconstructing missing data of a power system based on a graph neural network, characterized in that: the missing power system operation data at the current moment is constructed into a matrix and input into a trained graph neural network to obtain the complete power system operation data reconstructed at the current moment;

[0006] The missing power system operation data in the matrix are represented by fixed values ​​far from normal values;

[0007] The graph neural network is trained using historical operating data collected based on the power system.

[0008] Furthermore, the power system operation data includes the voltage amplitude, voltage phase angle, node injected active power, node injected reactive power and connection relationship between two nodes at each node; the constructed matrix includes:

[0009] The topological matrix A and the characteristic electrical characteristic matrix H that represent the connection relationship between two nodes in the power system t :

[0010]

[0011]

[0012] Among them, T ij Indicates the topological connection relationship of node ij. If it is 0, it means that the two nodes are not directly connected, and if it is 1, it means that the two nodes are directly connected. t m,i ,V t a,i ,P t i ,Q t i They represent the voltage amplitude, voltage phase angle, node injected active power and node injected reactive power at node i at time t respectively, and N is the total number of nodes in the power system.

[0013] Furthermore, the graph neural network is composed of several GCN layers and fully connected layers connected in sequence.

[0014] Furthermore, the graph neural network is trained based on historical operation data collected based on the power system, as follows:

[0015] The historical operation data of the power system are collected and constructed into matrices one by one. For the historical data matrix at a specific time section, part of the data is randomly discarded to simulate abnormal problems of data loss and data delay in the power system;

[0016] The matrix after discarding part of the data is used as the input of the graph neural network, and the corresponding original matrix is ​​used as the label. The graph neural network is trained with the goal of minimizing the loss function between the output of the graph neural network and the label to obtain a trained graph neural network.

[0017] Furthermore, in the matrix after discarding part of the data, the discarded data is replaced by a fixed value far from a normal value.

[0018] Furthermore, the loss function is the mean square error between the output of the graph neural network and the label.

[0019] A power system missing data reconstruction system based on graph neural network, comprising:

[0020] A data acquisition module is used to construct a matrix of the missing power system operation data at the current moment;

[0021] The data reconstruction module includes a trained graph neural network, which is used to construct the missing power system operation data at the current moment into a matrix and input it into a trained graph neural network to obtain the complete power system operation data reconstructed at the current moment.

[0022] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for reconstructing missing data of a power system based on a graph neural network is implemented.

[0023] A storage medium containing computer executable instructions, which, when executed by a computer processor, implement the above-mentioned method for reconstructing missing data of a power system based on a graph neural network.

[0024] The beneficial effects of the present invention are as follows: the present invention designs a method and system for reconstructing missing data of a network power system based on graph neural network, describes the state of the power system as a graph structure, and uses graph neural network to learn the complex spatiotemporal correlation between graph structure measurement data from historical data, so as to accurately and efficiently reconstruct the missing power system data. This solves the problem that the power system measurement data may have different degrees of missing data due to various reasons such as communication congestion, hardware failure, and transmission delay in the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The structure diagram of the graph convolutional neural network designed for the present invention;

[0026] Figure 2 The IEEE39 node system structure diagram used in the example;

[0027] Figure 3 This is the change graph of the mean square error indicator during the training process;

[0028] Figure 4 The electrical characteristic parameter matrix to be reconstructed for the power system at a certain moment;

[0029] Figure 5 It is the electrical characteristic parameter matrix of the power system after reconstruction at a certain moment. DETAILED DESCRIPTION

[0030] The present invention provides a method for reconstructing missing data of an electric power system based on a graph neural network, specifically: constructing the missing electric power system operation data at the current moment into a matrix and inputting it into a trained graph neural network to obtain the complete electric power system operation data reconstructed at the current moment.

[0031] The power system operation data includes the voltage amplitude, voltage phase angle, node injected active power, node injected reactive power and the connection relationship between two nodes at each node; the constructed matrix includes:

[0032] The topological matrix A and the characteristic electrical characteristic matrix H that represent the connection relationship between two nodes in the power system t The specific expressions are as follows:

[0033]

[0034]

[0035] Among them, T t ij Represents the topological connection relationship of node ij at time t. If it is 0, it means that the two nodes are not directly connected, and if it is 1, it means that the two nodes are directly connected. t m,i ,V t a,i ,P t i ,Q t i They represent the voltage amplitude, voltage phase angle, node injected active power and node injected reactive power at node i at time t respectively.

[0036] Among them, the missing power system operation data in the matrix are represented by fixed values ​​far from the normal values;

[0037] The graph neural network is trained based on the historical operation data collected by the power system. For example, the historical operation data of the power system are collected and constructed into electrical feature matrices one by one. At the time section t, in the electrical feature matrix H t Randomly replace 1-k electrical characteristics with fixed values ​​e t Replace (e t Far from the normal value), in order to simulate abnormal problems such as power system data loss and data delay. The processed electrical characteristic matrix is ​​recorded as H t 'As the input of the subsequent neural network, the matrix H before processing t As the target value of the subsequent neural network output. Using the mean square error as the loss function, the gradient descent method is used to train the trainable parameters of the neural network to minimize the mean square error, thereby obtaining a trained graph neural network.

[0038] Figure 1 The figure shows a graph convolutional neural network structure diagram. The graph convolutional neural network consists of several GCN layers (2 layers are shown in the figure) connected in sequence and a fully connected layer. The output of the GCN layer is shown in the following formula:

[0039]

[0040]

[0041]

[0042] Where H, A and H' are the input and output of the layer respectively; A is the topology matrix; H is the time series feature matrix of the power system; D is the degree matrix of the power system; I is the identity matrix; W is the parameter matrix that needs to be trained for the GCN layer.

[0043] In the figure, the output of the fully connected layer is shown as follows:

[0044]

[0045] Among them, X FC (l) and Y FC (l) are the input and output of the lth FC layer respectively; W (l) FC and b (l) They are the weight matrix and bias matrix of the lth FC layer, respectively, and are trainable parameters.

[0046] The present invention is further described below in conjunction with specific embodiments.

[0047] Example

[0048] This embodiment provides a method for reconstructing missing data of a power system based on a graph neural network.

[0049] like Figure 2 As shown, this embodiment uses the IEEE39 node system to simulate and verify the effectiveness of the proposed method. Figure 2 As shown, the total load is 6254MW. PMU (phase measurement value and angle) is installed at each node in the 39 system, and data is synchronized to the dispatch center at a frequency of 50Hz. In order to simulate abnormal problems such as power system data loss and data delay, 1-30 sample data are randomly set to 0 to represent abnormal values ​​when processing all moments.

[0050] 1. System data characterization and neural network training:

[0051] According to the above method, the 10 6 The electrical characteristic parameters of the group system are used as training data. Figure 1 The neural network is trained as shown in the figure. The curve of loss function changing with the number of iterations during training is as follows Figure 3 shown.

[0052] 2. Reconstruction of missing data in the power system:

[0053] At a certain time t, the system loses data due to communication failure. At this time, the power system characteristic matrix is ​​as follows: Figure 4 As shown (for ease of observation, missing data is represented by the symbol * in the figure, and is actually represented by a fixed value far from the normal value).

[0054] The characteristic matrix is ​​input into the neural network trained in the previous step to obtain the reconstructed power system characteristic parameter matrix as follows: Figure 5 The results show that the method of the present invention uses graph neural networks to learn the complex spatiotemporal correlations between measurement data from historical data, and can accurately and efficiently reconstruct the missing power system data.

[0055] Corresponding to the aforementioned embodiment of the method for reconstructing missing data of a power system based on a graph neural network, the present invention also provides an embodiment of a system for reconstructing missing data of a power system based on a graph neural network.

[0056] The system includes:

[0057] A data acquisition module is used to construct a matrix of the missing power system operation data at the current moment;

[0058] The data reconstruction module includes a trained graph neural network, which is used to construct the missing power system operation data at the current moment into a matrix and input it into a trained graph neural network to obtain the complete power system operation data reconstructed at the current moment.

[0059] For the system and electronic device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments, which will not be repeated here. The system embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Those of ordinary skill in the art can understand and implement it without paying creative labor.

[0060] Furthermore, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned graph neural network-based method for reconstructing missing data of a power system when executing the computer program.

[0061] The implementation process of the functions and effects in the above electronic device is specifically described in the implementation process of the corresponding steps in the above method, which will not be repeated here.

[0062] An embodiment of the present invention also provides a computer-readable storage medium, and the computer-executable instructions, when executed by a computer processor, implement the aforementioned method for reconstructing missing data of a power system based on a graph neural network.

[0063] The computer-readable storage medium may be an internal storage unit of any device with data processing capability described in any of the aforementioned embodiments, such as a hard disk or a memory. The computer-readable storage medium may also be any device with data processing capability, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capability and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capability, and may also be used to temporarily store data that has been output or is to be output.

[0064] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for reconstructing missing data of a power system based on graph neural network, characterized in that: Specifically, the missing power system operation data at the current moment is constructed into a matrix and input into a trained graph neural network to obtain the complete power system operation data reconstructed at the current moment; The missing power system operation data in the matrix are represented by fixed values ​​far from normal values; The graph neural network is obtained by training based on historical operation data collected based on the power system; The power system operation data includes the voltage amplitude, voltage phase angle, node injected active power, node injected reactive power and the connection relationship between two nodes at each node; the constructed matrix includes: The topological matrix A and the characteristic electrical characteristic matrix H that represent the connection relationship between two nodes in the power system t : Among them, T ij Indicates the topological connection relationship of node ij. If it is 0, it means that the two nodes are not directly connected, and if it is 1, it means that the two nodes are directly connected. t m,i ,V t a,i ,P t i ,Q t i They represent the voltage amplitude, voltage phase angle, node injected active power and node injected reactive power at node i at time t respectively, and N is the total number of nodes in the power system.

2. The method according to claim 1, characterized in that The graph neural network consists of several GCN layers and fully connected layers connected in sequence.

3. The method according to claim 1, characterized in that: The graph neural network is trained based on the historical operation data collected based on the power system, as follows: The historical operation data of the power system are collected and constructed into matrices one by one. For the historical data matrix at a specific time section, part of the data is randomly discarded to simulate abnormal problems of data loss and data delay in the power system; The matrix after discarding part of the data is used as the input of the graph neural network, and the corresponding original matrix is ​​used as the label. The graph neural network is trained with the goal of minimizing the loss function between the output of the graph neural network and the label to obtain a trained graph neural network.

4. The method according to claim 3, characterized in that In the matrix after discarding part of the data, the discarded data is replaced by a fixed value far from the normal value.

5. The method according to claim 3, characterized in that: The loss function is the mean square error between the output of the graph neural network and the label.

6. A power system missing data reconstruction system based on graph neural network, characterized in that: include: A data acquisition module is used to construct a matrix of the missing power system operation data at the current moment; The power system operation data includes the voltage amplitude, voltage phase angle, node injected active power, node injected reactive power and the connection relationship between two nodes at each node; the constructed matrix includes: The topological matrix A and the characteristic electrical characteristic matrix H that represent the connection relationship between two nodes in the power system t : Among them, T ij Indicates the topological connection relationship of node ij. If it is 0, it means that the two nodes are not directly connected, and if it is 1, it means that the two nodes are directly connected. t m,i ,V t a,i ,P t i ,Q t i They represent the voltage amplitude, voltage phase angle, node injected active power and node injected reactive power at node i at time t respectively, and N is the total number of nodes in the power system; The data reconstruction module includes a trained graph neural network, which is used to construct the missing power system operation data at the current moment into a matrix and input it into a trained graph neural network to obtain the complete power system operation data reconstructed at the current moment.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the method for reconstructing missing data of a power system based on a graph neural network as described in any one of claims 1 to 5.

8. A storage medium comprising computer executable instructions, which, when executed by a computer processor, implement the method for reconstructing missing data of a power system based on a graph neural network as described in any one of claims 1 to 5.

Images