A method and system for locating harmonic disturbance in distribution network

Through graph convolution neural network, the simulation data of the distribution network is trained and modeled to correct, which solves the problem of low harmonic source positioning accuracy in the distribution network, and achieves higher positioning accuracy and lower data requirements. Improve the safety and economic benefits of the power system.

CN118657069BActive Publication Date: 2024-11-05ELECTRIC POWER RES INST CHINA SOUTHERN POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411147811.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-05
Estimated Expiration
2044-08-21

AI Technical Summary

Technical Problem

The existing harmonic source positioning methods have low accuracy in the distribution network. Due to the measurement synchronization error and insufficient storage space, the harmonic source positioning is inaccurate, affecting the safety and economic benefits of the power system.

Method used

The graph convolution neural network is used to train the simulation data. By constructing the topological structure and feature data of the distribution network, mining harmonic hidden features, realizing accurate harmonic perturbation positioning, and correcting the feature data obtained by simulation under inherent errors , improve positioning accuracy.

Benefits of technology

It improves the accuracy of harmonic disturbance positioning in the distribution network, reduces dependence on synchronous measurements, reduces data storage requirements, and enhances the safety and economic benefits of the power system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118657069B_ABST
    Figure CN118657069B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of distribution network, and discloses a method and system for locating harmonic disturbances in a distribution network. Simulation data are trained by a graph convolutional neural network, and the hidden relationship between harmonic characteristics of each node in the network topology of the distribution network is mined by the graph convolutional neural network, so as to realize accurate harmonic disturbance positioning under pure simulation conditions, and correct the initial harmonic disturbance positioning model by characteristic data obtained by simulation under inherent error, so as to obtain a digital twin model suitable for locating harmonic disturbances in the distribution network, and further improve the accuracy of harmonic disturbance positioning in the distribution network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution networks, and in particular to a method and system for locating harmonic disturbances in a power distribution network. Background Art

[0002] In recent years, with the large-scale and decentralized access of distributed generation and power electronic equipment to the power grid, the number of harmonic sources in the power grid has increased sharply, the operating status has become more and more variable, and the harmonic pollution problem in the power grid has become more and more serious. The increasingly severe harmonic pollution problem will cause a series of problems, including overheating of electrical equipment, accelerated insulation aging, shortened service life, equipment failure and even combustion. In addition, harmonic problems will also trigger malfunctions of electrical equipment and cause confusion in electricity metering. These consequences not only endanger the safe operation of the power system, but also bring immeasurable economic losses to power users. Therefore, accurately tracing the source of harmonic disturbances in the distribution network can provide guidance for subsequent work such as formulating harmonic control plans. It is an important link in reducing harmonic hazards and improving power quality, and it has very important practical significance.

[0003] Existing harmonic source location methods are mainly divided into two categories: equivalent circuit method and harmonic state estimation method. These traditional methods rely heavily on obtaining accurate phase information of harmonic voltage and current from measuring equipment, that is, requiring the measuring equipment of each node to achieve synchronous measurement. However, any deviation in measurement synchronization will cause uncertain time errors between the harmonic voltage and current recorded at each node. For high-order harmonics, even microsecond errors can cause more serious phase errors. In addition, due to factors such as insufficient long-term monitoring storage space and delays in harmonic voltage and current synchronization, the harmonic data recorded by existing domestic power quality monitoring equipment are usually composed of statistical values. These data do not have the characteristics of real-time and synchronization, which will affect the accuracy of these methods in actual harmonic source location applications in distribution networks. Summary of the invention

[0004] The present invention provides a method and system for locating harmonic disturbances in a distribution network, which solves the technical problem of low accuracy in locating harmonic sources in the distribution network.

[0005] In view of this, a first aspect of the present invention provides a method for locating harmonic disturbances in a distribution network, comprising:

[0006] Acquire first distribution network simulation data under a plurality of different operating conditions, and extract first characteristic data of the first distribution network simulation data;

[0007] Using the plurality of the first feature data and the corresponding harmonic disturbance source positions, a graph convolutional neural network is trained to obtain an initial harmonic disturbance positioning model;

[0008] Acquire second distribution network simulation data under a plurality of different operating conditions, and extract second characteristic data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent error;

[0009] Using a plurality of the second characteristic data and the corresponding harmonic disturbance source positions, the initial harmonic disturbance positioning model is tested, and the initial harmonic disturbance positioning model is corrected using the test results to obtain a harmonic disturbance positioning model;

[0010] The distribution network operation data acquired in the current calculation period is input into the harmonic disturbance location model, and the harmonic disturbance source position in the distribution network in the current calculation period is output.

[0011] Optionally, the first distribution network simulation data is obtained by performing simulation tests based on a pre-built distribution network simulation model.

[0012] Optionally, the step of training a graph convolutional neural network using a plurality of the first feature data and their corresponding harmonic disturbance source positions, and obtaining an initial harmonic disturbance positioning model through training, comprises:

[0013] Constructing a graph structure of a graph convolutional neural network, wherein the graph structure is represented by G=(V, E, M), where G is the graph structure state, V is the node set, E is the edge set of the distribution network topology, and M is the adjacency matrix of the distribution network topology;

[0014] An electrical connection matrix is ​​constructed by the electrical distance between every two adjacent nodes in the distribution network, and the electrical connection matrix is ​​determined as an adjacency matrix of the distribution network topology;

[0015] The first feature data is used as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, the first feature data of the node and the first feature data of its adjacent nodes are feature aggregated in a graph form to obtain a fused spatial feature graph structure;

[0016] Mining the harmonic hidden features from the fused spatial feature graph structure, using the harmonic hidden features to classify all nodes, obtaining harmonic source nodes and non-harmonic source nodes;

[0017] Judging whether the classification accuracy of the node reaches a preset classification accuracy threshold according to the harmonic disturbance source position corresponding to the first feature data;

[0018] When the classification accuracy of the judgment node does not reach the preset classification accuracy threshold, the step of using the first feature data as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, performing feature aggregation on the first feature data of the node and the first feature data of its adjacent nodes in a graph form to obtain a fused spatial feature graph structure, until the classification accuracy of the judgment node reaches the preset classification accuracy threshold, stopping the iterative training, and training and outputting the initial harmonic disturbance positioning model.

[0019] Optionally, the graph convolutional neural network includes an input layer, multiple hidden layers and an output layer connected in sequence, wherein the propagation expression between layers in the multiple hidden layers is:

[0020]

[0021] In the formula, is the current hidden layer, (·) is a nonlinear activation function, is the degree matrix, is the diagonal adjacency matrix, is the previous hidden layer, is the weight, where

[0022]

[0023] In the formula, I n is the n-dimensional identity matrix.

[0024] Optionally, the second characteristic data is obtained by performing simulation tests on a distribution network simulation model built on an RT-LAB semi-physical simulation platform.

[0025] Optionally, the first characteristic data and the second characteristic data both adopt one or more of fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate and harmonic power.

[0026] In a second aspect, the present invention further provides a distribution network harmonic disturbance location system, comprising:

[0027] A first feature extraction module, used to obtain first distribution network simulation data under a plurality of different operating conditions, and extract first feature data of the first distribution network simulation data;

[0028] A model training module, used to train a graph convolutional neural network using the plurality of first feature data and their corresponding harmonic disturbance source positions, and to obtain an initial harmonic disturbance positioning model through training;

[0029] A second feature extraction module is used to obtain second distribution network simulation data under a plurality of different operating conditions, and extract second feature data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent error;

[0030] A model correction module, used for testing the initial harmonic disturbance positioning model by using a plurality of the second feature data and the corresponding harmonic disturbance source positions, and correcting the initial harmonic disturbance positioning model by using the test results to obtain the harmonic disturbance positioning model;

[0031] The disturbance locating module is used to input the distribution network operation data acquired in the current calculation period into the harmonic disturbance locating model, and output the harmonic disturbance source position in the distribution network in the current calculation period.

[0032] In a third aspect, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for locating harmonic disturbances in a distribution network as described in the first aspect.

[0033] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the steps of the method for locating harmonic disturbances in a distribution network as described in the first aspect are implemented.

[0034] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the method for locating harmonic disturbances in a distribution network as described in the first aspect.

[0035] It can be seen from the above technical solutions that the present invention has the following advantages:

[0036] The present invention trains simulation data through a graph convolutional neural network, and uses the graph convolutional neural network to mine the hidden relationship between the harmonic characteristics of each node in the distribution network topology, so as to achieve accurate harmonic disturbance positioning under pure simulation conditions, and correct the initial harmonic disturbance positioning model through the characteristic data obtained by simulation under inherent error, so as to obtain a digital twin model suitable for harmonic disturbance positioning of the distribution network, thereby further improving the accuracy of harmonic disturbance positioning of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of a method for locating harmonic disturbances in a distribution network provided by an embodiment of the present invention;

[0038] Figure 2A schematic diagram of the structure of a graph convolutional neural network provided by an embodiment of the present invention;

[0039] Figure 3 The equivalent topology of the IEEE 14 distribution network provided by the embodiment of the present invention;

[0040] Figure 4 A schematic diagram of the structure of a distribution network harmonic disturbance location system provided by an embodiment of the present invention;

[0041] Figure 5 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] An embodiment of the present invention provides a method for locating a harmonic disturbance in a distribution network. The embodiment of the present invention can be applicable to the situation of locating a harmonic disturbance in a distribution network. The method can be executed by a distribution network harmonic disturbance locating device. The distribution network harmonic disturbance locating device can be implemented in the form of hardware and / or software. The distribution network harmonic disturbance locating device can be configured in a computer device.

[0044] like Figure 1 As shown, an embodiment of the present invention provides a method for locating harmonic disturbances in a distribution network, comprising:

[0045] Step S1, obtaining first distribution network simulation data under a plurality of different operating conditions, and extracting first characteristic data of the first distribution network simulation data.

[0046] Among them, the first distribution network simulation data is obtained by simulation testing based on a pre-built distribution network simulation model.

[0047] In some embodiments, by obtaining distribution network topology information and related line parameters, and building a distribution network simulation model, the distribution network simulation model is loaded with harmonic current sources with different harmonic current amplitudes by setting different operating conditions, and the first distribution network simulation data under multiple different operating conditions is obtained.

[0048] Among them, the first characteristic data adopts one or more of the fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate and harmonic power of each node, and the first characteristic data are all average values ​​or effective values.

[0049] Step S2: Use multiple first feature data and their corresponding harmonic disturbance source positions to train the graph convolutional neural network, and train to obtain an initial harmonic disturbance positioning model.

[0050] In a graph convolutional neural network (GCN), two adjacent nodes are connected by edges to construct a graph. It extends the traditional convolution operation from Euclidean structure data (such as images, grids, etc.) to non-Euclidean structure data (graph structure data). From the spatial characteristics of the distribution network, its network topology is an irregular non-Euclidean spatial structure, which is essentially similar to a graph structure. Figure 2 As shown in Figure 1, the graph convolutional neural network includes an input layer, multiple hidden layers, and an output layer connected in sequence. However, it is worth noting that the distribution network is different from other graph structures. The degree of correlation between nodes in the distribution network is affected by the electrical correlation and impedance characteristics between the two. Establishing the graph structure of the distribution network based solely on a given topology or line length cannot fully reflect the degree of electrical and physical correlation between nodes.

[0051] Specifically, step S2 includes:

[0052] Step S201, construct a graph structure of a graph convolutional neural network, the graph structure is expressed as G=(V, E, M), where G is the graph structure state, V is the node set, E is the edge set of the distribution network topology, and M is the adjacency matrix of the distribution network topology.

[0053] Step S202: construct an electrical connection matrix by using the electrical distance between every two adjacent nodes in the distribution network, and determine the electrical connection matrix as an adjacency matrix of the distribution network topology.

[0054] It should be noted that in order to better capture the electrical connection characteristics between nodes in the power topology, the present invention uses an electrical connection matrix to describe the characteristics of the distribution network spatial topology, that is, the adjacency matrix M of the distribution network topology as a graph structure.

[0055] For an n-node distribution network, the power distance d between two adjacent nodes ij It can be expressed as:

[0056]

[0057] Where, X ii represents the self-impedance of the ith node, and X ij represents the mutual impedance from the i-th node to the j-th node.

[0058] Distance to power d ij After normalization, we can get:

[0059]

[0060] In the formula, is the normalized electric distance d ij .

[0061] The electrical connection matrix M is constructed as:

[0062]

[0063] Step S203: using the first feature data as the input of the graph convolutional neural network, based on the adjacency matrix of the distribution network topology, for each node, the first feature data of the node and the first feature data of its adjacent nodes are feature aggregated in a graph form to obtain a fused spatial feature graph structure.

[0064] Among them, the present invention applies GCN to the location of harmonic disturbances in distribution networks, and takes the characteristic data of electrical quantities (including voltage, current, power, etc.) of different nodes as input and inputs them into the GCN neural network. When the GCN neural network is running, it will feature aggregate the characteristic information of the electrical quantities (including voltage, current, power, etc.) of the current node with the characteristic data of its adjacent nodes.

[0065] Step S204: mining the harmonic hidden features from the fused spatial feature graph structure, and using the harmonic hidden features to classify all nodes to obtain harmonic source nodes and non-harmonic source nodes.

[0066] Among them, harmonic hidden features refer to features that can reveal the intrinsic structural characteristics of the graph. These features are obtained by spectral analysis of spatial feature graphs through spectral graph theory. Specifically, the features that nodes connected in the graph have but not connected in the graph are mined as harmonic hidden features.

[0067] The features that the harmonic source nodes in the spatial feature graph have but the non-harmonic source nodes do not have are classified, and the classification becomes more accurate with the continuous training process.

[0068] Step S205: judging whether the classification accuracy of the node reaches a preset classification accuracy threshold according to the harmonic disturbance source position corresponding to the first characteristic data;

[0069] It can be understood that, for each node, there is a target value of the harmonic disturbance source position corresponding to the first characteristic data, and the target value can be used to determine whether the classification result is accurate and calculate the classification accuracy of all nodes.

[0070] Step S206: When the classification accuracy of the judgment node does not reach the preset classification accuracy threshold, go to step S203 until the classification accuracy of the judgment node reaches the preset classification accuracy threshold, stop iterative training, and train and output the initial harmonic disturbance positioning model.

[0071] Among them, the graph convolutional neural network, in which the propagation expression between layers in multiple hidden layers is:

[0072]

[0073] In the formula, is the current hidden layer, (·) is a nonlinear activation function, which can be a ReLU function. is the degree matrix, is a diagonal matrix, is the diagonal adjacency matrix, is the previous hidden layer, is the weight, where

[0074]

[0075] In the formula, I n It is an n-dimensional identity matrix to avoid ignoring the characteristics of the node itself when multiplying it with the feature matrix.

[0076] When a graph is input, after passing through the hidden layer, the features of each node are transformed from X to Z through forward propagation, and Y is represented as the node label. The forward propagation formula can be expressed as:

[0077]

[0078] Softmax is the activation function, W is the trainable weight, superscript 0 is the initial weight, superscript l is the weight of the lth layer, X is the feature matrix, is the standardized adjacency matrix, where

[0079]

[0080] The goal of harmonic disturbance location is to determine the location of the hth harmonic source in the distribution network. This task can be expressed as a node classification problem. In order to determine the location of the harmonic source, in an embodiment of the present invention, the graph convolutional neural network includes a fully connected layer, which is the output positioning layer. The input of the fully connected layer consists of multiple units, and the predicted output also consists of multiple units. Each output value represents the possibility of the existence of a harmonic disturbance source at the node, represented by Y. When the model is trained to the accuracy of harmonic disturbance location reaching the requirement, the initial harmonic disturbance location model based on the graph convolutional neural network is trained.

[0081] Step S3, obtaining second distribution network simulation data under a plurality of different operating conditions, and extracting second characteristic data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent errors.

[0082] Among them, the second characteristic data is obtained by simulation testing on the distribution network simulation model built on the RT-LAB semi-physical simulation platform.

[0083] In some embodiments, the distribution network netlist model and real-time simulation model required for running the RT-LAB semi-physical simulation platform can realize high-speed real-time simulation of circuit topology on FPGA. The RT-LAB semi-physical simulation platform consists of a host computer, a slave computer and a control device, wherein the host computer realizes the development and modification of the distribution network topology, the slave computer is a real-time simulation machine, including a CPU and FPGA, and the control device can realize hardware-in-the-loop real-time simulation.

[0084] By running the RT-LAB semi-physical simulation platform, the voltage and current of each node can be displayed by an external oscilloscope, and the operation data of each working condition can be recorded and saved as the second distribution network simulation data. The second distribution network simulation data of each working condition is feature extracted by FFT analysis and other processing methods to obtain the second feature data. The second feature data uses one or more of fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate and harmonic power.

[0085] Since experiments on the RT-LAB semi-physical simulation platform will have more interference and inherent errors of equipment such as oscilloscopes than pure simulation models, considering the impact of inherent errors on positioning in the model is more in line with the actual engineering situation.

[0086] Step S4: using the plurality of second characteristic data and their corresponding harmonic disturbance source positions to test the initial harmonic disturbance positioning model, and using the test results to correct the initial harmonic disturbance positioning model to obtain the harmonic disturbance positioning model.

[0087] It can be understood that by inputting multiple second feature data into the initial harmonic disturbance positioning model, the output harmonic disturbance source position is compared with the target value of the original harmonic disturbance source position to determine the test accuracy, and the test accuracy is compared with the preset test accuracy threshold. If the test accuracy is lower than the preset test accuracy threshold, the network parameters of the initial harmonic disturbance positioning model are optimized until the test accuracy is not lower than the preset test accuracy threshold, and the optimized harmonic disturbance positioning model is output.

[0088] Step S5: input the distribution network operation data acquired in the current calculation period into the harmonic disturbance location model, and output the harmonic disturbance source position in the distribution network in the current calculation period.

[0089] Among them, by extracting features from the distribution network operation data obtained in the current calculation period, the extracted data is input into the harmonic disturbance positioning model, and the harmonic disturbance source position in the distribution network in the current calculation period is output.

[0090] It should be noted that the embodiment of the present invention trains simulation data through a graph convolutional neural network, and uses the graph convolutional neural network to mine the hidden relationship between the harmonic characteristics of each node in the distribution network topology, so as to achieve accurate harmonic disturbance positioning under pure simulation conditions, and correct the initial harmonic disturbance positioning model through the characteristic data obtained by simulation under inherent error, so as to obtain a digital twin model suitable for harmonic disturbance positioning of the distribution network, thereby further improving the accuracy of harmonic disturbance positioning of the distribution network.

[0091] In addition, in the context of insufficient storage space for long-term monitoring of power quality monitoring devices and delays in harmonic voltage and current synchronization, the embodiments of the present invention provide a method that only requires steady-state data that can be obtained in engineering practice, without the need for voltage and current delay-free synchronization data, with less data requirements, more accurate results and higher feasibility.

[0092] The following is a calculation example provided in conjunction with a method for locating harmonic disturbances in a distribution network provided in an embodiment of the present invention.

[0093] In this example, the modified IEEE 14 distribution network is taken as an example. The equivalent topology of the IEEE 14 distribution network is as follows: Figure 3 As shown in the figure, the load includes new energy sources such as photovoltaics and some harmonic emission devices. The output of new energy is random and due to the existence of high-frequency power electronic converters when connected to the grid, new energy is presented in two forms under different working conditions, namely harmonic source and no-load.

[0094] Building in Matlab / Simulink Figure 3 In the circuit simulation model shown in the figure, voltmeters and ammeters are added to each node and different types of operating conditions are set for simulation to obtain a simulation sample library. For example, new energy and harmonic wave transmitters are connected to some nodes, and each node is connected to an RL load, that is, no harmonics are emitted but there is active power and reactive power consumption. The node voltages and branch currents collected under various operating conditions in the sample library are feature extracted to extract the steady-state fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate, and harmonic power of each node under each operating condition.

[0095] The feature sets of each node under different working conditions in the simulation and the target set of harmonic disturbance source positions are used as samples and the established distribution network adjacency matrix is ​​input into the harmonic disturbance positioning model based on graph convolutional neural network for training. The present invention takes a single and three harmonic sources as embodiments, and names them as scenario 1 and scenario 2, generates 1200 samples respectively, and randomly shuffles them, making 60% of them as training set samples, 20% as validation set samples, and the remaining 20% ​​as model test set samples. The results of harmonic disturbance source positioning of the test set samples using the model of the present invention are shown in Table 1.

[0096] Table 1

[0097]

[0098] It can be seen from Table 1 that the harmonic disturbance positioning model based on graph convolutional neural network has a higher accuracy rate, and the harmonic disturbance positioning accuracy rates under the two working conditions are both above 95%.

[0099] Subsequently, the distribution network netlist model and real-time simulation model required for running the RT-LAB semi-physical simulation platform are built, and different working conditions are set similarly to the pure simulation model. However, since the model only needs to be corrected at this time, this embodiment generates 200 working condition samples. In the experiment, the semi-physical simulation platform is connected to an oscilloscope to collect the node voltage and branch current of each node, and the data is processed to extract the characteristic quantities under each working condition. The feature data set is input into the trained harmonic disturbance location model based on graph convolutional neural network, and the initial pure simulation model is corrected to form a harmonic disturbance location model in the distribution network based on digital twin. Taking scenario 2 of multiple harmonic sources as an example, 10 groups of working conditions are generated using the RT-LAB semi-physical simulation platform, and the model is verified. The verification results are shown in Table 2.

[0100] Table 2

[0101]

[0102] From the data in Table 2, it can be seen that among the 10 working conditions, 9 groups of positioning are completely accurate, and 1 group has a wrong positioning of 1 harmonic disturbance source, but the overall harmonic disturbance positioning accuracy is high, reaching more than 90%. Even if the input characteristics such as the harmonic disturbance source position and the harmonic emission value change randomly and are provided as steady-state values, the method provided by the embodiment of the present invention can quickly and accurately locate the harmonic disturbance source.

[0103] Based on the same inventive concept, an embodiment of the present application also provides a distribution network harmonic disturbance locating system for implementing the above-mentioned distribution network harmonic disturbance locating method. The implementation solution for solving the problem provided by the system is similar to the implementation solution recorded in the above-mentioned method. Therefore, the specific limitations in one or more distribution network harmonic disturbance locating system embodiments provided below can be referred to the limitations on the distribution network harmonic disturbance locating method in the above text, and will not be repeated here.

[0104] like Figure 4 As shown, a distribution network harmonic disturbance positioning system provided in an embodiment of the present application includes:

[0105] A first feature extraction module 100 is used to obtain first distribution network simulation data under a plurality of different operating conditions and extract first feature data of the first distribution network simulation data;

[0106] The first distribution network simulation data is obtained by performing simulation tests based on a pre-built distribution network simulation model.

[0107] The model training module 200 is used to train the graph convolutional neural network using multiple first feature data and their corresponding harmonic disturbance source positions, and to obtain an initial harmonic disturbance positioning model through training.

[0108] The model training module 200 is specifically used to construct a graph structure of a graph convolutional neural network, where the graph structure is represented by G=(V,E,M), where G is the graph structure state, V is a node set, E is an edge set of the distribution network topology, and M is an adjacency matrix of the distribution network topology; an electrical connection matrix is ​​constructed by the electrical distance between every two adjacent nodes in the distribution network, and the electrical connection matrix is ​​determined as the adjacency matrix of the distribution network topology; the first feature data is used as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, the first feature data of the node and the first feature data of its adjacent nodes are feature aggregated in a graph form to obtain a fused spatial feature graph structure; the harmonic hidden features are mined from the fused spatial feature graph structure, and all nodes are classified using the harmonic hidden features. Obtain harmonic source nodes and non-harmonic source nodes; determine whether the classification accuracy of the node reaches a preset classification accuracy threshold according to the harmonic disturbance source position corresponding to the first feature data; when it is determined that the classification accuracy of the node does not reach the preset classification accuracy threshold, use the first feature data as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, perform feature aggregation on the first feature data of the node and the first feature data of its adjacent nodes in a graph form to obtain a fused spatial feature graph structure, until the classification accuracy of the node is determined to reach the preset classification accuracy threshold, stop iterative training, and train and output an initial harmonic disturbance positioning model.

[0109] The graph convolutional neural network includes an input layer, multiple hidden layers, and an output layer connected in sequence, wherein the propagation expression between layers in the multiple hidden layers is:

[0110]

[0111] In the formula, is the current hidden layer, (·) is a nonlinear activation function, is the degree matrix, is the diagonal adjacency matrix, is the previous hidden layer, is the weight, where

[0112]

[0113] In the formula, I n is the n-dimensional identity matrix.

[0114] A second feature extraction module 300 is used to obtain second distribution network simulation data under a plurality of different operating conditions and extract second feature data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent error;

[0115] The second characteristic data is obtained by simulation testing on the distribution network simulation model built on the RT-LAB semi-physical simulation platform.

[0116] The first characteristic data and the second characteristic data both adopt one or more of fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate and harmonic power.

[0117] The model correction module 400 is used to test the initial harmonic disturbance positioning model using the plurality of second characteristic data and the corresponding harmonic disturbance source positions, and to correct the initial harmonic disturbance positioning model using the test results to obtain the harmonic disturbance positioning model;

[0118] The disturbance locating module 500 is used to input the distribution network operation data acquired in the current calculation period into the harmonic disturbance locating model, and output the harmonic disturbance source position in the distribution network in the current calculation period.

[0119] like Figure 5 As shown, an embodiment of the present invention further provides an electronic device 10, including a memory 20 and a processor 30, wherein a computer program is stored in the memory 20, and when the computer program is executed by the processor 30, the processor 30 executes the steps of the distribution network harmonic disturbance locating method as in any of the above embodiments.

[0120] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed, the steps of the method for locating harmonic disturbances in a power distribution network as described in any of the above embodiments are implemented.

[0121] An embodiment of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the distribution network harmonic disturbance locating method in any of the above embodiments.

[0122] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0123] In several embodiments provided by the present invention, it is understood that each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and a part of a module, a program segment or a code includes one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.

[0124] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, electronic devices, computer storage media and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0125] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0126] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0127] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for executing all or part of the steps of the method described in each embodiment of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0128] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating harmonic disturbances in distribution networks. It is characterized in that include: Acquire first distribution network simulation data under a plurality of different operating conditions, and extract first characteristic data of the first distribution network simulation data; The graph convolutional neural network is trained using the plurality of the first feature data and the corresponding harmonic disturbance source positions, and an initial harmonic disturbance positioning model is obtained through training, including: Constructing a graph structure of a graph convolutional neural network, wherein the graph structure is represented by G=(V, E, M), where G is the graph structure state, V is the node set, E is the edge set of the distribution network topology, and M is the adjacency matrix of the distribution network topology; The graph convolutional neural network includes an input layer, multiple hidden layers and an output layer connected in sequence, wherein the propagation expression between layers in the multiple hidden layers is: ; In the formula, is the current hidden layer, is a nonlinear activation function, is the degree matrix, is the diagonal adjacency matrix, is the previous hidden layer, is the weight, where ; In the formula, I n is the n-dimensional identity matrix; An electrical connection matrix is ​​constructed by the electrical distance between every two adjacent nodes in the distribution network, and the electrical connection matrix is ​​determined as an adjacency matrix of the distribution network topology; The first feature data is used as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, the first feature data of the node and the first feature data of its adjacent nodes are feature aggregated in a graph form to obtain a fused spatial feature graph structure; Mining the harmonic hidden features from the fused spatial feature graph structure, using the harmonic hidden features to classify all nodes, obtaining harmonic source nodes and non-harmonic source nodes; Judging whether the classification accuracy of the node reaches a preset classification accuracy threshold according to the harmonic disturbance source position corresponding to the first feature data; When the classification accuracy of the judgment node does not reach the preset classification accuracy threshold, the step of taking the first feature data as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, performing feature aggregation on the first feature data of the node and the first feature data of its adjacent nodes in a graph form to obtain a fused spatial feature graph structure, until the classification accuracy of the judgment node reaches the preset classification accuracy threshold, stopping the iterative training, and training and outputting the initial harmonic disturbance positioning model; Acquire second distribution network simulation data under a plurality of different operating conditions, and extract second characteristic data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent error; Using a plurality of the second characteristic data and the corresponding harmonic disturbance source positions, the initial harmonic disturbance positioning model is tested, and the initial harmonic disturbance positioning model is corrected using the test results to obtain a harmonic disturbance positioning model; The distribution network operation data acquired in the current calculation period is input into the harmonic disturbance location model, and the harmonic disturbance source position in the distribution network in the current calculation period is output.

2. The method for locating harmonic disturbances in a distribution network according to claim 1, It is characterized in that The first distribution network simulation data is obtained by performing simulation tests based on a pre-built distribution network simulation model.

3. The method for locating harmonic disturbances in a distribution network according to claim 1, It is characterized in that The second characteristic data is obtained by simulation testing on a distribution network simulation model built on the RT-LAB semi-physical simulation platform.

4. The method for locating harmonic disturbance in a distribution network according to any one of claims 1 to 3, It is characterized in that The first characteristic data and the second characteristic data both adopt one or more of fundamental voltage, fundamental current, harmonic voltage, harmonic current, voltage distortion rate, current distortion rate and harmonic power.

5. A distribution network harmonic disturbance location system, It is characterized in that include: A first feature extraction module, used to obtain first distribution network simulation data under a plurality of different operating conditions, and extract first feature data of the first distribution network simulation data; A model training module, used to train a graph convolutional neural network using the plurality of first feature data and their corresponding harmonic disturbance source positions, and to obtain an initial harmonic disturbance positioning model through training, comprising: Constructing a graph structure of a graph convolutional neural network, wherein the graph structure is represented by G=(V, E, M), where G is the graph structure state, V is the node set, E is the edge set of the distribution network topology, and M is the adjacency matrix of the distribution network topology; The graph convolutional neural network includes an input layer, multiple hidden layers and an output layer connected in sequence, wherein the propagation expression between layers in the multiple hidden layers is: ; In the formula, is the current hidden layer, is a nonlinear activation function, is the degree matrix, is the diagonal adjacency matrix, is the previous hidden layer, is the weight, where ; In the formula, I n is the n-dimensional identity matrix; An electrical connection matrix is ​​constructed by the electrical distance between every two adjacent nodes in the distribution network, and the electrical connection matrix is ​​determined as an adjacency matrix of the distribution network topology; The first feature data is used as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, the first feature data of the node and the first feature data of its adjacent nodes are feature aggregated in a graph form to obtain a fused spatial feature graph structure; Mining the harmonic hidden features from the fused spatial feature graph structure, using the harmonic hidden features to classify all nodes, obtaining harmonic source nodes and non-harmonic source nodes; Judging whether the classification accuracy of the node reaches a preset classification accuracy threshold according to the harmonic disturbance source position corresponding to the first feature data; When the classification accuracy of the judgment node does not reach the preset classification accuracy threshold, the step of taking the first feature data as the input of the graph convolutional neural network, and based on the adjacency matrix of the distribution network topology, for each node, performing feature aggregation on the first feature data of the node and the first feature data of its adjacent nodes in a graph form to obtain a fused spatial feature graph structure, until the classification accuracy of the judgment node reaches the preset classification accuracy threshold, stopping the iterative training, and training and outputting the initial harmonic disturbance positioning model; A second feature extraction module is used to obtain second distribution network simulation data under a plurality of different operating conditions, and extract second feature data of the second distribution network simulation data, wherein the second distribution network simulation data is obtained by simulation under inherent error; A model correction module, used for testing the initial harmonic disturbance positioning model by using a plurality of the second feature data and the corresponding harmonic disturbance source positions, and correcting the initial harmonic disturbance positioning model by using the test results to obtain the harmonic disturbance positioning model; The disturbance locating module is used to input the distribution network operation data acquired in the current calculation period into the harmonic disturbance locating model, and output the harmonic disturbance source position in the distribution network in the current calculation period.

6. An electronic device, It is characterized in that It comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for locating harmonic disturbances in a distribution network as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed, the steps of the method for locating harmonic disturbances in a power distribution network as claimed in any one of claims 1 to 4 are implemented.

8. A computer program product, It is characterized in that The computer program product comprises a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises program instructions, wherein when the program instructions are executed by a computer, the computer is caused to perform the steps of the method for locating harmonic disturbances in a distribution network as described in any one of claims 1 to 4.