Power transmission network fault positioning method, device, equipment, storage medium and chip

Abstract

The application discloses a power transmission network fault positioning method, device, equipment, storage medium and chip, and belongs to the technical field of power grids. The power transmission network comprises a plurality of power transmission lines, and a traveling wave measuring point is arranged at both ends of each power transmission line. The method comprises the following steps: determining a wide-area traveling wave topology of the power transmission network, wherein the wide-area traveling wave topology takes the traveling wave measuring point as a node, takes the power transmission line as an edge, each traveling wave measuring point corresponds to a traveling wave characteristic, and the traveling wave characteristic is obtained by modal decomposition on a traveling wave signal collected by the traveling wave measuring point; based on the wide-area traveling wave topology, the fault of each power transmission line is identified through a graph recognition network, and the fault line is positioned; and the traveling wave measuring points at both ends of the fault line are taken as the starting points, a plurality of ranging lines related to the fault line are determined, and the fault position is positioned in the plurality of ranging lines. The application improves the fault identification accuracy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power grid, and particularly relates to a power transmission network fault positioning method, device, equipment, storage medium and chip. BACKGROUND

[0002] In the process of fault diagnosis of power transmission network, traveling wave signals are widely used for fault detection and ranging due to their fast propagation speed and high positioning accuracy.

[0003] Traditional traveling wave fault positioning methods include single-end ranging method and double-end ranging method. The single-end ranging method uses the time difference between the initial traveling wave detected at a single end and the reflected traveling wave at the fault point to calculate the fault location. However, in a complex line network, the reflected traveling wave is reflected multiple times, and the source of the wave front is difficult to accurately identify. The double-end ranging method determines the fault point position by the time difference between the initial traveling waves arriving at the two ends of the line. However, when the distance between the two ends is long, the ranging accuracy will decrease due to the line mode wave speed attenuation, and the accuracy of the collected data at the measuring points is high. Failure to capture the wave front at any end measuring point will result in positioning failure.

[0004] Therefore, how to improve the accuracy of fault positioning in a complex power transmission network environment is a problem to be solved. SUMMARY

[0005] The present application aims to at least solve one of the technical problems in the related art. To this end, the present application provides a power transmission network fault positioning method, device, equipment, storage medium and chip to improve the accuracy of fault positioning.

[0006] In a first aspect, the present application provides a power transmission network fault positioning method, the power transmission network comprising a plurality of power transmission lines, each power transmission line having traveling wave measuring points at both ends; the method comprising:

[0007] determining a wide-area traveling wave topology of the power transmission network, the wide-area traveling wave topology having the traveling wave measuring points as nodes and the power transmission lines as edges, each traveling wave measuring point corresponding to a traveling wave feature, the traveling wave feature being obtained by modal decomposition of a traveling wave signal collected at the traveling wave measuring point;

[0008] based on the wide-area traveling wave topology, identifying faults of each power transmission line through a graph recognition network, and locating the fault line;

[0009] taking the traveling wave measuring points at both ends of the fault line as the starting point, determining a plurality of ranging lines related to the fault line, and locating the fault position in the plurality of ranging lines.

[0010] In the technical solution, the wide-area traveling wave graph data structure is constructed, the traveling wave measuring point is taken as a node, the overhead line is taken as an edge, the graph data structure is used to represent the topology relationship of the power transmission network, and basic data support is provided for fault location; the wide-area traveling wave topology is taken as input, the learnable graph recognition network is used to mine the wide-area traveling wave features and learn the nonlinear relationship between the multiple measuring points and the fault line, so that the fault line recognition is more accurate, and especially in the power grid with multiple T-connection structures, the robustness of fault line recognition is greatly improved; finally, multiple distance measurement lines are formed based on the double-end measuring points of the fault line, the topology range of fault location is further reduced, the multi-link distance measurement fusion mechanism is used for fault location, the error of a single distance measurement path is avoided, and the accuracy and stability of distance measurement are improved.

[0011] According to one embodiment of the application, based on the wide-area traveling wave topology, the fault of each power transmission line is recognized respectively by the graph recognition network, and the fault line is located, comprising:

[0012] Based on the correlation relationship between each traveling wave measuring point in the wide-area traveling wave topology, the adjacent measuring points of each traveling wave measuring point are determined respectively;

[0013] Based on the traveling wave features of each traveling wave measuring point and the adjacent measuring points of each traveling wave measuring point, global features are generated;

[0014] The importance of each traveling wave measuring point and its adjacent measuring points is determined respectively, and multiple positioning features are generated based on the importance;

[0015] The global features and the positioning features are fused to generate target features corresponding to each power transmission line;

[0016] Based on the target features, the fault of each power transmission line is recognized, and the determination result of each power transmission line is obtained, and based on the determination result of each power transmission line, the fault line is located.

[0017] In the above embodiment, by combining the topology of the power transmission network, the global fault propagation feature extraction is realized, the robustness of fault detection is improved, and by determining the importance, the key measuring points are highlighted, so that the accuracy of fault recognition is improved.

[0018] In a second aspect, the application provides a power transmission network fault location device, the power transmission network comprising multiple power transmission lines, and each power transmission line having traveling wave measuring points arranged at both ends; the device comprising:

[0019] A topology module is configured to determine a wide-area traveling wave topology of the power transmission network, the wide-area traveling wave topology taking the traveling wave measuring points as nodes and the power transmission lines as edges, each traveling wave measuring point corresponding to a traveling wave feature, the traveling wave feature being obtained by modal decomposition on a traveling wave signal collected by the traveling wave measuring point;

[0020] The identification module is used to identify faults in each transmission line and locate the faulty line based on the wide-area traveling wave topology and through a graph identification network.

[0021] The positioning module is used to determine multiple ranging lines related to the faulty line, starting from the traveling wave measurement points at both ends of the faulty line, and to locate the fault location among the multiple ranging lines.

[0022] In the above technical solution, a wide-area traveling wave graph data structure is constructed, with traveling wave measurement points as nodes and overhead lines as edges. The graph data structure represents the transmission network topology, providing basic data support for fault location. Using the wide-area traveling wave topology as input, a learnable graph recognition network is proposed to mine wide-area traveling wave features and learn the nonlinear relationship between multiple measurement points and the faulty line. This enables more accurate fault line identification, especially in power grids with multiple T-connections, significantly improving the robustness of fault line identification. Finally, multiple ranging lines are formed based on the two-end measurement points of the faulty line, further narrowing the topological range for fault location. A multi-link ranging fusion mechanism is used for fault location, avoiding errors from a single ranging path and improving the accuracy and stability of ranging.

[0023] Thirdly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power transmission network fault location method as described in the first aspect above.

[0024] Fourthly, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power grid fault location method as described in the first aspect above.

[0025] Fifthly, the present invention provides a chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the power grid fault location method as described in the first aspect above.

[0026] In a sixth aspect, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the power grid fault location method as described in the first aspect above.

[0027] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0028] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0029] Figure 1 is a flowchart of a power grid fault location method provided in some embodiments of the present application;

[0030] Figure 2 is a structural diagram of a typical traveling wave topology provided in some embodiments of the present application;

[0031] Figure 3 is a structural diagram of a wide-area traveling wave topology provided in some embodiments of the present application;

[0032] Figure 4 is a flowchart of variational modal decomposition provided in some embodiments of the present application;

[0033] Figure 5 is a principle diagram of a graph recognition network provided in some embodiments of the present application;

[0034] Figure 6 is a principle diagram of multi-link ranging provided in some embodiments of the present application;

[0035] Figure 7 is a structural diagram of a power grid fault location device provided in some embodiments of the present application;

[0036] Figure 8 is a structural diagram of a computer device provided in some embodiments of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be clearly described below with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of the present application.

[0038] Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above description of drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the specification and claims of the present application and the above description of drawings are used to distinguish different objects, not to describe a particular order or primary and secondary relationships.

[0039] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this invention can be combined with other embodiments.

[0040] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "attachment" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0041] In this invention, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this invention, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0042] In this invention, "multiple" refers to two or more (including two), similarly, "multiple groups" refers to two or more (including two groups), and "multiple pieces" refers to two or more (including two pieces).

[0043] In power systems, traveling wave measurement points (nodes) and overhead lines (edges) together form a natural graph data structure. When a fault occurs in a transmission line, the fault traveling wave will travel along the overhead line, undergoing multiple reflections and ultimately reaching various measurement points where it will mix and superimpose. During this process, the traveling wave waveforms at each measurement point exhibit two characteristics: firstly, they are related to the topology of the power grid; secondly, they also contain complex nonlinear relationships with the fault point.

[0044] In particular, transmission systems with multi-T connections and ring network structures can effectively improve the robustness of fault location. In complex power grid topologies (such as multi-T connections), the system is susceptible to traveling wave reflection, signal attenuation, and noise interference, leading to larger location errors.

[0045] In view of this, embodiments of the present invention provide a method for locating faults in a power transmission network. This method establishes a wide-area traveling wave topology based on traveling wave measurement points and transmission lines, identifies faulty lines based on the wide-area traveling wave topology, utilizes graph recognition networks to mine complex relationships between nodes, accurately identifies faulty lines, and finally combines adaptive weights calculated by the graph recognition network to locate faults based on multi-link ranging, thereby improving ranging accuracy and accurately locating faults.

[0046] The following detailed description, in conjunction with the accompanying drawings, of the power transmission network fault location method provided by the present invention through specific embodiments and application scenarios, will be provided in detail.

[0047] The power transmission network fault location method provided in this invention can be applied to computer equipment, specifically executed by hardware or software within the computer equipment. The computer equipment includes, but is not limited to, one or more of various desktop computers, laptops, smartphones, tablets, vehicle terminals, IoT devices, or portable wearable devices. IoT devices can be one or more of smart speakers, smart TVs, smart air conditioners, or smart vehicle devices. Portable wearable devices can be one or more of smartwatches, smart bracelets, or head-mounted devices. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, and big data and artificial intelligence platforms.

[0048] The power transmission network fault location method provided in this embodiment of the invention can be executed by a computer device or a functional module or entity within a computer device capable of implementing the power transmission network fault location method. The following description uses a computer device as the executing entity to illustrate the power transmission network fault location method provided in this embodiment of the invention.

[0049] like Figure 1 As shown, the power transmission network fault location method includes steps 110 to 130.

[0050] Step 110: Determine the wide-area traveling wave topology of the transmission network. The wide-area traveling wave topology uses traveling wave measurement points as nodes and transmission lines as edges. Each traveling wave measurement point corresponds to a traveling wave feature, which is obtained by performing mode decomposition on the traveling wave signals collected by the traveling wave measurement points.

[0051] The computer device obtains the traveling wave signals collected by each traveling wave measuring point in the power transmission network, and establishes a wide-area traveling wave topology taking the traveling wave measuring points as nodes and the power transmission lines as edges. The traveling wave measuring points (also referred to as measuring points) are usually located at positions such as transformer lines and line ends, and can collect original traveling wave signals transmitted on the power transmission lines. The traveling wave signal refers to a high-frequency voltage or current component that propagates along the power transmission line at high speed when a fault or other mutation event (such as lightning strike or switch operation) occurs on the power transmission line. Unlike the traditional power frequency signal, the traveling wave signal mainly has the characteristics of high frequency, fast propagation, bidirectional propagation, and transient state. In a complex multi-T-connected power transmission line, due to the characteristics of the traveling wave signal and problems such as wave speed attenuation and signal interference, the fault location accuracy is not high.

[0052] The wide-area traveling wave topology is a traveling wave signal propagation model based on the topology structure of the power transmission network. In the wide-area traveling wave topology, the traveling wave measuring points are taken as nodes (Node), and the power transmission lines are taken as edges (Edge) to represent the propagation path of the traveling wave signal and reflect the physical connection relationship between the traveling wave measuring points. The node of each node, i.e., the traveling wave feature of the traveling wave measuring point, describes the traveling wave signal attribute of the traveling wave measuring point.

[0053] The traveling wave feature refers to the signal parameter obtained by processing the traveling wave signal collected by the traveling wave measuring point, including but not limited to at least one of the instantaneous energy, center frequency, propagation time delay, and wave head polarity, and is used to characterize the dynamic change of the traveling wave signal. The traveling wave feature is usually extracted by modal decomposition or other signal processing methods and is taken as the node feature of each node in the power transmission network topology.

[0054] The traditional topology graph usually only shows the connection relationship between devices in the power transmission network, such as shown in FIG. 1, which shows the connection relationship between devices such as solar panels, wind power new energy units, and power sources. Unlike the traditional topology graph, the present application, on this basis, combines the traveling wave feature of each traveling wave measuring point and refines it into a more representative wide-area traveling wave topology, i.e., as shown in FIG. 2, which not only clearly represents the transmission path formed by each power transmission line in the power transmission network, but also accurately reflects the transmission characteristics of the traveling wave signal in the power transmission network, and can make the fault location more accurate. Figure 2 Figure 3 As shown in FIG. 2, the wide-area traveling wave topology not only clearly represents the transmission path formed by each power transmission line in the power transmission network, but also accurately reflects the transmission characteristics of the traveling wave signal in the power transmission network, and can make the fault location more accurate.

[0055] In step 120, based on the wide-area traveling wave topology, the fault of each power transmission line is identified and the fault line is located through the graph recognition network.

[0056] The traveling wave features of each traveling wave measuring point constitute an initial traveling wave feature matrix corresponding to the wide-area traveling wave topology.

[0057] ​The computer device identifies the probability of each power transmission line being faulty by using the wide-area traveling wave topology and the traveling wave feature matrix composed of the traveling wave features of the respective traveling wave features, in combination with deep learning technology and graph structure technology, and judges the power transmission lines by using a probability threshold, so as to respectively judge whether each power transmission line is a faulty line. The faulty line can be one or more.

[0058] In this way, the complex multi-T connection power transmission line fault location problem can be converted into a classification problem, that is, the classification of each power transmission line as a faulty line or a normal line, so as to achieve efficient fault location.

[0059] In some embodiments, the computer device takes the topology structure of the wide-area traveling wave topology and the traveling wave feature matrix composed of the respective traveling wave features of the topology nodes (i.e., traveling wave measurement points) as input, learns through a graph recognition network, and outputs the fault probability of each power transmission line, and then identifies the faulty line. Exemplarily, the graph recognition network includes a graph convolutional neural network (GCN) and a graph attention network (GAT), in other words, the graph recognition network can be a fusion of the graph convolutional neural network and the graph attention network, which can be called a graph convolutional attention network. It can learn the traveling wave features of the traveling wave nodes in the graph topology, and can also adaptively mine the correlation between the traveling wave measurement points from the topology structure by means of the attention mechanism. It is suitable for learning the nonlinear correlation between the wide-area multi-source traveling wave signal and the fault location in a complex power grid environment, and constructing a more stable and robust power grid fault line determination model. The correlation degree output by the model can be used for fusion of multiple source measurement points to achieve accurate fault location after subsequent processing.

[0060] Step 130, taking the traveling wave measurement points of the two ends of the faulty line as the starting point, determining a plurality of distance measurement lines related to the faulty line, and locating the fault position in the plurality of distance measurement lines.

[0061] After obtaining one or more faulty lines, the computer device analyzes each faulty line in detail.

[0062] Specifically, for any faulty line, the computer device takes the traveling wave measurement points of the two ends of the faulty line as the starting point, selects the adjacent measurement points of the traveling wave measurement points of the two ends, further narrows the range in the entire wide-area traveling wave topology, forms a plurality of distance measurement lines related to the faulty line, and locates the fault position from these distance measurement lines. The distance measurement line refers to the distance measurement path used to calculate the fault position. The adjacent measurement points of the traveling wave measurement points of the two ends of the faulty line can be the first-order adjacent measurement points, second-order adjacent measurement points, etc.

[0063] The computer device positions the fault location in multiple distance measurement lines, including: using single-end method, double-end method and the like to measure the distance between the fault location and the reference measurement point, so as to position the fault location. The reference measurement point can be a traveling wave measurement point at both ends of the fault line, or a preset traveling wave measurement point. Since there can be multiple distance measurement lines, the computer device can obtain multiple distance results, and then the computer device can fuse the multiple distance results, so as to obtain the final distance result. The fusion processing includes at least one of the following: mean value, median value, weighted sum processing and the like.

[0064] The power grid fault positioning method provided by the embodiment of the application constructs a wide-area traveling wave graph data structure, takes the traveling wave measurement point as a node and the overhead line as an edge, uses the graph data structure to represent the topology relationship of the power grid, provides basic data support for fault positioning, takes the wide-area traveling wave topology as input, uses the proposed learnable graph recognition network to mine the wide-area traveling wave features and learn the nonlinear relationship between the multiple measurement points and the fault line, can make the fault line identification more accurate, especially in the multi-T connection structure of the power grid, greatly improves the robustness of fault line identification, and finally forms multiple distance measurement lines based on the double-end measurement points of the fault line, further narrows the topology range of positioning faults, uses a multi-link distance measurement fusion mechanism to position faults, avoids the error of a single distance measurement path, and improves the accuracy and stability of distance measurement.

[0065] In power grid fault diagnosis, the traveling wave signal is widely used due to its fast propagation speed and sensitive fault response. However, the original traveling wave signal usually contains noise and multi-frequency aliasing components, and the traveling wave signal at a single moment cannot completely reflect the fault characteristics, resulting in low accuracy of directly using the original traveling wave signal for fault identification. Therefore, a signal processing method needs to be used to extract features from the traveling wave signal to enhance the robustness and stability of fault diagnosis.

[0066] In some embodiments, the method further includes: obtaining the original traveling wave signals collected by each of the traveling wave measurement points, respectively; for any traveling wave measurement point, performing modal decomposition on the original traveling wave signal collected by the traveling wave measurement point to obtain multi-dimensional modal components corresponding to the traveling wave measurement point; determining the energy values of the multi-dimensional modal components in multiple windows by setting window parameters, respectively; and performing dimension reduction processing on the modal components based on the energy values of the multi-dimensional modal components in the windows to obtain traveling wave features of a preset dimension.

[0067] The computer device obtains the traveling wave signals collected by each of the traveling wave measurement points in the power grid, respectively, to U ( t ) represents, t ∈[0,T ]. wherein, T is the sampling duration, usually several milliseconds to tens of milliseconds, and the sampling rate is usually between 1-10MHz. Thus, through high-frequency sampling of the traveling wave measuring point, the transient information of the traveling wave signal can be completely obtained, and the sensitivity of fault detection is improved.

[0068] For any traveling wave measuring point, the computer device decomposes the original traveling wave signal collected by it into k modal components by a modal decomposition method, and the k modal components constitute the k dimensional modal components of the traveling wave measuring point. k The number of modal components can be a preset value, or an adaptive modal selection algorithm (such as deep learning prediction k of the value) driven by data can be used to optimize the modal decomposition effect. The decomposition of the original traveling wave signal into k modal components can be expressed as: U ( t )→{ u 1( t ), u 2( t ),…, u k ( t )}.

[0069] For any original traveling wave signal, the computer device obtains preset window parameters, including time window size and window number, and divides the traveling wave signal into multiple windows, i.e. forms multiple energy blocks, and calculates the energy value of the multi-dimensional modal components in each window. Exemplarily, the computer device calculating the energy value of the multi-dimensional modal components in each window can be expressed as:

[0070] (1)

[0071] wherein, denotes the signal value of the i th sampling point, M is the number of sampling points in each window.

[0072] Exemplarily, assuming that there are 50 windows and the total number of sampling points is 500 to 1000, then each window contains about 10 to 20 sampling points.

[0073] Further, the computer device realizes dimension reduction processing of the modal components by normalization processing based on the energy values of the multi-dimensional modal components in each window, and obtains the preset dimension traveling wave features. The preset dimension is related to the window number, and usually the preset dimension is equal to the window number.

[0074] Exemplarily, the traveling wave feature of a preset dimension obtained by the computer device through the normalization processing can be expressed as:

[0075] (2)

[0076] wherein, max E is the maximum energy value of all modal components of the original traveling wave signal, and min E is the minimum energy value of all modal components of the original traveling wave signal.

[0077] In the above embodiment, the modal energy of each window is calculated through window division, so that the feature when the fault occurs is more prominent, the robustness is improved, and the accuracy of subsequent fault identification is improved. Through dimension reduction processing, the computational complexity is reduced, and the model calculation efficiency is improved.

[0078] Variational Mode Decomposition (VMD) is an adaptive signal decomposition method, which can decompose a complex signal into a plurality of intrinsic mode functions (IMFs) with limited bandwidth, and can adaptively adjust the center frequency of each mode to maximize the preservation of key features. In some embodiments, the computer device can perform variational mode decomposition on the traveling wave signals collected by the traveling wave measuring points, decompose the traveling wave signals into k intrinsic mode components, so that each modal component has a limited bandwidth, thereby removing noise and highlighting key frequency components. The frequency distribution of the final traveling wave feature is more clear, which can improve the stability of the traveling wave signal and help improve the accuracy of fault line identification.

[0079] To this end, for any traveling wave measuring point, the original traveling wave signal collected by the traveling wave measuring point is subjected to modal decomposition to obtain a multi-dimensional modal component corresponding to the traveling wave measuring point, including: for any traveling wave measuring point, the original traveling wave signal collected by the traveling wave measuring point is subjected to multi-layer decomposition to obtain a plurality of initial modal components; determining the initial center frequency of each initial modal component; under the constraint condition that the sum of the plurality of initial modal components is equal to the original traveling wave signal, the plurality of initial modal components and the corresponding initial center frequencies are iteratively updated a plurality of times until the iteration stopping condition is met, to obtain the multi-dimensional modal component corresponding to the traveling wave measuring point.

[0080] For any traveling wave measuring point, the computer device collects its original traveling wave signal, and sets the decomposition layer number k , and performs k layer decomposition by using the variational mode decomposition method, to obtain kan initial modal component. The initial modal component provides the basic frequency spectrum information of the traveling wave signal, providing input for subsequent optimization. Moreover, based on the decomposed eigenmodal components, the fault traveling wave is effectively characterized, and the complex signal is decomposed into simple parts each containing specific frequency information, so that the complexity of the original signal is greatly reduced, and the local mutation is more clear and the position information is more clear.

[0081] The computer device calculates the center frequency of each initial modal component to represent the frequency spectrum characteristics of the signal. Further, under the constraint that the sum of the plurality of initial modal components is equal to the original traveling wave signal, the computer device updates through multiple iterations to continuously optimize the modal components and the center frequencies until convergence, i.e., the iteration stopping condition is met. The iteration stopping condition includes but is not limited to that each initial modal component is aligned around the respective center frequency, and the sum of the bandwidths of each initial modal component is minimized.

[0082] Since the frequency components of the signal are not fixed, if the center frequency is not accurate, the decomposed modal components may be aliased or frequency shifted, and by recalculating each iteration, each modal component can be gathered around its correct center frequency.

[0083] Exemplarily, assume that a complex signal U ( t ) is decomposed into k components, the optimization goal is to decompose the sequence into modal components with limited bandwidths with center frequencies, while the sum of the estimated bandwidths of each modal is minimized, and the constraint condition is that the sum of all modes is equal to the original complex signal. The constrained variational model can be represented as:

[0084] (3)

[0085] where ω k is the center frequency of the k -dimensional eigenmodal component u k (t), u k is the k th eigenmodal component after decomposition, k is the number of decomposition layers, i.e., the number of eigenmodal components. is the Dirac function, * is the convolution operator, U is the line mode or zero mode component of the fault voltage traveling wave.

[0086] As shown in Figure 4 , the computer device initializes the modal component u k , the center frequency ω k and the Lagrange multiplier λ , from the iteration number n=0, sequentially for each modal component u k and the center frequency ω k is updated, and the Lagrange multiplier is updated at the same time λ Through repeated iterations n times, when the iteration stopping condition is met, the k modal components, i.e., the multi-dimensional modal components of the traveling wave measuring point, are obtained.

[0087] Further, by taking advantage of the quadratic method and the Lagrange multiplier method, an augmented Lagrange function is introduced to equivalently convert the above equation constraint optimization problem into an unconstrained optimization problem. Then, the computer device updates each modal component and the center frequency of each modal component, and solves by the alternating direction multiplier method, i.e., fixing the other two variables, updating one variable, thereby realizing iterative updating. The alternating optimization ensures that the decomposition of each modal has good bandwidth convergence and can adaptively adjust the frequency, and has strong robustness. Under the condition of meeting the iteration stopping condition, the iteration updating is stopped, and the final k dimensional eigenmodal component is obtained.

[0088] In the above embodiment, the traveling wave signal is decomposed into multiple band-limited signals by multi-layer decomposition, avoiding high-frequency noise interference and improving the separability of the traveling wave feature. By iteratively optimizing the modal component and the center frequency, and taking the sum of all modal components equal to the original signal as the constraint condition, not only the accuracy of signal decomposition is improved, but also the modal center frequency is more stable, which is helpful for subsequent feature extraction and fault diagnosis.

[0089] In the power transmission network, accurate identification of faults is crucial for safe operation of the power grid. Due to the complex topology of the transmission line, the propagation of the traveling wave signal on the line is affected by factors such as power grid topology, branch nodes, impedance characteristics, etc., resulting in complex spatiotemporal relationships of the traveling wave signal at different measuring points. Therefore, how to effectively utilize wide-area traveling wave data combined with topology information to accurately identify the fault line has become a problem to be solved. Traditional methods often analyze the traveling wave signal at each measuring point separately, ignoring the dynamic correlation between different measuring points, resulting in a decrease in fault identification accuracy in multi-T connection and ring network structures.

[0090] Based on this, in some embodiments, based on wide-area traveling wave topology, fault identification is performed on each transmission line using a graph recognition network to locate the faulty line. This includes: determining the adjacent measurement points of each traveling wave measurement point based on the correlation between each traveling wave measurement point in the wide-area traveling wave topology; generating global features based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points; determining the importance of each traveling wave measurement point and its adjacent measurement points, and generating multiple location features based on the importance; fusing the global features and location features to generate target features corresponding to each transmission line; performing fault identification on each transmission line based on the target features to obtain the judgment result for each transmission line, and locating the faulty line based on the judgment result for each transmission line.

[0091] Specifically, the computer device constructs the adjacency relationships between measurement points based on the wide-area traveling wave topology and the connection relationships of transmission lines. In some embodiments, the computer device can determine the direct adjacent measurement points of each traveling wave measurement point by defining an adjacency matrix.

[0092] (4)

[0093] Based on the correlation between various traveling wave measurement points, computer equipment can use graph convolutional neural networks to extract global features from the wide-area traveling wave topology based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points, thereby generating global features.

[0094] For example, the generation of global features by a computer device through a graph convolutional neural network can be represented as:

[0095] (5)

[0096] (6)

[0097] in, A It is an adjacency matrix. It is the first l The node feature matrix of the layer It is the first l The learnable weight matrix of the layer. This is the activation function.

[0098] Furthermore, based on the correlation between each traveling wave measurement point, the computer device can combine an attention mechanism to measure the importance of each traveling wave measurement point to its neighboring measurement points, and generate multiple localization features based on the importance.

[0099] Therefore, the computer equipment integrates global features and location features, and combines global information with local fault propagation characteristics to generate target features corresponding to each transmission line.

[0100] Finally, the computer equipment identifies faults in each transmission line based on target features, obtains a judgment result for each transmission line, and locates the faulty line based on the judgment result. In some embodiments, the computer equipment can identify faults based on target features using a binary classification method and output the fault probability of each transmission line. For any transmission line, if its fault probability exceeds a preset threshold, it is determined to be a faulty line; otherwise, if its fault probability does not exceed the preset threshold, it is determined to be a normal line, thereby improving the automation level of fault identification and achieving accurate fault location.

[0101] In the above embodiments, by combining the transmission network topology, global fault propagation feature extraction is achieved, which improves the robustness of fault detection. Furthermore, by determining the importance of key measurement points, the accuracy of fault identification can be improved.

[0102] In power transmission network fault detection, the propagation of traveling wave signals is affected by factors such as the transmission line topology, impedance characteristics, and branch connections. Analyzing the traveling wave characteristics at a single measurement point is insufficient to comprehensively reflect the global characteristics of fault propagation. Especially in power transmission networks with multiple T-junctions or ring network structures, traveling wave signals undergo multiple reflections and attenuation, making it difficult for information from a single measurement point to accurately characterize fault features. Using traveling wave data from only one measurement point for fault diagnosis may lead to misdiagnosis or missed diagnosis.

[0103] Therefore, in some embodiments, global features are generated based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points. This includes: for any traveling wave measurement point, the traveling wave characteristics of the target traveling wave measurement point are aggregated multiple times with the traveling wave characteristics of all its adjacent measurement points to obtain multiple aggregated features; wherein the traveling wave characteristics of the traveling wave measurement point in each aggregation process are based on the aggregated features obtained in the previous aggregation process; and the aggregated features obtained in the last aggregation are used as global features. Through multiple feature aggregations, traveling wave information is diffused and fused throughout the entire power grid topology, ultimately obtaining a global feature vector that can characterize global traveling wave characteristics.

[0104] Specifically, in each feature aggregation step, the computer device aggregates the initial traveling wave features of each traveling wave measurement point with the initial traveling wave features of its adjacent measurement points to obtain the traveling wave features output for that step. For example, suppose the initial traveling wave feature of traveling wave measurement point A is H. (0) The adjacent measurement points of traveling wave measurement point A include traveling wave measurement points B and D, and a total of [number] measurement points are measured. l Secondary feature aggregation yields H after the first feature aggregation. (1) And based on H (1) Perform a second feature aggregation to obtain H (2) ...until the l Sub-feature aggregation yields H (l)The feature aggregation manner includes, but is not limited to, weighted summation, etc. Exemplarily, each time of feature aggregation can be represented as formula (5) as described above.

[0105] Finally, after the secondary feature aggregation, the computer device can obtain the global feature of each traveling wave measuring point, which not only contains the information of the measuring point itself, but also fuses the global information in the topology structure. l

[0106] In the above embodiments, by performing feature aggregation on the traveling wave features of the traveling wave measuring point and its adjacent measuring points, the final feature of the measuring point can fully express the global topology information, thereby improving the accuracy and robustness of fault location.

[0107] In the fault detection of the power transmission network, the traveling wave signals propagate among multiple measuring points, and the traveling wave signals received by different measuring points may differ in amplitude, phase, arrival time, etc. These differences are influenced by factors such as the topology of the power transmission network, line impedance, traveling wave attenuation, and measuring point distance, resulting in different importance of each measuring point for fault identification. In fault identification, if the same weight is used for the measuring points with large contribution to fault identification and the measuring points with small contribution, the fault feature may be diluted. Especially in complex topologies (such as multi-T connection structure), the feature of a single measuring point is difficult to accurately describe the entire fault mode, which is prone to misjudgment.

[0108] Therefore, in some embodiments, the importance of each traveling wave measuring point and its adjacent measuring points is determined respectively, and multiple positioning features are generated based on the importance, including: for any traveling wave measuring point, determining the correlation between the traveling wave measuring point and any adjacent measuring point of the traveling wave measuring point; based on the correlation between the traveling wave measuring point and any adjacent measuring point of the traveling wave measuring point, determining the attention weight between the traveling wave measuring point and any adjacent measuring point of the traveling wave measuring point; based on the attention weight between the traveling wave measuring point and any adjacent measuring point of the traveling wave measuring point, performing aggregation processing on the traveling wave features of all adjacent measuring points to obtain the positioning feature corresponding to the traveling wave measuring point.

[0109] Specifically, for each traveling wave measuring point, the computer device calculates the correlation between the traveling wave measuring point and each adjacent measuring point based on one or more adjacent measuring points of the traveling wave measuring point. The correlation reflects the degree of association between the traveling wave measuring point and its adjacent measuring points.

[0110] ​The calculation manner of the correlation degree can be based on feature similarity, signal propagation characteristics, or learning weights of a deep learning model to ensure that the correlation degree can effectively reflect the fault correlation between the measuring points. For example, the computer device can calculate a cosine similarity as the correlation degree to measure the similarity in direction of the traveling wave feature vectors of two measuring points. For another example, the computer device can calculate a Pearson correlation coefficient as the correlation degree to measure the linear correlation between the traveling wave signals of two measuring points. For another example, the computer device can calculate the Euclidean distance between the traveling wave feature vectors of two measuring points as the correlation degree to measure the difference between the signals of the measuring points.

[0111] In some embodiments, for any traveling wave measuring point, the computer device automatically learns the correlation between the traveling wave measuring point and any of its adjacent measuring points through a graph convolution network to obtain the correlation degree of the two. Illustratively, the computer device outputs the correlation degree between the traveling wave measuring point and its adjacent measuring points through the graph convolution network, which can be represented as: i j

[0112] (7)

[0113] wherein, represents the correlation degree between the traveling wave measuring point and the traveling wave measuring point, i j is an attention parameter of the neural network, is a learnable parameter matrix, W represents a concatenation operation of vectors, LeakyReLU is a rectified linear unit, and are the respective traveling wave features of the traveling wave measuring point and the traveling wave measuring point. i j

[0114] Thus, for any traveling wave measuring point, the computer device can calculate the correlation degree between it and all adjacent measuring points. By calculating the correlation degree, the adjacent measuring points that have a greater impact on the current traveling wave measuring point can be identified, which provides a basis for subsequent attention weight calculation.

[0115] After calculating the correlation degree, the correlation degrees need to be converted into weight coefficients for subsequent feature aggregation. In some embodiments, for any traveling wave measuring point, the computer device can convert the correlation degree between the traveling wave measuring point and any adjacent measuring point into the attention weight between the traveling wave measuring point and the adjacent measuring point by means of softmax normalization, normalized weighting, or direct weighting.

[0116] ​​​​​​After obtaining the attention weight of each adjacent measurement point, the computer device aggregates the traveling wave feature of the measurement point with the traveling wave features of all adjacent measurement points of the measurement point to obtain a final positioning feature. The aggregation manner includes but is not limited to weighted summation, weighted average, linear transformation, or recursive updating through a multi-layer attention mechanism, etc. Since the attention weights of different measurement points are different, the final positioning feature can more accurately reflect the fault correlation between the measurement points.

[0117] In some embodiments, the computer device aggregates the traveling wave features of all adjacent measurement points by taking the attention weight of each adjacent measurement point as a weighting coefficient, so as to reflect the propagation of multi-source traveling wave positioning information in the local area to the center node, and obtain the positioning feature. For example, the computer device can calculate the positioning feature by the following formula:

[0118] (8)

[0119] wherein, is the corresponding positioning feature of the traveling wave feature i .

[0120] In the above embodiments, by calculating the correlation, the adjacent measurement points that have greater influence on the current measurement point can be identified, thereby providing a basis for subsequent attention weight calculation; and by calculating the attention weight, the measurement points that contribute more can play a more important role in the fault identification process, and the final positioning feature can more accurately represent the fault influence range, thereby reducing the interference of irrelevant measurement points.

[0121] Therefore, in some embodiments, based on the correlation between the targeted traveling wave measurement point and any adjacent measurement point of the targeted traveling wave measurement point, the attention weight between the targeted traveling wave measurement point and any adjacent measurement point of the targeted traveling wave measurement point is determined, including: based on the sum of the correlations between the targeted traveling wave measurement point and all adjacent measurement points of the targeted traveling wave measurement point, a total correlation is obtained; for any adjacent measurement point of the targeted traveling wave measurement point, the proportion of the correlation between the targeted traveling wave measurement point and the targeted adjacent measurement point in the total correlation is determined; and based on the proportion, the attention weight between the targeted traveling wave measurement point and the targeted adjacent measurement point is obtained.

[0122] Specifically, for any traveling wave measurement point, the computer device calculates the correlation between the traveling wave measurement point and each adjacent measurement point, respectively, and obtains a total correlation corresponding to the traveling wave measurement point based on the sum of the correlations of all adjacent measurement points. The total correlation can reflect the fault propagation influence range of the traveling wave measurement point in the entire set of adjacent measurement points.

[0123] Then, the computer device calculates the correlation ratio between the traveling wave measurement point and any one adjacent measurement point, that is, the proportion of the correlation in the total correlation. The proportion represents the contribution degree of the adjacent measurement point in the final feature calculation of the traveling wave measurement point. Therefore, the correlation ratio is used as the final attention weight to weight the traveling wave features of the adjacent measurement points, and the traveling wave features of the traveling wave measurement point are calculated by weighting the traveling wave features of the adjacent measurement points.

[0124] For example, the computer device can calculate the attention weight between the traveling wave measurement point i and its adjacent measurement points j by the following formula:

[0125] (9)

[0126] wherein, represents the adjacent measurement point of the traveling wave measurement point k , represents the number of adjacent measurement points of the traveling wave measurement point , and the traveling wave feature of the adjacent measurement point k .

[0127] In the above embodiment, through the attention weight calculation, in the fault identification process, the measurement points with greater contribution can play a more important role, the measurement point features with higher fault contribution degree are highlighted, so that the final features are more in line with the fault propagation law, and the fault identification accuracy is improved.

[0128] In the present application, the global feature can be extracted by a graph convolutional neural network, which can reflect the fault propagation mode of the entire power grid, but may lack accurate description of local key measurement points. The positioning feature can be calculated by a graph attention network, which emphasizes the correlation strength between measurement points and can accurately locate the fault influence range, but may ignore the global background information. Using the global feature alone may lose local details, resulting in large ranging error.

[0129] Therefore, in some embodiments, the global feature and the positioning feature are fused to generate target features corresponding to each power transmission line, including: determining a first fusion coefficient and a second fusion coefficient; performing feature fusion on the global features and the positioning features corresponding to each traveling wave measurement point based on the first fusion coefficient to obtain initial target features corresponding to each traveling wave measurement point; adjusting the initial target features corresponding to each traveling wave measurement point based on the second fusion coefficient to obtain target features between each traveling wave measurement point and the adjacent measurement points of each traveling wave measurement point; wherein any traveling wave measurement point and the adjacent measurement points of any traveling wave measurement point correspond to a power transmission line.

[0130] Specifically, the computer device obtains a first fusion coefficient and a second fusion coefficient. The fusion coefficients are used to balance the contribution degree of global features and positioning features to fault line identification. The fusion coefficients can be preset values or adaptive learning parameters obtained through training optimization. The first fusion coefficient is used to control the weight of global features and positioning features in initial fusion. The second fusion coefficient is used to further optimize the target features at the transmission line level, so that they are more in line with the fault propagation law.

[0131] The computer device first fuses the global features and the positioning features in a weighted summation manner to generate an initial target feature. Through preliminary fusion, the organic combination of global information and local information is ensured, and the fault identification capability at the measurement point level is improved. Exemplarily, based on the first fusion coefficient , the computer device fuses the global features and the positioning features to generate an initial target feature, which can be represented as follows:

[0132] (10)

[0133] wherein, is the positioning feature, is the global feature, is the initial target feature, represents the i th traveling wave measurement point.

[0134] The initial target feature is applicable to a single measurement point, and needs to be further extended to the transmission line level. Therefore, the computer device adjusts the initial target feature by using the second fusion coefficient, so that the target feature not only contains single measurement point information, but also fuses the information of adjacent measurement points, so as to be more in line with the fault propagation mode on the transmission line. Exemplarily, based on the second fusion coefficient , the computer device can calculate the target feature by the following formula:

[0135] (11)

[0136] wherein, the target feature is the fault feature representation of the transmission line connected with the traveling wave measurement point i and the traveling wave measurement point j . When the first fusion coefficient is 0 and the second fusion coefficient is 1, the model reflects the attention based on the original features. When the second fusion coefficient is 0, it reflects the characteristics of the model uniformly aggregating the neighbor node information.

[0137] Exemplarily, as Figure 5As shown, the graph recognition network in the embodiment of the present application includes a graph convolution network and a graph attention network, the graph convolution network is used to output global features, and the graph attention network is used to output positioning features, finally, the two are combined, the global features and the positioning features are jointly optimized through the feature fusion mode, so as to obtain more accurate and stable target features of the power transmission line level.

[0138] After the fusion of the global features and the positioning features is completed, the computer device obtains the target features of each measuring point. In order to determine which power transmission line has a fault, it is necessary to analyze and classify based on the target features, and finally output the recognition result of the fault line. In some embodiments, the computer device can use machine learning or deep learning methods for classification, such as threshold method, traditional machine learning methods (such as support vector machine, random forest, XGBoost, etc.), deep learning methods (such as graph neural network, etc.) or Bayesian inference, etc. The target features are analyzed and classified based on the target features, and finally the recognition result of the fault line is output. For example, the computer device can perform linear transformation on the target features through the graph neural network to output the probability that the corresponding power transmission line is the fault line, and then determine whether the power transmission line is the fault line. Thus, the fault line can be automatically identified in power grids of different complexities, reducing manual misjudgment and improving system efficiency.

[0139] In the above embodiment, the proportion of the global features and the positioning features is adjusted through the fusion coefficient, ensuring the adaptability in different scenarios, improving the flexibility of fault identification, and improving the fault identification capability of the single measuring point level; through the feature adjustment of the power transmission line level, the information of the fault propagation path is fully modeled, so that the target features are more consistent with the actual fault propagation mode.

[0140] In some embodiments, starting from the traveling wave measuring point at both ends of the fault line, a plurality of fault locating lines associated with the fault line are determined, including: taking the traveling wave measuring points at both ends of the fault line as candidate traveling wave measuring points, and determining the adjacent measuring points of the candidate traveling wave measuring points; based on the candidate traveling wave measuring points and their adjacent measuring points, a plurality of fault locating lines associated with the fault line are formed.

[0141] After one or more fault lines are determined, for each fault line, the computer device finds the traveling wave measuring points at both ends of the line. The traveling wave measuring points at both ends of the fault line are key nodes of the traveling wave signal propagation, which can be used to measure the arrival time of the traveling wave and perform fault location. Since the traveling wave signal propagates along multiple paths in the power transmission line, using only the traveling wave measuring points at both ends for fault location may have errors, so it is necessary to expand to more measuring points. To this end, the computer device finds the adjacent traveling wave measuring points based on the traveling wave measuring points at both ends of the fault line, forming a wider fault locating path, thereby expanding the range of the fault locating path by increasing the adjacent measuring points and improving the reliability of the fault locating data.

[0142] Further, the computer device takes the traveling wave measurement points at both ends of the fault line and the adjacent measurement points of each traveling wave measurement point as the core, traverses all feasible traveling wave transmission routes, forms multiple ranging paths, so as to calculate the fault distance on multiple paths and perform fusion optimization.

[0143] Exemplarily, as shown in Figure 6 The computer device determines multiple ranging routes based on the traveling wave measurement point b and the traveling wave measurement point c at both ends of the fault line: W ab , W bc , W cd , W cf , and W eb , wherein the traveling wave measurement point a and the traveling wave measurement point e are the first-order adjacent measurement points of the traveling wave measurement point b, and the traveling wave measurement point d and the traveling wave measurement point f are the first-order adjacent measurement points of the traveling wave measurement point c.

[0144] In the above embodiment, the ranging path range is expanded, the redundancy of the system is enhanced, and the ranging stability is improved; the multi-link fusion ranging can reduce the ranging error and improve the final ranging accuracy.

[0145] If the signal of a ranging route is affected by interference or noise, it may cause the ranging result to deviate. The fault position may be located at the intersection point of multiple ranging paths, and how to comprehensively consider multiple ranging results is a key problem. Therefore, in some embodiments, the fault position is located in the multiple ranging routes, including: respectively determining the initial distance between the fault position and the traveling wave measurement points at both ends of each ranging route; determining the path weight between the traveling wave measurement points at both ends of the fault line; based on the path weight, weighting and fusing the distances between the fault position and the traveling wave measurement points at both ends of all ranging routes to obtain the target distance between the fault position and the reference measurement point; the reference measurement point includes the traveling wave measurement points at both ends of the fault line.

[0146] Specifically, on each ranging route, the computer device calculates the initial distance from the fault point to the measurement points at both ends of the route by using the traveling wave time difference ranging method. For example, the computer device can calculate the initial distance on the ranging route of the traveling wave measurement point a and the traveling wave measurement point d by the following formula:

[0147] (12)

[0148] Wherein, c is the wave mode wave speed of the traveling wave, is the electrical distance of the a point and the d point.

[0149] For each ranging route, the calculation can be performed in the same way, for example, for multiple ranging routes: W ab , W bc , W cd , W cf , and W ebThe initial distances of the respective distances are D ab , D bc , D cd , D cf , and D eb . Thus, the initial distances from the fault point to different measuring points are calculated through multiple ranging lines, multiple independent ranging results are obtained, and the ranging robustness is improved.

[0150] However, the ranging results of different ranging paths may have errors, and therefore the path weight needs to be calculated to measure the importance of different paths. The path weight can be calculated according to the ranging signal quality, the ranging path length, and the historical data reliability (for example, if the error of a certain path is small in historical ranging, a higher weight is given).

[0151] In some embodiments, the path weight between the traveling wave measuring points is the attention weight between the traveling wave measuring points. In this way, the path weight has the ability of adaptive adjustment, which can significantly improve the ranging accuracy and reliability. Especially in multi-T connected power grids and ring network structures, the traveling wave signal may propagate along multiple paths, resulting in non-unique ranging results. The attention weight calculated in the above steps can automatically select the optimal path and reduce the ranging error.

[0152] For example, the computer device can fuse the initial distances based on the attention weight to obtain a target distance, which can be represented as:

[0153] (13)

[0154] Finally, the computer device can fuse all the ranging results by the weighted average method to calculate the target distance from the fault point to a reference measuring point. The reference measuring point can be a traveling wave measuring point at both ends of the fault line, or other traveling wave measuring points. Therefore, the target distance usually needs to be converted into a physical distance by combining the physical parameters and geographical information of the power grid, so as to locate the fault position in the actual power grid and perform subsequent operations such as maintenance.

[0155] In the above embodiments, the path weight can improve the contribution of high-quality ranging paths, reduce the influence of low-quality ranging paths on the final result, and fuse the results of different ranging lines by weighted fusion, reduce single path error, and improve the stability and accuracy of the final ranging.

[0156] In order to more accurately model the law of wave propagation, on the basis of any one of the above embodiments or a combination thereof, a wide-area traveling wave topology of the power transmission network is determined, including: based on the structural information in the power transmission network, a graph data model is constructed with traveling wave measuring points as nodes and power transmission lines as edges; based on at least one of the electrical distance, impedance characteristics or historical traveling wave propagation data of the power transmission line, the topological connection relationship in the graph data model is weighted processed to obtain the wide-area traveling wave topology of the power transmission network.

[0157] Specifically, the computer device constructs a graph data model of the power transmission network with traveling wave measuring points as nodes of the graph and power transmission lines as edges of the graph, wherein each traveling wave measuring point collects a traveling wave signal, and an adjacency relationship is constructed according to the connection relationship of the power transmission lines, and the graph structure formed represents the topology of the power transmission network and can be used for fault propagation analysis.

[0158] In addition, the traveling wave signal propagation characteristics of the power transmission line are not only determined by the physical connection relationship, but also affected by the electrical characteristics and the historical traveling wave propagation data. The electrical characteristics include at least one of the electrical distance and the impedance characteristics.

[0159] The electrical distance refers to the equivalent physical length corresponding to the traveling wave signal propagation in the power transmission line, which is related to the length, material, structure and electromagnetic characteristics of the power transmission line. The shorter the electrical distance, the faster the traveling wave signal propagation, the smaller the attenuation, and the higher the signal correlation between measuring points. The longer the electrical distance, the longer the signal propagation time, which may cause multiple reflections and attenuation, resulting in a decrease in ranging accuracy.

[0160] The impedance characteristics refer to the degree of hindering of the traveling wave signal propagation of the power transmission line. The greater the impedance, the greater the traveling wave propagation loss, the more obvious the signal attenuation, and the more difficult the fault location. The smaller the impedance, the more stable the traveling wave signal propagation, and the higher the signal correlation between measuring points.

[0161] The historical traveling wave propagation data refer to the records of the traveling wave signal propagation characteristics between different measuring points in the long-term operation process of the power transmission network. For example, the computer device can calculate the correlation between measuring points by counting the past traveling wave signal arrival time, amplitude, frequency, attenuation, etc. If the traveling wave propagation patterns of two measuring points remain consistent over a long period of time, it indicates that they are closely related in the traveling wave propagation path and can be given a higher weight in the weighted topology. If the historical traveling wave propagation data of two measuring points have low similarity, it indicates that their traveling wave correlation is low, and the weight should be reduced.

[0162] Therefore, the computer device can respectively determine the respective weight coefficients of the electrical distance, the impedance characteristics or the historical traveling wave propagation data, and weight process the topological connection relationship in the graph data model to obtain the wide-area traveling wave topology of the power transmission network. The weight coefficient may be, for example, a preset value.

[0163] In the above embodiments, the connection relationship between the transmission lines is clearly defined through the graph data model, providing a data basis for the traveling wave propagation path analysis, especially suitable for complex power grid topologies such as multi-T connection, and can improve the adaptability of fault detection. Moreover, by weighting the topological connection relationship, the graph data model can more accurately reflect the traveling wave propagation characteristics, enhancing the physical reality of the topological model and improving the accuracy of fault detection.

[0164] The present application proposes a power grid fault location method based on a graph recognition network (i.e. a graph convolutional attention network, L-CAT), which utilizes wide-area traveling wave information and graph neural network technology to improve the precision and robustness of fault detection. The method includes core links such as traveling wave signal preprocessing, wide-area traveling wave graph data construction, fault line identification, multi-link distance fusion, and fault point positioning. In the traveling wave signal preprocessing stage, the traveling wave signal is decomposed using a variational mode decomposition algorithm to highlight the transient characteristics of the traveling wave front, while filtering out the baseband signal and noise interference, making the local mutation of the signal more clear and identifiable, and improving the efficiency and accuracy of subsequent processing. Compared with traditional signal processing methods, variational mode decomposition has higher adaptability and denoising ability, and is especially suitable for fault location of complex multi-T connected transmission lines. The extracted traveling wave features are normalized by "maximum-minimum value" to obtain standardized node features, improving the generalization ability and positioning accuracy of the model.

[0165] In terms of wide-area traveling wave graph data construction, the present application constructs a wide-area traveling wave graph data structure, taking traveling wave measuring points as nodes and transmission lines as edges, and weighting the topological relationship using electrical distance, impedance characteristics or historical traveling wave propagation data to make it more consistent with the traveling wave propagation law. The graph convolutional network is used to aggregate the features near the nodes to effectively model the topology structure of the power grid; the graph attention network adaptively calculates the adjacency weights between measuring points to highlight important traveling wave transmission paths and improve the accuracy of fault location.

[0166] In the fault line identification link, the graph recognition network combines the global feature extraction capability of the graph convolutional network and the local attention mechanism of the graph attention network, and through mining wide-area traveling wave features and learning the nonlinear relationship between multiple measuring points and fault lines, it realizes accurate identification of fault lines. The graph recognition network adopts an adaptive weight adjustment mechanism, which interpolates between the graph convolutional network and the graph attention network through two scalar parameters to generate a more robust model, better adapting to changes in traveling wave features at different fault locations, and improving the robustness and accuracy of fault location.

[0167] In terms of fault point ranging, a multi-link ranging fusion method is adopted, double-end measuring points of the fault line and first-order adjacent measuring points thereof are selected to form multiple ranging links, and the ranging results are weighted and averaged by using the weights output by the graph recognition network to improve the accuracy of fault ranging.

[0168] Compared with the traditional method, the innovation of the application lies in:

[0169] 1) The variational mode decomposition algorithm decomposes the traveling wave signal, reduces the signal complexity, and enhances the transient characteristics;

[0170] 2) The adaptive weight fusion of the graph recognition network model combines the information of multiple measuring points to realize accurate fault positioning;

[0171] 3) The combination of the graph convolution network and the graph attention network can aggregate node features and highlight key measuring points to improve fault recognition accuracy;

[0172] 4) The multi-link ranging fusion strategy improves the ranging stability and accuracy under complex topological structure.

[0173] Therefore, the fault positioning system of the graph recognition network model provided by the application covers traveling wave measuring point arrangement, graph data construction, model training optimization, fault line identification and fault point calculation, and provides a complete solution for fast and accurate positioning of power transmission line faults.

[0174] The power transmission network fault positioning method provided by the embodiment of the application can be executed by a power transmission network fault positioning device.

[0175] The embodiment of the application also provides a power transmission network fault positioning device applied to a computer equipment. Figure 7 As shown in the figure, the power transmission network fault positioning device comprises a topology module 701, an identification module 702 and a positioning module 703.

[0176] The topology module 701 is used to determine the wide-area traveling wave topology of the power transmission network, and the wide-area traveling wave topology takes the traveling wave measuring point as the node and the power transmission line as the edge.

[0177] The identification module 702 is used to identify the fault of each power transmission line through the graph recognition network based on the wide-area traveling wave topology, and locate the fault line.

[0178] The positioning module 703 is configured to determine a plurality of fault location lines related to the fault line from a traveling wave measurement point at a double end of the fault line, and locate the fault position in the plurality of fault location lines.

[0179] The power transmission network fault positioning device provided by the embodiment of the present application constructs a wide-area traveling wave graph data structure, takes a traveling wave measurement point as a node and an overhead line as an edge, uses the graph data structure to represent the topology relationship of the power transmission network, and provides basic data support for fault positioning. The wide-area traveling wave topology is taken as input, a learnable graph recognition network is used to mine wide-area traveling wave features and learn the nonlinear relationship between the multiple measurement points and the fault line, which can make the fault line identification more accurate, especially in the power grid with multi-T connection structure, greatly improving the robustness of fault line identification. Finally, a plurality of fault location lines are formed based on the double-end measurement points of the fault line, the topology range for locating the fault is further reduced, a multi-link fault location fusion mechanism is used for fault positioning, the error of a single fault location path is avoided, and the accuracy and stability of fault location are improved.

[0180] In some embodiments, the identification module is further configured to determine adjacent measurement points of each traveling wave measurement point based on the association relationship between the traveling wave measurement points in the wide-area traveling wave topology, generate a global feature based on the traveling wave features of each traveling wave measurement point and the adjacent measurement points of each traveling wave measurement point, determine the importance of each traveling wave measurement point and its adjacent measurement points, generate a plurality of positioning features based on the importance, generate a target feature corresponding to each power transmission line by fusing the global feature and the positioning features, perform fault identification on each power transmission line based on the target feature, obtain a determination result of each power transmission line, and locate the fault line based on the determination result of each power transmission line.

[0181] In some embodiments, the identification module is further configured to, for any traveling wave measurement point, perform multiple aggregation processing on the traveling wave feature of the traveling wave measurement point and the traveling wave features of all adjacent measurement points of the traveling wave measurement point, to obtain a plurality of aggregated features, wherein the traveling wave feature of the traveling wave measurement point in each aggregation processing is obtained based on the aggregated feature obtained in the last aggregation processing, and the aggregated feature obtained in the last aggregation processing is taken as the global feature.

[0182] In some embodiments, the identification module is further configured to, for any traveling wave measurement point, determine the correlation between the traveling wave measurement point and any adjacent measurement point of the traveling wave measurement point, determine the attention weight between the traveling wave measurement point and any adjacent measurement point of the traveling wave measurement point based on the correlation, and perform aggregation processing on the traveling wave features of all adjacent measurement points based on the attention weight between the traveling wave measurement point and any adjacent measurement point of the traveling wave measurement point, to obtain a positioning feature corresponding to the traveling wave measurement point.

[0183] In some embodiments, the identifying module is further configured to obtain a total correlation degree based on a sum of the correlation degrees between the targeted traveling wave measuring point and all adjacent measuring points of the targeted traveling wave measuring point; determine, for any adjacent measuring point of the targeted traveling wave measuring point, a proportion of the correlation degree between the targeted traveling wave measuring point and the targeted adjacent measuring point in the total correlation degree; and obtain an attention weight between the targeted traveling wave measuring point and the targeted adjacent measuring point based on the proportion.

[0184] In some embodiments, the identifying module is further configured to determine a first fusion coefficient and a second fusion coefficient; perform feature fusion on the global feature and the positioning feature corresponding to each traveling wave measuring point based on the first fusion coefficient to obtain an initial target feature corresponding to each traveling wave measuring point; and adjust the initial target feature corresponding to each traveling wave measuring point based on the second fusion coefficient to obtain a target feature between each traveling wave measuring point and an adjacent measuring point of each traveling wave measuring point; wherein any traveling wave measuring point and any adjacent measuring point of each traveling wave measuring point correspond to a power transmission line.

[0185] In some embodiments, the positioning module is further configured to take the traveling wave measuring points at both ends of the fault line as candidate traveling wave measuring points, and determine the adjacent measuring points of the candidate traveling wave measuring points; and form a plurality of distance measurement lines associated with the fault line based on the candidate traveling wave measuring points and the adjacent measuring points thereof.

[0186] In some embodiments, the positioning module is further configured to determine an initial distance between the fault location and each traveling wave measuring point at both ends of each distance measurement line, respectively; determine a path weight between the traveling wave measuring points at both ends of the fault line; and perform weighted fusion on the distances between the fault location and the traveling wave measuring points at both ends of all distance measurement lines based on the path weight to obtain a target distance between the fault location and a reference measuring point; wherein the reference measuring point includes the traveling wave measuring points at both ends of the fault line.

[0187] In some embodiments, the apparatus further includes a preprocessing module configured to obtain original traveling wave signals respectively collected by each traveling wave measuring point; perform modal decomposition on the original traveling wave signals collected by any traveling wave measuring point to obtain a plurality of multi-dimensional modal components corresponding to the targeted traveling wave measuring point; determine energy values of the multi-dimensional modal components in a plurality of windows by setting window parameters; and perform dimension reduction processing on the modal components based on the energy values of the multi-dimensional modal components in the windows to obtain traveling wave features of a preset dimension.

[0188] In some embodiments, the preprocessing module is further configured to, for any traveling wave measuring point, perform multi-layer decomposition on a raw traveling wave signal collected by the traveling wave measuring point to obtain a plurality of initial modal components; determine initial center frequencies of the initial modal components; and perform multiple iterations of updating on the initial modal components and the corresponding initial center frequencies under a constraint condition that a sum of the initial modal components is equal to the raw traveling wave signal, until an iteration stop condition is met, to obtain a multi-dimensional modal component corresponding to the traveling wave measuring point.

[0189] In some embodiments, the topology module is further configured to construct a graph data model with the traveling wave measuring points as nodes and the transmission lines as edges based on structural information in the power transmission network; and perform weighted processing on the topological connection relationship in the graph data model based on at least one of the electrical distance, the impedance characteristics, or the historical traveling wave propagation data of the transmission lines, to obtain a wide-area traveling wave topology of the power transmission network.

[0190] The power transmission network fault locating apparatus in the embodiments of the present application can be a computer device or a component in a computer device, such as an integrated circuit or a chip. The computer device can be a terminal device or a server. For example, the computer device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted computer device, a Mobile Internet Device (MID), an Augmented Reality (AR) / Virtual Reality (VR) device, a robot, a wearable device, an Ultra-mobile Personal Computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), and can also be a server, a Network Attached Storage (NAS), a Personal Computer (PC), a Television (TV), a teller machine, or a self-service machine, and the like, and the embodiments of the present application are not limited in this regard.

[0191] The power transmission network fault locating apparatus in the embodiments of the present application can be a device with an operating system. The operating system can be a Windows operating system, an Android operating system, an IOS operating system, or other possible operating systems, and the embodiments of the present application are not limited in this regard.

[0192] The power transmission network fault locating apparatus provided in the embodiments of the present application can implement the method embodiments Figure 1 The processes implemented by the method embodiments are not repeated here to avoid repetition.

[0193] In some embodiments, such as Figure 8 As shown, this embodiment of the invention also provides a computer device 800, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the program is executed by the processor 801, it implements the various processes of the above-described method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here.

[0194] It should be noted that the computer device in the embodiments of the present invention includes the mobile computer device and the non-mobile computer device described above.

[0195] This invention also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described power transmission network fault location method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0196] The processor is the processor in the computer device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0197] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described power transmission network fault location method.

[0198] The processor is the processor in the computer device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0199] This invention also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described power transmission network fault location method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0200] It should be understood that the chip mentioned in the embodiments of the present invention may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0201] It should be noted that, as used in this document, the terms "include," "includes," or "including" are intended to be non-limiting, such that recitation of the named item or items does not mean that any or all of the named item or items are required or contained. In addition, it should be noted that the scope of the methods and apparatus of the embodiments of the present application are not limited by the order of the steps or sequence for performing the steps, as some steps can occur in different orders and / or concurrently with other steps besides those depicted and described herein. Also, descriptions of features or aspects within each embodiment should typically be considered as available for combinations with other features or aspects in the remaining embodiments. Furthermore, the described features, concepts, aspects, and / or steps can be used in various combinations, and other features, concepts, aspects, and / or steps can be utilized or otherwise determined to use in accordance with the disclosed subject matter.

[0202] Those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a plurality of instructions for making a terminal (which can be a mobile phone, computer, server, or network device, etc.) execute the method described in each embodiment of the present application.

[0203] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-described specific embodiments, and the above-described specific embodiments are merely illustrative, not restrictive, and those skilled in the art can make many forms without departing from the purpose of the present application and the scope protected by the claims.

[0204] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the exemplary description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0205] All the embodiments and optional embodiments of the present application can be combined with each other to form new technical solutions, if not particularly stated.

[0206] All the technical features and optional technical features of the present application can be combined with each other to form new technical solutions, if not particularly stated.

[0207] All the steps of the present application can be performed in sequence or randomly, preferably in sequence, if not particularly stated. For example, the method comprises steps (a) and (b), which means that the method can comprise steps (a) and (b) performed in sequence, or steps (b) and (a) performed in sequence. For example, it is mentioned that the method can further comprise step (c), which means that step (c) can be added to the method in any sequence, for example, the method can comprise steps (a), (b) and (c), or steps (a), (c) and (b), or steps (c), (a) and (b), etc.

[0208] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of fault location in a power transmission network, characterized by, The power transmission network includes multiple transmission lines, and traveling wave measurement points are set at both ends of each transmission line; the method includes: A wide-area traveling wave topology of the power transmission network is determined. The wide-area traveling wave topology has traveling wave measurement points as nodes and the transmission lines as edges. Each traveling wave measurement point has a corresponding traveling wave feature, which is obtained by performing mode decomposition on the traveling wave signal collected by the traveling wave measurement point. Global features are generated based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points. Calculate the correlation between each traveling wave measurement point and its neighboring measurement points, and use the proportion of the correlation in the total correlation as the attention weight between each traveling wave measurement point and its neighboring measurement points; the correlation is used to reflect the fault correlation between measurement points; Based on the attention weight, the traveling wave features of each traveling wave measurement point and all its adjacent measurement points are aggregated to obtain the location features; the location features are used to characterize the fault impact range. Based on the first fusion coefficient, the global features and localization features corresponding to each traveling wave measurement point are fused to obtain the initial target features of each traveling wave measurement point. Based on the second fusion coefficient, the initial target features of each traveling wave measurement point and its adjacent measurement points are fused to obtain the target features corresponding to each transmission line. Based on the target characteristics, determine whether each transmission line is a faulty line; For each faulty line, multiple ranging lines are formed by selecting traveling wave measuring points at both ends and their adjacent measuring points. The distance between the fault location and the reference measuring point is measured in each ranging line to obtain multiple distance results. By utilizing the attention weights between the traveling wave measurement points at both ends and their adjacent measurement points, multiple distance results are weighted and fused to obtain the target distance between the fault location and the reference measurement point.

2. The method of claim 1, wherein, The method further includes: determining the adjacent measurement points of each traveling wave measurement point based on the correlation between each traveling wave measurement point in the wide-area traveling wave topology.

3. The method of claim 1, wherein, The generation of global features based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points includes: For any traveling wave measurement point, the traveling wave feature of the target traveling wave measurement point is aggregated multiple times with the traveling wave features of all adjacent measurement points of the target traveling wave measurement point to obtain multiple aggregated features; wherein, the traveling wave feature of the traveling wave measurement point in each aggregation process is based on the aggregated feature obtained in the previous aggregation process. The aggregated features obtained from the last aggregation are used as global features.

4. The method of claim 1, wherein, The method further includes: The total correlation is obtained by summing the correlations between the target traveling wave measurement point and all its adjacent measurement points. For any adjacent measurement point of the target traveling wave measurement point, determine the proportion of the correlation between the target traveling wave measurement point and the target adjacent measurement point to the total correlation.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: A first fusion coefficient and a second fusion coefficient are determined; wherein, the first fusion coefficient is used to control the weights of global features and localization features in the initial fusion; and the second fusion coefficient is used to make the target features conform to the fault propagation law at the transmission line level.

6. The method of claim 1, wherein, The reference measurement points include traveling wave measurement points at both ends of the faulty line.

7. The method of claim 1, wherein, The method further includes: The original traveling wave signals collected at each traveling wave measurement point were acquired separately; For any traveling wave measurement point, perform mode decomposition on the original traveling wave signal collected at the target traveling wave measurement point to obtain the multi-dimensional mode components corresponding to the target traveling wave measurement point. By setting window parameters, the energy values ​​of the multidimensional modal components in multiple windows are determined respectively; Based on the energy values ​​of the multidimensional modal components in each window, the modal components are subjected to dimensionality reduction processing to obtain traveling wave features of a preset dimension.

8. The method according to claim 7, characterized in that, For any traveling wave measurement point, modal decomposition is performed on the original traveling wave signal acquired at the target traveling wave measurement point to obtain the multidimensional modal components corresponding to the target traveling wave measurement point, including: For any traveling wave measurement point, the original traveling wave signal collected at the target traveling wave measurement point is decomposed into multiple layers to obtain multiple initial mode components. Determine the initial center frequency of each initial modal component; Using the constraint that the sum of the multiple initial modal components equals the original traveling wave signal, the initial modal components and their corresponding initial center frequencies are iteratively updated multiple times until the iteration stop condition is met, thereby obtaining the multidimensional modal components corresponding to the target traveling wave measurement point.

9. The method according to claim 1 or 7, characterized in that, The determination of the wide-area traveling wave topology of the transmission network includes: Based on the structural information in the power transmission network, a graph data model is constructed with traveling wave measurement points as nodes and transmission lines as edges; Based on at least one of the electrical distance, impedance characteristics, or historical traveling wave propagation data of the transmission line, the topological connection relationship in the graph data model is weighted to obtain the wide-area traveling wave topology of the transmission network.

10. A power transmission network fault location device, characterized in that, The power transmission network includes multiple transmission lines, and each transmission line has traveling wave measuring points at both ends; the device includes: The topology module is used to determine the wide-area traveling wave topology of the power transmission network. The wide-area traveling wave topology has the traveling wave measurement points as nodes and the transmission lines as edges. Each traveling wave measurement point has a corresponding traveling wave feature, which is obtained by performing mode decomposition on the traveling wave signals collected by the traveling wave measurement points. The identification module is used to generate global features based on the traveling wave characteristics of each traveling wave measurement point and its adjacent measurement points; calculate the correlation between each traveling wave measurement point and its adjacent measurement points, and use the proportion of the correlation in the total correlation as the attention weight between each traveling wave measurement point and its adjacent measurement points; the correlation is used to reflect the fault correlation between measurement points; aggregate the traveling wave characteristics of each traveling wave measurement point and all its adjacent measurement points based on the attention weight to obtain localization features; the localization features are used to characterize the fault impact range; perform feature fusion on the global features and localization features corresponding to each traveling wave measurement point based on a first fusion coefficient to obtain the initial target features of each traveling wave measurement point, and perform feature fusion on the initial target features of each traveling wave measurement point and its adjacent measurement points based on a second fusion coefficient to obtain the target features corresponding to each transmission line; and determine whether each transmission line is a faulty line based on the target features. The positioning module is used to select traveling wave measurement points at both ends and their adjacent measurement points to form multiple ranging lines for each faulty line, and measure the distance between the fault location and the reference measurement point in each ranging line to obtain multiple distance results; using the attention weight between the traveling wave measurement points at both ends and their adjacent measurement points, the multiple distance results are weighted and fused to obtain the target distance between the fault location and the reference measurement point.

11. A computer 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 power transmission network fault location method as described in any one of claims 1-9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the power transmission network fault location method as described in any one of claims 1-9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the power transmission network fault location method as described in any one of claims 1-9.

14. A chip, characterized in that, The chip includes a processor and a communication interface, the communication interface and the processor being coupled together, the processor being used to run programs or instructions to implement the power grid fault location method as described in any one of claims 1-9.

Images