Power distribution network single-end positioning method based on global traveling wave time-frequency diagram and related device

Abstract

The invention relates to the technical field of data processing, and discloses a power distribution network single-end positioning method based on a global traveling wave time-frequency diagram and a related device, and the method comprises the steps: analyzing the transmission characteristics of fault traveling waves under a single branch line based on the global traveling wave time-frequency diagram of a power distribution network fault traveling wave signal, and obtaining the transmission characteristics of the single branch line; based on the global traveling wave time-frequency diagram, analyzing the transmission characteristics of the fault traveling wave under the cross-branch line to obtain the transmission characteristics of the cross-branch line; constructing a power distribution network fault positioning model based on a residual network according to the single-branch line transmission characteristic and the cross-branch line transmission characteristic to obtain an initial power distribution network fault positioning model; based on a graph hybrid convolutional neural network, optimizing and improving the initial power distribution network fault positioning model according to the global traveling wave time-frequency graph to obtain a target power distribution network fault positioning model; according to the target power distribution network fault positioning model, single-end fault positioning is carried out on the multi-branch power distribution network, and a more accurate target fault positioning result can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and power protection, and in particular to a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram and a related device. Background Art

[0002] The distribution network operating environment is complex and changeable, with frequent adverse conditions and numerous branch nodes, which significantly increases the difficulty of fault location. If the fault point is not effectively found, it will not only directly lead to power outages for users, but may also cause major personal safety accidents, thus posing a serious threat to the stable operation of the power grid.

[0003] Among the electrical faults in the distribution network, the single-phase arc grounding fault is the most widespread and common. Therefore, the protection of this type of fault requires not only that the protection system act quickly and accurately to ensure the safe operation of the system, but also that the protection system can provide accurate fault location information for maintenance personnel to accurately troubleshoot the fault. The current traditional traveling wave positioning technology is often subject to many challenges in the application process, such as sampling frequency and synchronization deviation, difficulty in identifying the traveling wave head, environmental noise interference, complex network topology structure, and large transition resistance. Summary of the invention

[0004] The present invention provides a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram and a related device to solve the technical problem of how to improve the accuracy of fault location results.

[0005] In a first aspect, a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram is provided, comprising:

[0006] Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed, and the transmission characteristics of the single branch line are obtained;

[0007] Based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed, and the transmission characteristics of the cross-branch line are obtained;

[0008] According to the transmission characteristics of single branch lines and cross-branch lines, a distribution network fault location model is constructed based on the residual network to obtain an initial distribution network fault location model;

[0009] Based on graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency graph to obtain the target distribution network fault location model.

[0010] According to the target distribution network fault location model, single-end fault location is performed on the multi-branch distribution network to obtain the target fault location result.

[0011] In a second aspect, a single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram is provided, comprising:

[0012] The first processing module is used to analyze the transmission characteristics of the fault traveling wave under the single branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, and obtain the transmission characteristics of the single branch line;

[0013] The second processing module is used to analyze the transmission characteristics of the fault traveling wave under the cross-branch line based on the global traveling wave time-frequency diagram to obtain the transmission characteristics of the cross-branch line;

[0014] The third processing module is used to construct a distribution network fault location model based on a residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics, so as to obtain an initial distribution network fault location model;

[0015] The fourth processing module is used to optimize and improve the initial distribution network fault location model based on the graph hybrid convolutional neural network and the global traveling wave time-frequency graph to obtain the target distribution network fault location model;

[0016] The fifth processing module is used to perform single-end fault location on the multi-branch distribution network according to the target distribution network fault location model to obtain a target fault location result.

[0017] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned single-end positioning method for the distribution network based on the global traveling wave time-frequency diagram are implemented.

[0018] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned single-end positioning method of the distribution network based on the global traveling wave time-frequency diagram are implemented.

[0019] In the scheme implemented by the above-mentioned single-end positioning method of distribution network based on global traveling wave time-frequency diagram and related devices, based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the single branch line transmission characteristics, and based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics. According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, the distribution network fault location model can be further constructed based on the residual network to obtain the initial distribution network fault location model, so that the initial distribution network fault location model can be optimized and improved based on the graph hybrid convolutional neural network and the global traveling wave time-frequency diagram to obtain the target distribution network fault location model, and then the multi-branch distribution network can be single-ended The fault location is performed based on the target distribution network fault location model to obtain a more accurate target fault location result. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0021] Figure 1 It is a flow chart of a single-end positioning method of a distribution network based on a global traveling wave time-frequency diagram in one embodiment of the present invention;

[0022] Figure 2A is a schematic diagram of a multi-branch power grid simulation model in one embodiment of the present invention;

[0023] Figure 2B is a schematic diagram of transient fault waveforms of different fault branches in one embodiment of the present invention;

[0024] Figure 2C is a schematic diagram of the amplitude attenuation of the traveling wave transmission function in one embodiment of the present invention;

[0025] Figure 2D is a schematic diagram of a transient waveform when a same branch fails in one embodiment of the present invention;

[0026] Figure 2E is a schematic diagram of the relationship between the mode wave velocity and the frequency in one embodiment of the present invention;

[0027] Figure 2F It is a schematic diagram of the time-frequency of wavelet traveling waves of different faults in the same branch in one embodiment of the present invention;

[0028] Figure 3A is a schematic diagram of zero-mode voltage traveling waves at the head end of different branch faults in one embodiment of the present invention;

[0029] Figure 3B is a schematic diagram of line mode voltage traveling waves at the head end of different branch faults in one embodiment of the present invention;

[0030] Figure 4A is a schematic diagram of the structure of a Convnext network in one embodiment of the present invention;

[0031] Figure 4B is a schematic diagram of the structure of Convnext Block in one embodiment of the present invention;

[0032] Figure 4C is a schematic structural diagram of a residual module in one embodiment of the present invention;

[0033] Figure 5It is a Convnext-Resnet18 hybrid neural network model in one embodiment of the present invention;

[0034] Figure 6 is a schematic diagram of a fault location process in an embodiment of the present invention;

[0035] Figure 7 It is a structural schematic diagram of a single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram in one embodiment of the present invention;

[0036] Figure 8 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0037] Fig. 9 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] The distribution network is an indispensable part of the power system. Its line layout is extensive and the structure is complex. It usually presents a tree-like multi-branch and multi-segmented form. Therefore, fault location of multi-branch distribution network is more complicated than that of transmission network. The main technical means of fault location of distribution network at present include impedance method, distribution automation method and traveling wave fault location method. These methods have made breakthroughs in both theory and practice, providing strong support for the stable operation of power system.

[0040] Although existing methods can locate the fault section of the distribution network to a certain extent, most of them rely on a single model to establish the mapping relationship between the fault signal and the fault location. There is a lack of in-depth utilization of fault information, which is easily disturbed by different fault conditions, resulting in limited positioning accuracy. When dealing with radial distribution networks with multiple branches, due to the complexity of the network structure, a single model is often difficult to accurately capture the precise mapping relationship between the fault signal and the specific location, and cannot accurately locate the fault.

[0041] When a fault occurs, the global wavelet time-frequency diagram of the line-mode traveling wave contains rich fault features, and these information are not significantly related to the precise calibration of the wave velocity, the natural frequency of the traveling wave, and the wave head. If the association between the time-frequency image and the fault location in the traveling wavelet time-frequency diagram can be effectively utilized, accurate fault location of the multi-branch structure of the distribution network under the condition of a weak high-resistance grounding fault can be achieved. The present invention constructs a mapping relationship between the line-mode traveling wave time-frequency diagram and the fault distance, and proposes a single-end positioning method for a multi-branch distribution network based on the mining of the global traveling wave time-frequency diagram features.

[0042] The multi-branch distribution network fault location model proposed in the present invention takes the residual convolutional neural network as the core to construct a comprehensive location model. The model deeply explores the time-frequency domain distribution characteristics of the global traveling wave time-frequency diagram, successfully establishes a single mapping network between each branch time-frequency diagram and the fault distance, and realizes accurate fault location under the radial network structure of the single mother line of the distribution network. This method does not need to pay attention to the extraction of the natural frequency of the traveling wave and the identification of the wave head, and can maintain a high location accuracy even in the case of high-resistance grounding faults.

[0043] See also Figure 1 As shown, Figure 1 A schematic flow chart of a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram provided by an embodiment of the present invention includes the following steps:

[0044] S10: Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the transmission characteristics of the single branch line.

[0045] When a fault occurs in the distribution network line, the fault transient wave generated by it will propagate along the line. When the transient wave encounters a fault point or busbar during propagation, refraction and reflection will occur, causing the stable electrical signal to suddenly change and become distorted.

[0046] Use PSCAD / EMTDC to establish Figure 2A The 10kv distribution network model shown in the figure sets HT1, IT1, T1T3, T3T2, and T2G as five experimental lines, namely L1, L2, L3, L4, and L5. Experimental simulation is carried out on each line respectively. A new fault point is taken as a distance of 0.5km to simulate different fault conditions. An integrated capacitive coupler traveling wave detection device is installed at the H point at the head end of the multi-branch line of the distribution network to detect the traveling wave waveform generated by each fault point of the distribution line, and the transient waveform is subjected to wavelet time-frequency transformation to establish a branch sample library for each region. Single-phase grounding faults f1 and f2 are set at 3.5km and 4.5km from the T1 node on branch IT1, and single-phase grounding faults f3 and f4 are set on branches T1T3 and T2T3. The initial phase angle of the fault is 90°, and the transition resistance is set to 100Ω.

[0047] Assuming that a single-phase grounding fault f occurs in HT1, the first traveling wave detected at the H terminal is directly propagated from the fault point; the second traveling wave detected has experienced propagation from the fault point to the branch node T1, reflected at T1, refracted at the fault point, and finally arrived at the H terminal; the third traveling wave is captured by the H terminal again after being reflected between the H terminal and the fault point. For any given detection point, the transient signal detected is essentially the result of the superposition of each traveling wave in a specific time sequence after the initial traveling wave of the fault is refracted at the wave impedance discontinuity. The reverse traveling wave and forward traveling wave of the fault voltage detected at the H terminal can be represented by the following formulas (1.1) and (1.2), respectively.

[0048]

[0049] Among them, E1 is the initial traveling wave; Q is the transmission function; K and W are the refraction coefficient and reflection coefficient of the fault traveling wave at the wave impedance discontinuity point, respectively; the subscripts of Q, W, and H represent the transmission path of the fault traveling wave; B and F in the U subscript represent the reverse traveling wave and the forward traveling wave, respectively.

[0050] When a single-phase grounding fault occurs in different branch lines, the traveling waves detected at the head end will also follow the rules similar to those of formula (1.1) and formula (1.2). Due to the uniqueness of the structure of each node in the distribution network, the refraction, reflection, transmission path and arrival timing of the traveling waves at each node will be different. These differences will lead to changes in the specific forms of each superposition term when expressing the transient characteristics of the traveling wave. Under different fault lines and different fault distances, the transient waveforms of the traveling waves detected at the H end of the distribution network bus head when a single-phase grounding fault occurs will show significant differences.

[0051] It should be understood that, based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the single branch line transmission characteristics, which refers to the process of how to obtain the single branch line transmission characteristics. Among them, in step S10, that is, based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the single branch line transmission characteristics, including the following steps:

[0052] S21: Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the transmission characteristics of the single branch line, including:

[0053] S22: determining the traveling wave arrival time data of the distribution network fault traveling wave signal based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal;

[0054] S23: based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, calculating the propagation distance of the traveling wave to obtain the traveling wave propagation distance data;

[0055] S24: determining reference traveling wave velocity data according to the traveling wave arrival time data and the traveling wave propagation distance data;

[0056] S25: performing difference analysis on the reference traveling wave velocity data and the actual traveling wave velocity data to obtain traveling wave velocity analysis data;

[0057] S26: Based on the amplitude data of the global traveling wave time-frequency diagram, performing a traveling wave amplitude attenuation analysis based on the time-frequency diagram to obtain traveling wave amplitude analysis data;

[0058] S27: Based on the reflection data of the global traveling wave time-frequency diagram, performing traveling wave reflection characteristic analysis based on the time-frequency diagram to obtain traveling wave reflection characteristic analysis data;

[0059] S28: Determine the single-branch line transmission characteristics based on the traveling wave velocity analysis data, the traveling wave amplitude analysis data and the traveling wave reflection characteristic analysis data.

[0060] For example, Figure 2B As shown, Figure 2B The transient waveforms of different fault branches are shown as examples. During the transmission of the fault traveling wave along the line, the initial traveling wave amplitude is related to the transmission distance x and the frequency component f contained in itself. Optionally, refer to the following formula (1.3):

[0061]

[0062] 其中， A(ω) 为线路传输函数，w为角频率，e为自然常数， γ 为传播系数，x为传输距离， γ(ω) is the transmission coefficient function, R0, L0, C0, G0 are the resistance, inductance, capacitance, and conductance per unit length of the conductor, respectively, and are all frequency-dependent functions; j is an imaginary unit, α(ω) is the attenuation coefficient, β(ω) is the phase coefficient, π is the circumference, and f is the fault wave frequency. Obtain the frequency-dependent parameters of the 10kV distribution line, and use formula (1.3) to obtain the law of how the amplitude of the transmission function is affected by the transmission distance and the wave frequency.

[0063] like Figure 2C As shown, Figure 2C The schematic diagram of the amplitude attenuation of the traveling wave transmission function is shown as an example. As the transmission distance increases, the traveling wave amplitude will gradually decay. The farther the fault distance is, the weaker the traveling wave amplitude propagated from the fault point to the head end detection is. The high-frequency component of the fault traveling wave decays faster than the low-frequency component. The traveling wave amplitude detected by the head end is greatly related to the transmission distance of the fault feeder.

[0064] Depend on Figure 2C It can be seen that the transformation process of the traveling wave amplitude is actually the same as energy transfer. In the fault line, the energy of the traveling wave significantly exceeds that of the non-fault line. The traveling wave propagation energy between different fault branches also shows obvious differences. With the help of continuous wavelet transform technology, the signal decomposition of the traveling wave time domain image is carried out, which significantly enhances the characteristic information of the traveling wave time-frequency image. The multi-branch lines in the distribution network also show different characteristics in their refraction and reflection traveling wave processes under different fault scenarios.

[0065] like Figure 2D As shown, Figure 2D The transient waveform diagram of the same branch fault is shown as an example. Optionally, with the help of line transfer function expression formula (1.3), the traveling wave velocity can be further expressed as the following formula (1.4) and formula (1.5):

[0066]

[0067] Where, v(ω) is the phase velocity of the traveling wave, which is a function of the angular frequency; ω is the angular frequency; β(ω) is a function related to the line parameters and frequency; R0, L0, C0, G0 are the resistance, inductance, capacitance, and conductance per unit length of the conductor, respectively; π is pi, f0 is the rated frequency of the system, and f is the fault traveling wave frequency; v n (f) is the traveling wave velocity of phase n when the traveling wave frequency is f; R n , L n , C n , G n They are the resistance, inductance, capacitance and conductance per unit length of the n-phase line.

[0068] From formula (1.4), we know that the speed of traveling wave is related to its own frequency. There is a difference in the speed of zero-mode traveling wave and line-mode traveling wave. The measurement error of traveling wave speed will affect the accuracy of traditional traveling wave fault location method. Therefore, the speed expression of different mode components can be obtained by formula (1.4) as follows, and the influence of frequency change on mode speed can be qualitatively analyzed. And the relationship between mode speed and frequency can be obtained from formula (1.5), as follows: Figure 2E shown.

[0069] After Karenbell transform processing, the zero-mode and line-mode components are obtained, among which the traveling wave transmission speeds of the three mode components gradually increase with the increase of frequency, the 1-mode and 2-mode wave speeds gradually tend to be stable after the frequency reaches 10kHz, and the 0-mode traveling wave transmission speed has the same increasing trend as the frequency. In view of the uncertainty of the wave speed, if the zero-mode traveling wave is used to process the global time-frequency diagram, the head-end zero-mode traveling wave will be greatly affected by interference. Therefore, the line-mode traveling wave that is not easily affected by frequency interference is used as the fault model sample library data in the present invention.

[0070] Optionally, after performing wavelet analysis on the traveling wave waveform detected by the device, the following can be obtained: Figure 2F The full-domain time-frequency diagram of the traveling wave is shown. Figure 2F It can be seen from the different fault conditions shown in the figure that there are obvious differences in the time-frequency diagrams of different fault distances in the same branch of the distribution network. In the case of short distances, the high-frequency fault information is more obvious, and the high-frequency component of the traveling wave in the fault line is severely attenuated as the distance increases. The arrival timing and frequency distribution of each wave head of the fault traveling wave at different fault distances in the same branch are different.

[0071] For different fault distances of branch lines of multi-branch distribution grid faults, the refraction and reflection process of the traveling wave and the timing of reaching the endpoint are slightly different. However, in different frequency bands, the fault point location is different, the attenuation degree of the high frequency band of the wavelet time-frequency graph is different, and the time-frequency graph under different branch faults is also different, and the difference is still quite obvious. When the fault point is close to the monitoring point, the time-frequency energy of the traveling wave is obviously present in the high-frequency area. When the fault point is far away from the monitoring point, the frequency bands of the traveling wave are obviously attenuated with the fault distance during transmission, the low-frequency component exists steadily, and the time-frequency energy of the traveling wave gradually concentrates in the low-frequency area.

[0072] Obviously, it can be concluded that:

[0073] 1) According to the different traveling wave transmission rates and attenuation differences of each frequency band under different fault branches, the multi-branch power grid fault line can divide each branch line into different fault sections. When a ground fault occurs in each fault section, the traveling wave transmitted from the fault point to the head end of the line is different;

[0074] 2) For different fault points under the same fault branch, the degree of high-frequency attenuation of the traveling wave varies with the change of fault distance, and there are obvious differences in the time-frequency diagram of the traveling wave;

[0075] 3) There are obvious differences in the global characteristics of wavelet time-frequency diagrams under different fault distances. The global wavelet time-frequency diagram is used to establish a connection with the fault distance. The time-frequency characteristics in the global wavelet time-frequency diagram are fully mined to characterize the mapping relationship. The arrival timing of each wave head in the global wavelet time-frequency diagram is deeply mined. The complex nonlinear relationship between input features and output values ​​is learned by data regression to process the fault distance, and an accurate fault location model for the distribution network is obtained.

[0076] S20: Based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics.

[0077] Since it is impossible to install multiple sensors on each branch line to monitor the line operation status, the actual application will inevitably involve the cross-branch transmission and detection of the traveling wave signal. Therefore, it is necessary to analyze the transmission characteristics of the traveling wave signal after it propagates across the branches. Figure 2AIn the simulation model shown in the figure, the ground fault of lines L1, L2, L3 and L4 in phase A is used as a typical case for simulation analysis. In the simulation process, the total duration is set to 0.01 seconds, and the fault is set to occur at 0.005 seconds. Through experimental simulation in the PSCAD-Matlab simulation environment, the three-phase voltage and current are obtained at the head end of the multi-branch power grid line. The line mode traveling wave is obtained by Karen Bell transformation. Figure 3A The waveform of the voltage zero-mode traveling wave detected at the H end of the distribution network busbar head when a fault occurs is shown. Figure 3B The voltage line mode waveform under fault condition is depicted.

[0078] Depend on Figure 3B It can be seen that the images detected at the head end of the fault in different branches of the distribution network are different. The difference in the traveling wave signals of different branch lines is mainly due to the refraction and reflection of each node and the influence of the traveling wave transmission function. The fault traveling waves detected at the head end of different branches and different fault distances are mainly the difference in refraction and reflection during the traveling wave transmission process. The transmission rate and attenuation level of each frequency component of the traveling wave are different during the transmission process. Theoretically, when different grounding faults occur in each branch of the multi-branch line of the distribution network, the traveling waves detected at the head end of the bus and the transformed panoramic fault traveling wave time-frequency diagram also have obvious differences.

[0079] S30: constructing a distribution network fault location model based on a residual network according to the single-branch line transmission characteristics and the inter-branch line transmission characteristics, and obtaining an initial distribution network fault location model.

[0080] In the accurate fault location of distribution network, a major problem is how to simultaneously determine the fault branch and accurately measure the fault distance of this branch when a fault occurs. Among them, judging the multi-branch fault branch is a traditional feature screening problem, while the measurement of fault distance is a nonlinear regression problem of data feature quantity. It is particularly important to determine the fault branch and fault distance after extracting the global traveling wave fault information.

[0081] It should be understood that, according to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, the distribution network fault location model is constructed based on the residual network to obtain the initial distribution network fault location model, which refers to the process of constructing and obtaining the initial distribution network fault location model. Among them, in step S30, that is, according to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, the distribution network fault location model is constructed based on the residual network to obtain the initial distribution network fault location model, including the following steps:

[0082] S31: constructing a transmission feature vector based on the single-branch line transmission characteristics and the inter-branch line transmission characteristics;

[0083] S32: Determine the number of network layers and module structure based on the residual network and the transmission feature vector to obtain a distribution network fault location network structure;

[0084] S33: constructing and obtaining a fault branch identification module based on the distribution network fault location network structure;

[0085] S34: constructing and obtaining a fault distance measurement module based on the distribution network fault location network structure;

[0086] S35: Determine an initial distribution network fault location model based on the fault branch identification module and the fault distance measurement module.

[0087] After in-depth analysis of the time-frequency characteristics of the global waveform of the traveling wave, the present invention adopts a two-layer path analysis scheme, that is, first using the Convnext model to identify the faulty branch on the global wavelet time-frequency diagram detected at the head end, and then using the residual convolutional neural network to regress the data feature quantity, and using the key time-frequency domain information and time-frequency feature differences of each traveling wave in the global wavelet time-frequency diagram to determine the faulty branch, and then using Resnet18 to reduce the dimensionality of the high-dimensional time-frequency features, and according to the arrival timing of each frequency wave head reflecting the fault distance information, a two-layer path output is established to specifically judge the faulty branch and locate the fault in the actual distribution network.

[0088] Convnext is an image classification model proposed by DeepMind in 2022. It builds on the success of Transformer and proposes a new convolution operation called deep convolution (DCN). DCN is a divide-and-conquer convolution operation that decomposes the convolution into a series of smaller convolutions, each of which is applied to a different part of the input feature map. This allows the model to focus on different local areas of the feature map, thereby improving the representational power of the model. The architecture of the Convnext model is similar to Transformer. It consists of multiple stages, each of which contains a DCN block, a normalization layer, and an activation function, and the depth of the model increases with the number of stages. Convnext is trained using contrastive learning.

[0089] Contrastive learning is a self-supervised learning technique that teaches the model to cluster positive samples (images belonging to the same class) together and separate negative samples (images belonging to different classes). The network structure diagram is as follows: Figure 4A As shown in Figure 1, each ConvnextBlock is an inverted residual structure consisting of a convolutional layer, layer normalization (LN) and an activation function. The specific parts are as follows: Figure 4B shown.

[0090] This application applies the relevant principles of the Convnext model to the construction of the fault branch identification part. Using the deep convolution (DCN) operation of Convnext, the global wavelet time-frequency map is input into the Convnext structure composed of multiple stages. Each stage contains a DCN block, a normalization layer, and an activation function. The Convnext model is trained by contrastive learning, and the global wavelet time-frequency maps belonging to the same fault branch are clustered together as positive samples, and those of different fault branches are separated as negative samples, so that the model can learn the differences in the time-frequency map features of different fault branches, thereby distinguishing the fault branch.

[0091] ResNet-18 solves the gradient problem by introducing residual connections, breaking the layer-by-layer connection method of traditional networks. Among them, ResNet-18 is one of the simplest and most classic residual networks. It uses 18 layers of depth and builds the network by stacking residual blocks.

[0092] The core of ResNet-18 is the residual block, which consists of two convolutional layers, and each convolutional layer has a shortcut connection with identity mapping. This shortcut connection directly bypasses the convolutional layer and adds the input feature map to the output feature map. There are large gradient changes between different layers of the network, and the network can effectively backpropagate gradient information. Through residual connections, ResNet-18 can effectively solve the problems of gradient vanishing and gradient exploding. When the network deepens, traditional networks are prone to gradients that are too small or even disappear, resulting in the inability to effectively train and converge the network. The residual connection can directly add the input feature map to the output feature map, so that the gradient can be better propagated and maintain a larger gradient amplitude. Its typical residual module structure is as follows: Figure 4C shown.

[0093] The fault distance measurement part is built based on ResNet-18. The feature vector constructed previously is input into the ResNet-18 network. Through the stacking of residual blocks, the network can reduce the dimensionality of high-dimensional time-frequency features. For example, as the data propagates in the network, through the residual connection, the network can gradually extract the key features related to the fault distance while avoiding the gradient vanishing problem. According to the characteristics such as the arrival timing of each frequency wave head reflecting the fault distance information, the network continuously adjusts the parameters during the training process to establish a mapping relationship between the features and the fault distance, and realizes the measurement of the fault distance.

[0094] BN is batch normalization in neural networks, which can solve the gradient diffusion and gradient explosion problems of deep networks during training. Its main principle is to normalize the input features so that the features maintain zero mean and unit variance during training, making the network insensitive to the scale of the input data, thereby avoiding the problem of performance degradation of deep networks and accelerating network training and convergence. The batch normalization process used in the present invention is as follows:

[0095]

[0096] in, is the sample mean, x i is the i-th sample in the sample data, i is the sample index, m is the number of samples, is the sample variance, is the standardized sample value, ε is a very small integer, y i is the value after linear transformation, γ and β are the coefficients of linear transformation; BN γ,β (x i ) represents the i Batch normalization is performed and then linear transformation is performed.

[0097] The deep residual network is an algorithm based on SkipConnection published by Kaiming He et al. of Microsoft Research Asia in 2015. The network structure of Resnet-18 is shown in Table 1:

[0098] Table 1

[0099]

[0100] S40: Based on a graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model.

[0101] The multi-distribution distribution network fault location model built by this patent (i.e. the above-mentioned target distribution network fault location model) is built with reference to the classic CNN classic small-scale model and Resnet-18 network structure. Considering the complexity of the multi-distribution distribution network fault line location problem, the Convnext fault branch judgment model Resnet-18 model consists of 1 convolution layer and 8 residual basic modules, including 17 convolution layers. The global traveling wave time-frequency diagram retains the time-frequency energy characteristics of the distribution network fault traveling wave signal and fully mines the deep-level characteristics of the disturbance signal. Based on the idea of ​​feature map regression, while learning the wavelet time-frequency energy characteristics, the deep-level characteristics are mined to build a higher accuracy fault location model.

[0102] It should be understood that, based on the graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain the target distribution network fault location model, which refers to the process of how to obtain the target distribution network fault location model. Among them, in step S40, that is, based on the graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain the target distribution network fault location model, including the following steps:

[0103] S41: Based on a graph hybrid convolutional neural network, the initial distribution network fault location model is structurally adjusted according to the global traveling wave time-frequency graph to obtain a reconstructed distribution network fault location model;

[0104] S42: dynamically adjusting the model parameters based on the adaptive moment estimation method to obtain the optimized model parameters;

[0105] S43: Iteratively updating the reconstructed distribution network fault location model based on the optimized model parameters and hyperparameters to obtain a hybrid neural network model;

[0106] S44: Performing dual-channel mixed convolution processing based on the initial distribution network fault location model and the hybrid neural network model to obtain a target distribution network fault location model.

[0107] The jump connection of the Resnet-18 network crosses the identity mapping of some layers, and the result of the identity mapping is added to the output of the convolution layer, which is more conducive to the feature mining of the wavelet time-frequency map. Components such as the convolution layer, pooling layer, and activation function work together to effectively map the feature data in the global wavelet time-frequency map to the hidden layer feature space. This process not only helps to extract key information from the input data, but also realizes the effective abstraction and conversion of features.

[0108] The fully connected layer at the end of the Resnet-18 network maps the features learned in the front part of the network to the fault distance, achieving accurate prediction and classification of the target task. Through a series of mapping and conversion processes, the Resnet-18 network can make full use of the feature information of the global wavelet time-frequency graph to achieve accurate inference of the fault distance. The network model structure is as follows Figure 5 shown.

[0109] This patent introduces an adaptive moment estimation method that can dynamically adjust the learning rate for each parameter or weight. The core idea of ​​the adaptive moment estimation method is to calculate the gradient value of the fault distance measurement error with respect to the model parameters, and then calculate the first-order moment estimation and the second-order moment estimation, and on this basis correct the deviation of the first-order moment estimation and the second-order moment estimation, and finally update the optimized parameters.

[0110] For parameters with larger gradients, the learning rate will be reduced accordingly to avoid oscillation caused by too fast parameter updates. For parameters with smaller gradients, the learning rate will be increased accordingly to accelerate convergence. This adaptive adjustment of the learning rate helps the model converge faster and more accurately during training. This method is particularly suitable for dealing with oscillation problems in the iterative process of distribution network fault traveling wave signals. This method has good adaptability and robustness. Through this optimization strategy, the performance of the model in the distribution network fault location task can be significantly improved. The iterative process is as follows:

[0111]

[0112] in, is the variable θ i The value at the k+1th iteration, is the variable θ i The value at the kth iteration, is the variable θ at iteration k i The gradient value, η is the learning rate, To update the variable v i The value after the kth iteration, is the intermediate variable s i The value after the kth iteration, ε is the regularization term, is the updated variable v in the kth iteration i The value of is the value of the hyperparameter β1 at the kth iteration, is the intermediate variable s in the kth iteration i The value of is the value of the hyperparameter β2 at the kth iteration, is the updated variable v in the k+1th iteration i The value of β1 is a hyperparameter, E R is the error function, is the error function for the variable θ i The partial derivative of (k) is the variable θ i The value at the kth iteration, is the intermediate variable s in the k+1th iteration i The value of β2 is a hyperparameter. Among them, 0≤β1<1, 0≤β2<1.

[0113] Based on the above theory, this patent constructs a hybrid neural network model for the problem of fault branch identification in the distribution network. The input of the model is the line mode traveling wave diagram of the distribution network collected in real time, which is converted into a global traveling wave time-frequency diagram through running continuous wavelet transform.

[0114] Inside the model, the Convnext module is responsible for extracting frequency domain distribution information from the input time-frequency diagram. This information is further processed by the fully connected layer. When the input image enters the Convnext model, it first passes through a series of convolutional layers, normalization layers, activation functions and other structures for forward propagation. In this process, each convolutional layer will use its corresponding weights and bias items to perform convolution operations on the input feature map, thereby extracting different features in the image and mapping them into a probability distribution yg through the fully connected layer.

[0115] The global average pooling layer compresses the spatial dimension of the feature map into a feature vector, and the fully connected layer uses this feature vector as input and outputs the predicted probability distribution of each category through linear transformation and activation function. Finally, this predicted distribution yg is used to accurately identify the fault branch:

[0116]

[0117] y k =f(w k down(y mn )+b k )(5.4)

[0118] y g =f(w g x+b g )=[q1,q2,q3...q l ] (5.5)

[0119] in, is the output value of node z in the j-th layer network; f is the activation function; is the output value of node x in the previous layer (j-1 layer); is the weight connecting the j-1th layer x node and the jth layer node z; is the bias term of the l-th layer node z; y k is the output value of the kth layer; w k is the weight parameter; down(y mn ) is for y mn Perform downsampling operation, y mn is the input data after processing; b k is the bias term of the kth layer; y g is the final output value of the model; w g is the weight parameter; x is the input data; b g is the bias term of the g-th layer; [q1,q2,q3...q l ] is the specific form of the output result, that is, the probability distribution vector of each branch.

[0120] Formula (5.5) is the calculation formula of the pooling layer of the hybrid convolutional neural network, in which the pooling layer uses a characteristic position in the global traveling wave time-frequency diagram and the overall characteristics of its adjacent outputs as the network output at this position, down(.) is the maximum (max) pooling function, yk is the output of the pooling layer, where x is the feature vector, yg is the fully connected network output, and the probability of identifying and classifying the global traveling wave time-frequency diagram into the i-th branch is q1, and l represents the total number of branches in the distribution network.

[0121] In the method mentioned in this patent, the Resnet18 model uses the global traveling wave time-frequency diagram as input and processes it through a series of convolutional layers, residual blocks, and fully connected layers. When Convnext determines the fault branch, Resnet18 performs the fault data regression task for this branch. The convolutional layers and pooling layers are consistent, but the last layer is usually a fully connected layer associated with the fault distance, and the number of its output units matches the dimension of the regression target.

[0122] The regression target is the distance value associated with the image. Through the back-propagation algorithm and the gradient descent optimizer, the weights and biases of the model can be adjusted to minimize the difference between the predicted distance and the actual distance. After enough training iterations, the Resnet18 model should be able to learn an effective mapping relationship from the input image to the target distance. For image data regression, the residual convolutional neural network uses the weights and bias parameters generated by its own fully connected layer to map the feature vector Rt to the fault point location y2.

[0123] y2=f[R1,w2,b2]=R t T W2+b2

[0124] Among them, y2 is the location of the fault point, which is the final target of the model; f is the activation function; R t T is the feature vector; W2 is the weight; b2 is the bias.

[0125] The basic logic of the dual-channel hybrid convolutional neural network fault location model proposed in this patent is as described above. In the proposed Convnext-Resnet18 dual-channel multi-branch distribution network, Convnext mainly focuses on capturing the spatial frequency domain features in the waveform of the global time-frequency graph of the line-mode traveling wave, and extracts the instantaneous frequency distribution features of its global traveling wave time-frequency graph by stacking convolution layers, compressing the spatial dimensions of its feature graph into feature item vectors corresponding to different branches for model classification; while Resnet-18 performs fine processing on the frequency and wave head arrival time features of the line-mode traveling wave, and uses the rich time-frequency information in the global traveling wave time-frequency graph for regression ranging. Therefore, by cleverly connecting these two network models with different structures in parallel, the scheme proposed in this patent can comprehensively and deeply extract fault features from the global wavelet time-frequency graph, both in the spatial frequency domain and the time domain. After multiple iterative training, the resulting distribution network fault location model has shown excellent performance in mining the global traveling wavelet time-frequency graph features for fault ranging.

[0126] S50: performing single-end fault location on the multi-branch distribution network according to the target distribution network fault location model to obtain a target fault location result.

[0127] The overall implementation idea of ​​the method proposed in the present invention is to use the difference in the traveling wave line mode voltage at the head end after a distribution network fault occurs, generate time-frequency images of the traveling wave signal using continuous wavelet transform, use the global traveling wave time-frequency diagram as the data set, input the constructed model, use the convolution layer in the residual convolutional neural network to deeply mine the implicit features of the image, transform the features through the fully connected layer to correspond to the sample label, perform image data regression according to the implicit image features, and measure the fault distance.

[0128] The fault location process of the present invention is as follows: Figure 6 As shown in the figure, the specific steps of multi-distribution power grid fault location based on time-frequency multi-scale image and deep residual neural network are as follows:

[0129] 1) At the initial end of the distribution network, the traveling wave line mode components of the multi-branch distribution network are collected and imported into MATLAB, and continuous wavelet transform is performed on them to obtain a three-dimensional multi-scale time-frequency diagram, which is then converted into a two-dimensional time-frequency sequence.

[0130] 2) The time-frequency image is processed by dimensionality reduction and the Convnext-Resnet-18 residual convolutional neural network is used for image recognition and feature extraction. The actual fault distance is used as the label of the corresponding sample of each branch to form a multi-branch power grid fault sample set. The sample set is divided into training set and test set according to 4:1.

[0131] 3) Construct the Convnext-Resnet18 model as in the previous steps, set the basic parameters of the fault location model and the weights of each convolution layer, and set the number of iterations, learning rate, and number of samples required for each iteration in combination with the Adam optimization algorithm. Substitute the simulation data into the Convnext-Resnet18 model, and initialize the cumulative amount and square cumulative amount of the gradient. Use the data graph obtained in the test set to verify the fault branch judgment and ranging effect of the Convnext-Resnet18 model proposed in the present invention. If the positioning accuracy and branch judgment can reach the expected accuracy, the training is terminated. If the expected accuracy is not reached, readjust the parameters and learning rate until the accuracy requirements required by the present invention are met.

[0132] 4) After a fault occurs in the distribution network, the fault line is first located and the line selection result is obtained. Then, the integrated capacitive coupler studied by the research team is used at the head end to detect the initial transient traveling wave at the head end. The 3ms time window after the arrival of the transient traveling wave is selected, and the obtained data is processed by applying continuous wavelet transform. The global wavelet time-frequency graph is input into the trained distribution network fault location model, and the specific weight parameters of each convolution layer and pooling layer are loaded. The bias item is set for each convolution layer, and the last layer (full connection layer category probability) is replaced with a fully connected layer that outputs a single value (fault distance). The trained model is used to determine the fault line and predict the fault distance. The detected data samples are updated to the total database, and the weight and bias item parameters are updated according to the Convnext-Resnet18 model.

[0133] In order to optimize the computational efficiency of the residual convolutional layer network, this method requires that the dimension of the traveling wave line mode wavelet time-frequency diagram be reconstructed to 224 × 224 using the bilinear interpolation method when building the model. In order to reduce the impact of different working conditions on the traveling wavelet time-frequency energy amplitude, the maximum and minimum normalization method is used to normalize the values ​​of each element in the matrix to the interval [0,1], thereby improving the accuracy of positioning.

[0134] It can be seen that in the above scheme, based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed to obtain the single branch line transmission characteristics, and based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics. According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model can be further constructed based on the residual network to obtain an initial distribution network fault location model, so that based on the graph hybrid convolutional neural network, the initial distribution network fault location model can be optimized and improved according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model, and then according to the target distribution network fault location model, single-end fault location can be performed on multi-branch distribution networks to obtain a more accurate target fault location result.

[0135] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0136] In one embodiment, a single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram is provided, and the single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram corresponds one-to-one to the single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram in the above-mentioned embodiment. As shown in FIG3 , the single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram includes a first processing module 101, a second processing module 102, a third processing module 103, a fourth processing module 104, and a fifth processing module 105. Each functional module is described in detail as follows:

[0137] The first processing module 101 is used to analyze the transmission characteristics of the fault traveling wave under the single branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, and obtain the transmission characteristics of the single branch line;

[0138] The second processing module 102 is used to analyze the transmission characteristics of the fault traveling wave under the cross-branch line based on the global traveling wave time-frequency diagram to obtain the cross-branch line transmission characteristics;

[0139] The third processing module 103 is used to construct a distribution network fault location model based on a residual network according to the single branch line transmission characteristics and the inter-branch line transmission characteristics, so as to obtain an initial distribution network fault location model;

[0140] A fourth processing module 104 is configured to optimize and improve the initial distribution network fault location model based on a graph hybrid convolutional neural network and according to the global traveling wave time-frequency graph to obtain a target distribution network fault location model;

[0141] The fifth processing module 105 is used to perform single-end fault location on the multi-branch distribution network according to the target distribution network fault location model to obtain a target fault location result.

[0142] In one embodiment, the first processing module 101 is used to analyze the transmission characteristics of the fault traveling wave under the single branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal to obtain the single branch line transmission characteristics, which is specifically used to:

[0143] Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, determining the traveling wave arrival time data of the distribution network fault traveling wave signal;

[0144] Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the propagation distance of the traveling wave is calculated to obtain the traveling wave propagation distance data;

[0145] Determining reference traveling wave velocity data according to the traveling wave arrival time data and the traveling wave propagation distance data;

[0146] Performing difference analysis on the reference traveling wave velocity data and the actual traveling wave velocity data to obtain traveling wave velocity analysis data;

[0147] Based on the amplitude data of the global traveling wave time-frequency diagram, performing a traveling wave amplitude attenuation analysis based on the time-frequency diagram to obtain traveling wave amplitude analysis data;

[0148] Based on the reflection data of the global traveling wave time-frequency diagram, performing traveling wave reflection characteristic analysis based on the time-frequency diagram to obtain traveling wave reflection characteristic analysis data;

[0149] Based on the traveling wave velocity analysis data, the traveling wave amplitude analysis data and the traveling wave reflection characteristic analysis data, a single branch line transmission characteristic is determined.

[0150] In one embodiment, the third processing module 103 is used to construct a distribution network fault location model based on a residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics to obtain an initial distribution network fault location model, specifically for:

[0151] Constructing a transmission feature vector based on the single-branch line transmission characteristics and the inter-branch line transmission characteristics;

[0152] Determine the number of network layers and module structure based on the residual network and the transmission feature vector to obtain a distribution network fault location network structure;

[0153] Based on the distribution network fault location network structure, a fault branch identification module is constructed and obtained;

[0154] Based on the distribution network fault location network structure, a fault distance measurement module is constructed and obtained;

[0155] Based on the fault branch identification module and the fault distance measurement module, an initial distribution network fault location model is determined.

[0156] In one embodiment, the fourth processing module 104 is used to optimize and improve the initial distribution network fault location model based on the graph hybrid convolutional neural network according to the global traveling wave time-frequency graph to obtain a target distribution network fault location model, specifically for:

[0157] Based on a graph hybrid convolutional neural network, the structure of the initial distribution network fault location model is adjusted according to the global traveling wave time-frequency graph to obtain a reconstructed distribution network fault location model;

[0158] Dynamically adjust the model parameters based on the adaptive moment estimation method to obtain the optimized model parameters;

[0159] Based on the optimization model parameters and hyperparameters, the reconstructed distribution network fault location model is iteratively updated to obtain a hybrid neural network model;

[0160] Based on the initial distribution network fault location model and the hybrid neural network model, dual-channel hybrid convolution processing is performed to obtain the target distribution network fault location model.

[0161] In one embodiment, the fourth processing module 104 is used to iteratively update the reconstructed distribution network fault location model based on the optimization model parameters and hyperparameters to obtain a hybrid neural network model, which is specifically used to:

[0162] The following formula is used to implement the iterative update of the reconstructed distribution network fault location model based on the optimized model parameters and hyperparameters to obtain the hybrid neural network model:

[0163]

[0164] in, is the variable θ i The value at the k+1th iteration, is the variable θ i The value at the kth iteration, is the variable θ at iteration k i The gradient value, η is the learning rate, To update the variable v i The value after the kth iteration, is the intermediate variable s i The value after the kth iteration, ε is the regularization term, is the updated variable v in the kth iteration i The value of is the value of the hyperparameter β1 at the kth iteration, is the intermediate variable s in the kth iteration i The value of is the value of the hyperparameter β2 at the kth iteration, is the updated variable v in the k+1th iteration i The value of β1 is a hyperparameter, E R is the error function, is the error function for the variable θ i The partial derivative of (k) is the variable θ i The value at the kth iteration, is the intermediate variable s in the k+1th iteration i The value of β2 is a hyperparameter.

[0165] The present invention provides a single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram. Based on the global traveling wave time-frequency diagram of a distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under a single branch line are analyzed to obtain the single branch line transmission characteristics, and based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics. According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model can be further constructed based on a residual network to obtain an initial distribution network fault location model, so that the initial distribution network fault location model can be optimized and improved based on a graph hybrid convolutional neural network according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model, and then according to the target distribution network fault location model, single-end fault location can be performed on a multi-branch distribution network to obtain a more accurate target fault location result.

[0166] For the specific limitations of the single-ended positioning device for a distribution network based on a global traveling wave time-frequency diagram, please refer to the limitations of the single-ended positioning method for a distribution network based on a global traveling wave time-frequency diagram in the above text, which will not be repeated here. Each module in the above-mentioned single-ended positioning device for a distribution network based on a global traveling wave time-frequency diagram can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0167] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in FIG4. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, the functions or steps on the service side of a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram are realized.

[0168] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it realizes the functions or steps of the client side of a single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram.

[0169] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0170] Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed, and the transmission characteristics of the single branch line are obtained;

[0171] Based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics;

[0172] According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model is constructed based on a residual network to obtain an initial distribution network fault location model;

[0173] Based on a graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model;

[0174] According to the target distribution network fault location model, single-end fault location is performed on the multi-branch distribution network to obtain a target fault location result.

[0175] The present invention provides a computer device, which analyzes the transmission characteristics of the fault traveling wave under a single branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal to obtain the single branch line transmission characteristics, and analyzes the transmission characteristics of the fault traveling wave under the cross-branch line based on the global traveling wave time-frequency diagram to obtain the cross-branch line transmission characteristics. According to the single branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model can be further constructed based on a residual network to obtain an initial distribution network fault location model, so that the initial distribution network fault location model can be optimized and improved based on a graph hybrid convolutional neural network according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model, and then single-end fault location can be performed on a multi-branch distribution network according to the target distribution network fault location model to obtain a more accurate target fault location result.

[0176] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0177] Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed, and the transmission characteristics of the single branch line are obtained;

[0178] Based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the cross-branch line transmission characteristics;

[0179] According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model is constructed based on a residual network to obtain an initial distribution network fault location model;

[0180] Based on a graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model;

[0181] According to the target distribution network fault location model, single-end fault location is performed on the multi-branch distribution network to obtain a target fault location result.

[0182] The present invention provides a computer-readable storage medium, which analyzes the transmission characteristics of the fault traveling wave under a single branch line based on a global traveling wave time-frequency diagram of a distribution network fault traveling wave signal to obtain the single branch line transmission characteristics, and analyzes the transmission characteristics of the fault traveling wave under the cross-branch line based on the global traveling wave time-frequency diagram to obtain the cross-branch line transmission characteristics. A distribution network fault location model can be further constructed based on a residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics to obtain an initial distribution network fault location model, so that the initial distribution network fault location model can be optimized and improved based on a graph hybrid convolutional neural network according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model, and then single-end fault location can be performed on a multi-branch distribution network according to the target distribution network fault location model to obtain a more accurate target fault location result.

[0183] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0184] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0185] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0186] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A single-end positioning method for distribution network based on global traveling wave time-frequency diagram, characterized in that: The method comprises: Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the transmission characteristics of the fault traveling wave under the single branch line are analyzed, and the transmission characteristics of the single branch line are obtained; Based on the global traveling wave time-frequency diagram, the transmission characteristics of the fault traveling wave under the cross-branch line are analyzed to obtain the transmission characteristics of the cross-branch line; According to the single-branch line transmission characteristics and the cross-branch line transmission characteristics, a distribution network fault location model is constructed based on a residual network to obtain an initial distribution network fault location model; Based on a graph hybrid convolutional neural network, the initial distribution network fault location model is optimized and improved according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model; According to the target distribution network fault location model, single-end fault location is performed on the multi-branch distribution network to obtain a target fault location result.

2. The single-end positioning method for distribution network based on global traveling wave time-frequency diagram according to claim 1 is characterized in that: The global traveling wave time-frequency diagram based on the distribution network fault traveling wave signal is used to analyze the transmission characteristics of the fault traveling wave under the single branch line, and obtain the transmission characteristics of the single branch line, including: Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, determining the traveling wave arrival time data of the distribution network fault traveling wave signal; Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the propagation distance of the traveling wave is calculated to obtain the traveling wave propagation distance data; Determining reference traveling wave velocity data according to the traveling wave arrival time data and the traveling wave propagation distance data; Performing difference analysis on the reference traveling wave velocity data and the actual traveling wave velocity data to obtain traveling wave velocity analysis data; Based on the amplitude data of the global traveling wave time-frequency diagram, performing a traveling wave amplitude attenuation analysis based on the time-frequency diagram to obtain traveling wave amplitude analysis data; Based on the reflection data of the global traveling wave time-frequency diagram, performing traveling wave reflection characteristic analysis based on the time-frequency diagram to obtain traveling wave reflection characteristic analysis data; Based on the traveling wave velocity analysis data, the traveling wave amplitude analysis data and the traveling wave reflection characteristic analysis data, a single branch line transmission characteristic is determined.

3. The single-end positioning method for distribution network based on global traveling wave time-frequency diagram according to claim 2 is characterized in that: The construction of the distribution network fault location model based on the residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics to obtain the initial distribution network fault location model includes: Constructing a transmission feature vector based on the single-branch line transmission characteristics and the inter-branch line transmission characteristics; Determine the number of network layers and module structure based on the residual network and the transmission feature vector to obtain a distribution network fault location network structure; Based on the distribution network fault location network structure, a fault branch identification module is constructed and obtained; Based on the distribution network fault location network structure, a fault distance measurement module is constructed and obtained; Based on the fault branch identification module and the fault distance measurement module, an initial distribution network fault location model is determined.

4. The single-end positioning method for distribution network based on global traveling wave time-frequency diagram according to claim 3 is characterized in that: The graph-based hybrid convolutional neural network optimizes and improves the initial distribution network fault location model according to the global traveling wave time-frequency diagram to obtain a target distribution network fault location model, including: Based on a graph hybrid convolutional neural network, the structure of the initial distribution network fault location model is adjusted according to the global traveling wave time-frequency graph to obtain a reconstructed distribution network fault location model; Dynamically adjust the model parameters based on the adaptive moment estimation method to obtain the optimized model parameters; Based on the optimization model parameters and hyperparameters, the reconstructed distribution network fault location model is iteratively updated to obtain a hybrid neural network model; Based on the initial distribution network fault location model and the hybrid neural network model, dual-channel hybrid convolution processing is performed to obtain the target distribution network fault location model.

5. The single-end positioning method for distribution network based on global traveling wave time-frequency diagram according to any one of claims 1 to 4, characterized in that: The reconstructed distribution network fault location model is iteratively updated based on the optimized model parameters and hyperparameters to obtain a hybrid neural network model, including: The following formula is used to implement the iterative update of the reconstructed distribution network fault location model based on the optimized model parameters and hyperparameters to obtain the hybrid neural network model: in, is the value of variable θi at the k+1th iteration, is the value of variable θi at the kth iteration, is the variable θ at iteration k i The gradient value, η is the learning rate, To update the variable v i The value after the kth iteration, is the intermediate variable s i The value after the kth iteration, ε is the regularization term, is the updated variable v in the kth iteration i The value of is the value of the hyperparameter β1 at the kth iteration, is the intermediate variable s in the kth iteration i The value of is the value of the hyperparameter β2 at the kth iteration, is the updated variable v in the k+1th iteration i The value of β1 is a hyperparameter, E R is the error function, is the error function for the variable θ i The partial derivative of (k) is the value of variable θi at the kth iteration, is the value of the intermediate variable si in the k+1th iteration, and β2 is a hyperparameter.

6. A single-end positioning device for a distribution network based on a global traveling wave time-frequency diagram, characterized in that: The single-end positioning device for the distribution network based on the global traveling wave time-frequency diagram comprises: The first processing module is used to analyze the transmission characteristics of the fault traveling wave under the single branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, and obtain the transmission characteristics of the single branch line; A second processing module is used to analyze the transmission characteristics of the fault traveling wave under the cross-branch line based on the global traveling wave time-frequency diagram to obtain the cross-branch line transmission characteristics; A third processing module is used to construct a distribution network fault location model based on a residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics, so as to obtain an initial distribution network fault location model; A fourth processing module is used to optimize and improve the initial distribution network fault location model based on a graph hybrid convolutional neural network and according to the global traveling wave time-frequency graph to obtain a target distribution network fault location model; The fifth processing module is used to perform single-end fault location on the multi-branch distribution network according to the target distribution network fault location model to obtain a target fault location result.

7. The single-end positioning device for distribution network based on global traveling wave time-frequency diagram according to claim 6, characterized in that: The first processing module is used to analyze the transmission characteristics of the fault traveling wave under the single-branch line based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal to obtain the transmission characteristics of the single-branch line, specifically for: Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, determining the traveling wave arrival time data of the distribution network fault traveling wave signal; Based on the global traveling wave time-frequency diagram of the distribution network fault traveling wave signal, the propagation distance of the traveling wave is calculated to obtain the traveling wave propagation distance data; Determining reference traveling wave velocity data according to the traveling wave arrival time data and the traveling wave propagation distance data; Performing difference analysis on the reference traveling wave velocity data and the actual traveling wave velocity data to obtain traveling wave velocity analysis data; Based on the amplitude data of the global traveling wave time-frequency diagram, performing a traveling wave amplitude attenuation analysis based on the time-frequency diagram to obtain traveling wave amplitude analysis data; Based on the reflection data of the global traveling wave time-frequency diagram, performing traveling wave reflection characteristic analysis based on the time-frequency diagram to obtain traveling wave reflection characteristic analysis data; Based on the traveling wave velocity analysis data, the traveling wave amplitude analysis data and the traveling wave reflection characteristic analysis data, a single branch line transmission characteristic is determined.

8. The single-end positioning device for distribution network based on global traveling wave time-frequency diagram according to claim 7, characterized in that: The third processing module is used to construct a distribution network fault location model based on a residual network according to the single branch line transmission characteristics and the cross-branch line transmission characteristics to obtain an initial distribution network fault location model, specifically for: Constructing a transmission feature vector based on the single-branch line transmission characteristics and the inter-branch line transmission characteristics; Determine the number of network layers and module structure based on the residual network and the transmission feature vector to obtain a distribution network fault location network structure; Based on the distribution network fault location network structure, a fault branch identification module is constructed and obtained; Based on the distribution network fault location network structure, a fault distance measurement module is constructed and obtained; Based on the fault branch identification module and the fault distance measurement module, an initial distribution network fault location model is determined.

9. 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, the single-end positioning method for the distribution network based on the global traveling wave time-frequency diagram as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the single-end positioning method for a distribution network based on a global traveling wave time-frequency diagram as described in any one of claims 1 to 5 is implemented.

Images