Power distribution network fault voltage sag positioning method

Abstract

The invention provides a power distribution network fault voltage sag positioning method, which comprises the steps of constructing a power distribution network model based on a topological structure of a power distribution network; determining a characteristic matrix and an electrical connection matrix based on the power distribution network model; inputting the characteristic matrix and the electrical connection matrix into a voltage sag source section positioning model, and outputting and determining the section level position of the fault; and determining a fault occurrence position based on the section level position of the fault. According to the method, on the premise that an additional monitoring device does not need to be additionally arranged, time-space characteristic information of voltage and current can be effectively mined, high-precision section positioning and accurate positioning of the sag source of the power distribution network are achieved while the light weight of the algorithm is kept, the efficiency and accuracy of fault positioning are remarkably improved, and the positioning precision exceeds 99%.

Description

Technical Field

[0001] The present application relates to the technical field of distribution networks, and in particular to a method for locating a fault voltage sag in a distribution network. Background Art

[0002] Voltage sag is one of the most significant power quality issues in power systems, often caused by short-circuit faults, and accounts for approximately 80% of all power quality issues. This problem causes significant economic losses to power systems annually, particularly for industrial users, and has become a major constraint to their normal operation. When a short-circuit fault occurs in the power grid, reclosing can trigger multi-level voltage sags at multiple nodes, impacting sensitive equipment multiple times and potentially causing system failure, resulting in even more serious economic consequences. Compared to non-fault-related voltage sag sources, which are typically located at different points from the equipment installation point, fault-induced voltage sags are sudden and uncertain. Therefore, research on their source location typically focuses on identifying short-circuit fault sources. Effectively locating the source of voltage sags not only helps clarify responsibility and impact assessment between power suppliers and users, but is also crucial for ensuring the safe and stable operation of the power grid.

[0003] Under the new circumstances, with the continuous advancement of high-quality development of distribution networks, power quality issues are increasingly receiving widespread attention. Grid companies are continuously increasing their investment in power quality monitoring systems (PQMS), and various monitoring devices are widely deployed in all aspects of the power system to achieve systematic monitoring and in-depth analysis of power disturbance data, making "digitalization" gradually become an important feature of modern power grids. Existing technologies can only vaguely locate the source of voltage sags at the segment level and cannot accurately locate the specific fault point. This leads to a vague scope during troubleshooting and difficulty in quickly identifying the problem. This results in long troubleshooting time and a wide range, and the source of the fault cannot be quickly and effectively determined, affecting the timeliness and efficiency of problem handling. As a result, the voltage sag problem cannot be solved in a timely manner, seriously affecting the stability of the power grid and the reliability of user electricity use, and hindering the development of the power grid in the direction of intelligence and high reliability.

[0004] Therefore, how to improve the efficiency and accuracy of locating voltage sag sources is an urgent problem to be solved. Summary of the Invention

[0005] In order to overcome the above technical defects, the present application provides a method for locating a fault voltage sag in a distribution network. To achieve the above purpose, the present application is implemented according to the following technical solutions:

[0006] This application provides a method for locating a distribution network fault voltage sag, comprising:

[0007] Based on the topological structure of the distribution network, a distribution network model is constructed;

[0008] Based on the distribution network model, determining a characteristic matrix and an electrical connection matrix;

[0009] Inputting the characteristic matrix and the electrical connection matrix into a voltage sag source section location model, and outputting a section-level location of the fault;

[0010] The fault location is determined based on the section-level location of the fault.

[0011] Optionally, determining a characteristic matrix based on the distribution network model includes:

[0012] Based on the distribution network model, determining the effective value of the node voltage and the effective value of the node current;

[0013] A characteristic matrix is determined based on the node voltage effective value and the node current effective value.

[0014] Optionally, the voltage sag source section location model includes a normalization module, a spatiotemporal graph convolution module, a section location module and an output module connected in sequence.

[0015] Optionally, inputting the characteristic matrix and the electrical connection matrix into a voltage sag source section location model to output a determined section-level location of the fault includes:

[0016] Inputting the characteristic matrix into the normalization module to obtain a sag fault characteristic;

[0017] Inputting the sag fault feature and the electrical connection matrix into the spatiotemporal graph convolution module to obtain a first output matrix;

[0018] Inputting the first output matrix into the section positioning module to determine the sag source section where the fault is located;

[0019] The temporary drop source section where the fault is located is input into the output module to determine the section-level position where the fault is located.

[0020] Optionally, determining the fault location based on the fault section-level location includes:

[0021] Use power quality monitoring devices to determine the three-phase voltage during voltage sag to identify the fault type;

[0022] Based on the identifiable fault type and the fault section-level location, a fault occurrence location is determined.

[0023] Optionally, determining the fault location based on the identifiable fault type and the fault section-level location includes:

[0024] Step S601: Based on the identifiable fault type, a corresponding first sag fault is set at a first midpoint position of the fault section level position;

[0025] Step S602: determining, based on the first sag fault and the first node at the fault section level, a sum of first amplitudes of drops in effective voltage values of each phase voltage at the first node during a simulated fault;

[0026] Step S603: determining, based on the first sag fault and the second node at the fault section level, a sum of second amplitudes of drops in effective values of voltages of respective phases at the second node during a simulated fault;

[0027] Step S604: obtaining a sum of third amplitudes of voltage drops of each phase effective value when the first node is truly faulty;

[0028] Step S605: obtaining a sum of fourth amplitudes of drops in effective values of voltage sources of each phase when the second node is truly faulty;

[0029] Step S606: determining a first amplitude difference based on the sum of the first amplitudes and the sum of the third amplitudes;

[0030] Step S607: determining a second amplitude difference based on the sum of the second amplitudes and the sum of the fourth amplitudes;

[0031] Step S608: Determine the fault location based on the first amplitude difference and the second amplitude difference.

[0032] Optionally, determining the fault location based on the first amplitude difference and the second amplitude difference includes:

[0033] determining whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero;

[0034] If so, determining whether the first distance value between the first node and the first midpoint position meets a preset accuracy threshold;

[0035] If so, the interval range corresponding to the first distance value is determined as the fault location.

[0036] Optionally, when the determination result of determining whether the first distance between the first node and the first midpoint satisfies a preset accuracy threshold is no, the method includes:

[0037] The first midpoint position is updated to the second node, and the position between the updated second node and the first node is updated to the section-level position where the fault is located, and steps S601-S608 are repeated.

[0038] Optionally, when the determination result of determining whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero is no, the method includes:

[0039] Determining whether a second distance value between the second node and the first midpoint satisfies a preset accuracy threshold;

[0040] If so, the interval range corresponding to the second distance value is determined as the fault location.

[0041] Optionally, when the determination result of determining whether the second distance between the second node and the first midpoint satisfies a preset accuracy threshold is no, the method includes:

[0042] The first midpoint position is updated to the first node, and the position between the updated first node and the second node is updated to the section-level position where the fault is located, and steps S601-S608 are repeated.

[0043] This application has the following beneficial effects:

[0044] The method proposed in this application can effectively mine the temporal and spatial characteristic information of voltage and current without adding additional monitoring devices. While keeping the algorithm lightweight, it can achieve high-precision section positioning and precise positioning of the temporary sag source in the distribution network, significantly improving the efficiency and accuracy of fault location, with a positioning accuracy of over 99%.

[0045] In addition to the above-described purposes, features and advantages, the present application has other purposes, features and advantages. The present application will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0047] Figure 1 This is a flow chart of a method for locating a voltage sag in a distribution network fault provided by an embodiment of the present application;

[0048] Figure 2 Schematic diagram of the spatiotemporal correlation of the operating status of the distribution network provided by an embodiment of the present application;

[0049] Figure 3 Schematic diagram of the processing process of the first output feature in the segment positioning module provided in an embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of a search for accurate fault location provided by an embodiment of the present application;

[0051] Figure 5 This is a schematic diagram of a secondary search for accurate fault location provided by an embodiment of the present application;

[0052] Figure 6 This is a schematic diagram of the node distribution network model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The embodiments of the present application are described in detail below with reference to the accompanying drawings, but the present application can be implemented in many different ways as defined and covered by the claims.

[0054] In order to solve the problems raised by the background technology, such as Figure 1 As shown, the present application provides a method for locating a distribution network fault voltage sag, comprising:

[0055] Step S101: constructing a distribution network model based on the topological structure of the distribution network;

[0056] To effectively model the temporal dynamics and spatial dependencies of voltage and current, the distribution network topology is represented as a topological graph by treating power sources and loads as nodes and the power lines connecting the nodes as edges. To fully utilize spatial information, a generalized graph is used to model the distribution network, rather than treating individual node voltages and branch currents as independent discrete components.

[0057] Since the operating state of each node is not only affected by itself in the current time segment, but also has a complex spatiotemporal coupling relationship with the operating state of its adjacent nodes in the adjacent time segments. The schematic diagram of the spatiotemporal correlation of the distribution network operating state is as follows: Figure 2 shown.

[0058] Step S102: determining a characteristic matrix and an electrical connection matrix based on the distribution network model;

[0059] When using algorithms to identify the source of voltage sags, while selecting more fault features can help improve location accuracy, uploading large amounts of fault data from edge power quality monitoring devices to the grid company's main station consumes significant communication and storage resources, a problem often overlooked in existing research. Therefore, to deploy voltage sag source location algorithms within the business application systems of grid company main stations and achieve accurate and lightweight location, emphasis should be placed on streamlining feature selection, prioritizing a small number of representative features to minimize resource consumption while ensuring location accuracy.

[0060] From a temporal perspective, based on the characteristics of the PQDIF (Power Quality Data Interchange Format) standard and the fact that the change in node voltage when a voltage sag occurs is the most intuitive variable, the node voltage RMS value was selected as one of the key features. When a voltage sag occurs due to a fault, the current at the fault point will suddenly increase, so the node current RMS value was selected as the second key feature. Therefore, the voltage RMS value and current RMS value of each node were selected to construct a feature matrix. At time t, the feature matrix can be represented as a two-dimensional matrix Xt, namely:

[0061]

[0062] Where N represents the number of nodes in the distribution network, V t 1 , I t 1 It represents the average value of the effective value of the three-phase voltage and three-phase current of phases a, b, and c at node 1.

[0063] From a spatial perspective, the distribution network can be viewed as a graph structure. The state of each node in the distribution network is constantly affected by adjacent nodes and more distant nodes. The connection relationship between nodes reflects its spatial characteristics. Therefore, in an N-node distribution network, an N×N electrical connection matrix A can be obtained based on the distribution network topology, where the matrix elements A ij =1 indicates that the power lines between nodes are connected, A ij =0 means no connection.

[0064] Step S103: inputting the characteristic matrix and the electrical connection matrix into a voltage sag source section location model, and outputting a determined section-level location of the fault;

[0065] When training the neural network model, when a voltage sag fault occurs in a certain section, it can be represented as a voltage sag fault between two nodes. The labels of these two nodes are marked as 1, and the labels of the remaining nodes are marked as 0. At this point, the voltage sag source location problem is converted into a graph classification problem based on STGCN, that is, the feature X is represented as t The electrical connection matrix A is input into the trained STGCN voltage sag source section location model, which can output the predicted nodes. The lines connecting the nodes represent the predicted voltage sag source sections.

[0066] This application uses a voltage sag source segment location model to achieve segment-level fault location. The voltage sag source segment location model is an improved STGCN algorithm. This model has been improved on the basis of the original STGCN model, making it more suitable for locating voltage sag sources. This location model consists of four parts: a normalization module, a spatiotemporal graph convolution module, a segment location module, and an output module. The specific content and processing procedures of these four modules are described in detail below:

[0067] The normalization module is mainly used to normalize the input feature matrix to obtain the sag fault feature. The sag fault feature is in the form of a matrix with dimensions (B, N, T, 2), where B refers to the number of samples selected for a training session, that is, the length of each processed data, N refers to the number of distribution network nodes, T refers to the time segment length, and 2 refers to the number of features, including the average value of the effective value of each phase of the three-phase voltage and three-phase current. The above normalization process can be expressed as:

[0068]

[0069] Among them, x i is the input feature X t The elements in x imax and x imin are the maximum and minimum values of the i-th input feature.

[0070] The spatiotemporal graph convolution module consists of two spatiotemporal convolution modules, each of which includes two time-domain convolution layers for extracting temporal features and one spatial-domain convolution layer for extracting spatial features. That is, the time-domain convolution layers and the spatial-domain convolution layers are used alternately. The above-mentioned sag fault features and electrical connection matrix are sequentially input into the time-domain convolution layer and the spatial-domain convolution layer, and then passed through a time activation function before being input into the time-domain convolution layer. This process can be expressed as follows:

[0071]

[0072] Among them, X l and X l+1 and represent the input and output of the fault feature after a spatiotemporal convolution module, Θ t represents the convolution kernel on the spatial domain convolution layer, Γ0 l , Γ1 l Denote the kernels of the two time-domain convolutional layers, σ denotes the activation function, and G denotes a directed graph, which contains two pieces of information: the input sag fault characteristics and the electrical connection matrix.

[0073] In the time dimension, the time domain convolution layer uses the gated linear unit (GLU) as a nonlinear layer to implement one-dimensional causal convolution. This method does not rely on previous outputs, which not only effectively alleviates the gradient vanishing problem but also retains the nonlinear expression ability of the model. The expression of the time domain convolution layer is

[0074]

[0075] Among them, α, β represent the input through GLU, the Hadamard product between the elements of ⊙, M represents the time length of the input fault feature, K t represents the size of the input feature neighborhood, and C represents the size of the output dimension.

[0076] In the spatial dimension, the spatial domain convolution layer adopts the form of graph convolution. First, the electrical connection matrix is Laplace transformed to aggregate the information of the domain. Each node can also update its state according to its adjacent nodes. Finally, a nonlinear transformation is performed based on its own information. The relevant transformation expression is

[0077]

[0078] Where D is the degree matrix, which is a diagonal matrix. σ(·) represents a nonlinear activation function, such as the ReLU function. (l +1) is the current hidden feature layer after mining spatial features through graph convolution, and H (l) is the fault node feature matrix of the previous layer. W is the trainable weight.

[0079] After the sag fault features and the electrical connection matrix pass through two spatiotemporal graph convolution modules, the hidden time and space characteristics are extracted to obtain the first output matrix of (B, N, 1).

[0080] like Figure 3 As shown in the figure, the specific processing process of the first output matrix in the segment localization module is as follows: the first output matrix is input into the segment localization module, first passing through a fully connected layer and a pooling layer to obtain a second output matrix of dimension (B, N). The segment localization module further determines whether there is a temporary drop source in each segment within the time period. When designing the segment localization module, the width of the convolution kernel of this module is set to 3 and the size is set to 1. The time dimension decreases by 2 after each pass through this module, and the number of channels also changes synchronously.

[0081] The second output matrix is input to the ReLU activation function, and it is mapped to the value range of (0,1) to obtain the sag source probability matrix. Finally, the sag source criterion is used to determine whether there is a sag source in each segment. That is, if the sag source probability of each node exceeds the predetermined threshold, the sag source standard is met, and the node is determined to be the beginning or end of the sag source segment, otherwise it is not. The output of the segment positioning module Y = [Y1, Y2, Y3, ... Y N ], where N is the number of nodes in the distribution network. The output elements are both 0 and 1. If the outputs of two nodes are both 1, that is, if the outputs of the nodes at both ends of a line are both 1, then the line is determined to be in the section where the voltage sag source is located. This line is then considered the section where the fault is located, thus achieving the voltage sag source's segment location.

[0082] After determining the sag source section where the fault is located, it is input to the output module, and the output module uses this as the section-level location of the fault.

[0083] Step S104: Determine the fault location based on the section-level location of the fault.

[0084] After successfully locating the voltage sag section in step S103, searching for the voltage sag source from the beginning to the end of the section would require many searches and be slow. Therefore, accurately locating the voltage sag source based on a binary search algorithm from computer science can significantly improve location efficiency. The binary search algorithm can narrow the voltage sag source's location to half of the section after each search. This allows for fast and accurate location, with high precision.

[0085] The specific implementation process is as follows:

[0086] First, a power quality monitoring device is used, and then the fault type can be identified based on the three-phase voltage during the voltage sag period, and then the fault location can be determined based on the identifiable fault type and the fault segment level location. Figure 4 Let's explain the specific determination process:

[0087] First, it is assumed that the fault is located at the section level between the first node a (also understood as the left node) and the second node b (also understood as the right node).

[0088] Step S601: according to the identifiable fault type, a corresponding first sag fault is set at the first midpoint position z0 of the fault section level position.

[0089] A corresponding first sag fault z0 is set at the midpoint between the first node a and the second node b. This can be understood as injecting a corresponding fault current to simulate the occurrence of a voltage sag.

[0090] Step S602: determining, based on the first sag fault and the first node at the fault section level, a sum of first amplitudes of drops in effective voltage values of each phase voltage at the first node during a simulated fault;

[0091] At this time, calculate the sum of the first amplitude of the drop of the effective value of the voltage of each phase when the first node a is simulated fault (first pause fault) a0 , the unit is per unit value.

[0092] Step S603: determining, based on the first sag fault and the second node at the fault section level, a sum of second amplitudes of drops in effective voltage values of each phase voltage at the second node during a simulated fault;

[0093] At this time, calculate the sum E of the second amplitude of the effective value of the voltage of each phase when the second node b is in the simulated fault (first pause fault) b0 , the unit is per unit value.

[0094] Step S604: Obtain the sum of the third amplitudes of the drop in the effective value of the voltage of each phase when the first node is truly faulty;

[0095] At this time, the third sum E of the voltage effective value drops of each node a when the first node a is in a real fault (the current fault) is obtained. a , the unit is also per unit value.

[0096] Step S605: obtaining a sum of fourth amplitudes of drops in effective values of voltage sources of each phase when the second node is truly faulty;

[0097] At this time, the fourth sum E of the voltage effective value drops of the second node b when the second node b is in a real fault (the current fault) is obtained. b , the unit is also per unit value.

[0098] Step S606: determining a first amplitude difference based on the sum of the first amplitudes and the sum of the third amplitudes;

[0099] At this time, the sum of the third amplitude E a Subtract the sum of the first amplitude E a0 , get the first amplitude difference ΔE a0 .

[0100] Step S607: determining a second amplitude difference based on the sum of the second amplitudes and the sum of the fourth amplitudes;

[0101] At this time, the sum of the fourth amplitude E b Subtract the sum of the second amplitude E b0 , and obtain the second amplitude difference ΔE b0 .

[0102] Step S608: Determine the fault location based on the first amplitude difference and the second amplitude difference.

[0103] At this time, it is determined whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero. If so, ΔE a0 >0, and ΔE b0 <0, it indicates that the first sag point z0 is far from the actual sag point and close to the second node b. At this time, the actual fault point is between the first node a and the first sag point z0, that is, closer to the first node a than to the second node b. Because it is farther away from point a, the sag depth during the simulated fault is smaller than during the actual sag. At this point, it is determined whether the first distance between the first node a and the first midpoint meets the preset accuracy threshold. If so, the interval corresponding to the first distance value is determined as the fault location. That is, the fault location is within the interval between the first node a and the first midpoint.

[0104] It should be noted that the preset accuracy threshold is limited according to actual needs and the specific site, and this application does not make any specific limitations here.

[0105] If the first distance value does not meet the preset accuracy threshold, because the first search is generally unlikely to accurately locate the fault, then Figure 4 The first midpoint position z0 in the figure is updated to the second node b, and the position between the updated second node and the first node, that is, the position between the first node a and the original first temporary landing point z0, is updated to the section-level position where the fault is located, as shown in FIG. Figure 5 As shown, the above steps S601-S608 are then repeated to reconfirm the fault location.

[0106] If the judgment result of whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero is no, that is, ΔE a0 <0, while ΔE b0 >0, it indicates that the first sag point z0 is far from the actual sag point and close to the first node a. At this point, the actual fault point is between the second node b and the first sag point z0, that is, closer to the second node b than to the first node a. Because it is farther away from point b, the sag depth during the simulated fault is smaller than during the actual sag. At this point, a determination is made as to whether the second distance between the second node a and the first midpoint meets a preset accuracy threshold. If so, the interval corresponding to the second distance is determined as the fault location. This means that the fault location is within the interval between the second node b and the first midpoint.

[0107] If the second distance value does not meet the preset accuracy threshold, Figure 4The first midpoint position z0 in the diagram is updated to the first node a, and the position between the updated first node and the second node, that is, the position between the second node b and the original first temporary landing point z0, is updated to the section-level position where the fault is located. Then, the above steps S601-S608 are repeated to reconfirm the fault location.

[0108] Experimental simulation stage

[0109] This embodiment takes the improved IEEE 14-node distribution network model as an example and builds a Figure 1 The simulation model of the topology shown in the figure is connected to a distributed power source at a specific node in order to verify the positioning capability of the model in the distributed photovoltaic access scenario. Figure 6 shown.

[0110] The system voltage level is set to 10kV and the frequency is 50Hz. The distributed photovoltaic system adopts an inverter photovoltaic model. The distribution network model has a total of 14 nodes, 13 loads, 13 power lines, and is connected to 3 distributed photovoltaic power sources. The access points are at nodes 5, 8, and 14, with capacities of 300kW, 200kW, and 200kW respectively.

[0111] This distribution network model was built in MATLAB / Simulink for simulation. The present invention takes into account most situations to obtain sag data: different locations on the line, different sag types, different distributed photovoltaic outputs, and simulates multiple sag types for each of the 13 power lines. Among them, the sag types include single-phase grounding, two-phase short circuit, two-phase grounding and three-phase short circuit, a total of 4 common sag types, each of which generated 1000 samples. And considering that the output change of the load fluctuates within the range of ±20%, the data measured by the power quality monitoring device also adds 30dB white noise to simulate the real working conditions.

[0112] When dividing the data, 60% of the data is used as the training data set, 20% of the data is used as the validation data set, and 20% of the data is used as the test data set. The hyperparameters of the model are set as follows: learning rate is 0.001, batch size is 16, and training epochs are 100.

[0113] In the segment positioning, the positioning accuracy is shown in Table 1.

[0114] Table 1 Comparison of model accuracy

[0115]

[0116] As can be seen from Table 1, the method proposed in the present invention can accurately locate the four common types of temporary sags in the distribution network fault location of distributed photovoltaic access: single-phase grounding, two-phase short circuit, two-phase grounding, and three-phase short circuit, and the positioning accuracy exceeds 99%.

[0117] Taking a three-phase short circuit as an example, in the section between node 3 and node 11, the three-phase short circuit position is set to a position 20% away from node 11. According to the precise positioning method proposed in the present invention, node 11 is set as point a, node 3 is set as point b, and the sum of the amplitudes of the effective value drops of each phase voltage is E. a and E b They are 2.82pu and 1.73pu respectively.

[0118] The sag source current is injected into the midpoint position of each round of search in turn. In the first round of search, ΔEa0>0 is obtained, and ΔE b0 <0, which means it is closer to node a. In subsequent searches, binary search method is continuously used to accurately locate the source of the temporary drop, and the judgment criterion is ΔE a0 The positioning error interval is set to 10 -2 , accurately locating the source to within 1% of the search segment length. After seven rounds of searching, the source of the dip was determined to be located within a range of 19.53% to 20.31% from node 3, meeting the expected accuracy. To increase the accuracy requirement, simply continue searching.

[0119] In summary, it can be seen that the method for locating voltage sags in distribution network faults proposed in this application can effectively mine the spatiotemporal characteristic information of voltage and current without adding additional monitoring devices, and achieve high-precision segment positioning and precise positioning of the distribution network sag source while keeping the algorithm lightweight, significantly improving the efficiency and accuracy of fault positioning, with a positioning accuracy of more than 99%. In addition, this application introduces an efficient binary search strategy, narrows the positioning range layer by layer through step-by-step search, and successfully locates the sag source position to a range with an error of less than 1% in the simulation, with good engineering practicality and dynamic adaptability. The three-phase voltage and current effective values used as input features comply with the power quality data standard in the Power Quality Data Interchange Format (PQDIF), making it easy to integrate into the existing cloud power quality monitoring platform without the need for additional modification of the original system architecture. It has the advantages of simple deployment, low resource consumption, and strong economy, and is suitable for large-scale promotion and application of new distribution networks.

[0120] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for locating a voltage sag in a distribution network fault, characterized in that: include: Based on the topological structure of the distribution network, a distribution network model is constructed; Based on the distribution network model, determining a characteristic matrix and an electrical connection matrix; Input the characteristic matrix and the electrical connection matrix into the voltage sag source section location model, and output the determined section-level location of the fault; The fault location is determined based on the section-level location of the fault.

2. The method according to claim 1, characterized in that The determining of the characteristic matrix based on the distribution network model includes: Based on the distribution network model, determining the effective value of the node voltage and the effective value of the node current; A characteristic matrix is determined based on the node voltage effective value and the node current effective value.

3. The method according to claim 1, characterized in that The voltage sag source section location model includes a normalization module, a spatiotemporal graph convolution module, a section location module and an output module which are connected in sequence.

4. The method according to claim 3, characterized in that The inputting the characteristic matrix and the electrical connection matrix into the voltage sag source section location model and outputting the determined section-level location of the fault includes: Inputting the characteristic matrix into the normalization module to obtain a sag fault characteristic; Inputting the sag fault feature and the electrical connection matrix into the spatiotemporal graph convolution module to obtain a first output matrix; Inputting the first output matrix into the section positioning module to determine the sag source section where the fault is located; The temporary drop source section where the fault is located is input into the output module to determine the section-level position where the fault is located.

5. The method according to claim 1, wherein The determining of the fault location based on the fault segment-level location includes: Use power quality monitoring devices to determine the three-phase voltage during voltage sag to identify the fault type; Based on the identifiable fault type and the fault section-level location, a fault occurrence location is determined.

6. The method according to claim 5, characterized in that The determining of the fault location based on the identifiable fault type and the fault segment-level location includes: Step S601: Based on the identifiable fault type, a corresponding first sag fault is set at a first midpoint position of the fault section level position; Step S602: determining, based on the first sag fault and the first node at the fault section level, a sum of first amplitudes of drops in effective voltage values of each phase voltage at the first node during a simulated fault; Step S603: determining, based on the first sag fault and the second node at the fault section level, a sum of second amplitudes of drops in effective values of voltages of respective phases at the second node during a simulated fault; Step S604: obtaining a sum of third amplitudes of voltage drops of each phase effective value when the first node is truly faulty; Step S605: obtaining a sum of fourth amplitudes of drops in effective values of voltage sources of each phase when the second node is truly faulty; Step S606: determining a first amplitude difference based on the sum of the first amplitudes and the sum of the third amplitudes; Step S607: determining a second amplitude difference based on the sum of the second amplitudes and the sum of the fourth amplitudes; Step S608: Determine the fault location based on the first amplitude difference and the second amplitude difference.

7. The method according to claim 6, characterized in that The determining a fault location based on the first amplitude difference and the second amplitude difference includes: determining whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero; If so, determining whether the first distance value between the first node and the first midpoint position meets a preset accuracy threshold; If so, the interval range corresponding to the first distance value is determined as the fault location.

8. The method according to claim 7, characterized in that When the determination result of determining whether the first distance between the first node and the first midpoint satisfies a preset accuracy threshold is no, the method further includes: The first midpoint position is updated to the second node, and the position between the updated second node and the first node is updated to the section-level position where the fault is located, and steps S601-S608 are repeated.

9. The method according to claim 7, characterized in that When the result of determining whether the first amplitude difference is greater than zero and the second amplitude difference is less than zero is no, the method further includes: Determining whether a second distance value between the second node and the first midpoint satisfies a preset accuracy threshold; If so, the interval range corresponding to the second distance value is determined as the fault location.

10. The method according to claim 9, characterized in that When the result of determining whether the second distance between the second node and the first midpoint satisfies the preset accuracy threshold is no, the method further includes: The first midpoint position is updated to the first node, and the position between the updated first node and the second node is updated to the section-level position where the fault is located, and steps S601-S608 are repeated.

Images