Power distribution network cable short-circuit fault locating method considering power supply network topological relationship

Abstract

The application relates to the technical field of short-circuit fault positioning, in particular to a power distribution network cable short-circuit fault positioning method considering power supply network topological relations, which comprises the following steps: taking a protection switch as a topological node and a cable line connected with each protection switch as a topological edge in a power distribution network, and extracting topological structure information representing physical connection relations of the power distribution network; synchronously collecting electrical signals of each topological node in each monitoring period; calculating the adjacent electrical similarity of any two topological nodes, and clustering all the topological nodes according to the adjacent electrical similarity; calculating the electrical abnormality degree of each topological node, screening the clustering clusters and obtaining candidate fault nodes according to the electrical abnormality degree; calculating impedance angle variation, screening fault positioning points from the candidate fault nodes, and evaluating the connection relations and the sign change direction of the impedance angle variation between different fault positioning points, so as to position the short-circuit fault of the cable line. The application improves the accurate positioning of the short-circuit fault in a complex multi-branch network.

Description

Technical Field

[0001] This application relates to the field of short-circuit fault location technology, specifically to a method for locating short-circuit faults in distribution network cables that takes into account the power supply network topology. Background Technology

[0002] With the continuous expansion of urban power distribution networks, complex power supply network structures such as ring networks and multi-source power supply are becoming increasingly common, and power distribution networks exhibit typical high-density, multi-branch topology characteristics. When a short-circuit fault occurs in a power distribution network cable, quickly and accurately locating the fault section is of great significance for narrowing the power outage area and shortening the fault recovery time.

[0003] Existing fault location methods mainly rely on directly detecting voltage or current changes at both ends of the line to locate short-circuit faults. However, in the actual operation of complex distribution networks, the normal switching of large-capacity loads and the random output of distributed power sources can cause significant fluctuations in voltage and current, making it easy to misjudge normal electrical fluctuations as short-circuit faults. Secondly, existing methods lack effective integration of the physical topology of the entire network and in-depth exploration of the spatial collaborative propagation mechanism. They fail to utilize the spatial aggregation characteristics of transient electrical anomalies caused by short-circuit faults propagating along the cable line topology to the upstream and downstream of the fault point for multi-node collaborative discrimination. This makes it difficult to identify fault sections in complex distribution network structures with multiple branches and multiple power sources, easily leading to misjudgments and omissions in fault location. Summary of the Invention

[0004] To address the aforementioned technical issues, a method for locating short-circuit faults in distribution network cables that takes into account the power supply network topology is provided.

[0005] The solution to the technical problem presented in this application is to provide a method for locating short-circuit faults in distribution network cables that takes into account the topology of the power supply network, including the following steps:

[0006] In the distribution network, protection switches are used as topology nodes and cable lines connecting each protection switch are used as topology edges. Topology information that characterizes the physical connection relationship of the distribution network is extracted; electrical signals of each topology node are collected synchronously during each monitoring period.

[0007] For each topology node, time series prediction and comparison of electrical signals for each monitoring period are performed to extract electrical residual information of each topology node; the correlation of electrical residual information between two topology nodes with a connection relationship is evaluated using topology structure information, the adjacent electrical similarity of the two topology nodes is calculated, and all topology nodes are clustered accordingly.

[0008] The average distribution level of electrical residual information and adjacent electrical similarity of other topological nodes that are connected to each topological node are analyzed. The electrical anomaly degree of each topological node is calculated, and clusters are selected and candidate fault nodes are obtained accordingly.

[0009] Based on the electrical signals of each candidate fault node, the impedance angle is calculated and the change in impedance angle between adjacent monitoring periods is extracted. Based on the change in impedance angle, the fault location point is selected from the candidate fault nodes. The connection relationship between different fault location points and the sign change direction of the change in impedance angle are evaluated to locate the short circuit fault in the cable line.

[0010] Preferably, the electrical signal includes a voltage signal and a current signal.

[0011] Preferably, the extraction of electrical residual information for each topology node includes:

[0012] For each topology node, starting from each sampling point of the voltage signal, backtrack one power frequency cycle and calculate the effective value of all voltage amplitudes within that power frequency cycle, which is then used as the effective voltage value of that sampling point; similarly, for the current signal, calculate the effective current value of each sampling point.

[0013] The effective values ​​of voltage and current are collectively referred to as electrical effective values. For each topology node, before each monitoring period, based on the electrical effective values ​​of all sampling points within the preset period, the electrical prediction value of each sampling point within each monitoring period is output using a time prediction model. The electrical residual between the electrical prediction value and the actual electrical effective value of the same sampling point is calculated, wherein the electrical residual includes voltage residual and current residual.

[0014] For each topology node, the voltage residuals and current residuals of all sampling points in each monitoring period are spliced ​​together to form an electrical residual sequence, which serves as electrical residual information.

[0015] Preferably, the calculation process of the adjacent electrical similarity is as follows: for any two topological nodes, if the two topological nodes have a connection relationship in the topological structure information, then the adjacent electrical similarity of the two topological nodes is the correlation degree of the electrical residual sequence between the two topological nodes; otherwise, the adjacent electrical similarity is assigned a value of 0.

[0016] Preferably, the clustering of all topological nodes includes: constructing an N×N dimensional similarity matrix S, where N represents the number of all topological nodes, and the rows and columns of matrix S represent each topological node. For any element in the similarity matrix... , which is the first Row topology node and the first The electrical similarity between adjacent nodes in the column topology is calculated. Based on the similarity matrix, a spectral clustering algorithm is used to cluster all topology nodes to obtain multiple clusters.

[0017] Preferably, the calculation of the electrical anomaly degree of each topology node includes: selecting other topology nodes that are connected to each topology node based on the topology information, defining them as adjacent nodes; calculating the mean of the electrical residual sequence of each adjacent node as the average residual; calculating the product of the adjacent electrical similarity between each topology node and each of its adjacent nodes and the average residual of the adjacent node, and summing the product of each topology node and all its adjacent nodes as the electrical anomaly degree of the topology node.

[0018] Preferably, the step of screening clusters and obtaining candidate fault nodes includes: calculating the mean of the electrical anomaly degree of all topological nodes in each cluster as the fault confidence degree; selecting clusters with fault confidence degrees greater than a preset discrimination threshold, and defining all topological nodes contained in the screened clusters as candidate fault nodes.

[0019] Preferably, the calculation of the impedance angle and extraction of the impedance angle change of the candidate fault node between adjacent monitoring periods includes:

[0020] For each candidate fault node, voltage phasors and current phasors are extracted from the voltage and current signals of each monitoring period, and complex impedance is calculated. The complex impedance is then subjected to a four-quadrant arctangent function operation to extract the impedance angle.

[0021] The difference between the impedance angle of each candidate fault node in each monitoring period and the impedance angle in the previous monitoring period is taken as the impedance angle change.

[0022] Preferably, the step of screening out fault location points includes: selecting candidate fault nodes whose absolute value of impedance angle change is greater than a preset fault threshold and marking them as fault location points.

[0023] Preferably, the step of assessing the connection relationship between different fault location points and the sign of the impedance angle change to locate short-circuit faults in cable lines includes: based on topology information, defining two fault location points with a connection relationship as a pair of adjacent location points; if the signs of the impedance angle changes of the two fault location points in a pair of adjacent location points are opposite, then it is determined that there is a short-circuit fault in the cable line between the pair of adjacent location points; otherwise, no short-circuit fault has occurred in the distribution network.

[0024] This application has at least the following beneficial effects:

[0025] This application extracts electrical residuals through time series prediction, stripping away background interference from normal load fluctuations in the power grid and effectively amplifying the electrical mutation characteristics caused by short-circuit faults. It utilizes topological information to evaluate the correlation of residual information and calculate adjacent electrical similarity only between connected topological nodes. This ensures that the calculation of adjacent electrical similarity follows the actual physical propagation path of the distribution network and avoids false associations caused by accidental data similarity between non-adjacent nodes. Based on this, clustering of all topological nodes automatically merges spatially clustered areas affected by the same fault into one cluster, achieving preliminary separation of the fault-affected domain from the normal operating area and effectively eliminating interference from isolated disturbances in fault area identification. Deeply fusing the average residual, representing the degree of absolute electrical anomaly mutation, with the adjacent electrical similarity, representing spatial state correlation, calculates the electrical anomaly degree. Based on this, clusters are selected and candidate fault nodes are obtained. This allows for strict constraints on high similarity using the actual fault amplitudes of adjacent nodes in the physical topology, effectively eliminating "low residual but high similarity" faults. Interference in areas of stable power grid operation significantly amplifies the distinction between actual fault clusters and normal fluctuation areas, thus reliably and accurately defining the boundaries of the actual fault impact range and avoiding misjudging large branch areas of normal fluctuation as candidate fault nodes. Utilizing the impedance change characteristics caused by short-circuit faults, the impedance angle change is calculated. This physical characteristic is not directly affected by random power output from photovoltaic new energy sources or the switching of ordinary large loads, further enabling the precise elimination of candidate fault nodes that have not undergone impedance qualitative changes and extracting fault location points adjacent to the actual fault location. Furthermore, by analyzing the sign change of impedance angle changes between fault location points with interconnected relationships, short-circuit faults in cable lines are located. This deeply explores the underlying physical laws governing the directional reversal of short-circuit current upstream and downstream of the fault point caused by short-circuit faults. Under the constraints of topological connections, fault location is advanced from "node screening" to "segment locking," avoiding false fault phenomena caused by large appliance switching and random power output from photovoltaic new energy sources in the distribution network, and improving the accuracy of short-circuit fault location in complex multi-branch networks. Attached Figure Description

[0026] The following section provides a more detailed description of the distribution network cable short-circuit fault location method that takes into account the power supply network topology, in conjunction with the accompanying drawings.

[0027] Figure 1 A flowchart illustrating the steps of a distribution network cable short-circuit fault location method considering power supply network topology provided in this application embodiment;

[0028] Figure 2The diagram below illustrates the equivalent circuit of a short circuit in a distribution network cable, as provided in this application embodiment. In this circuit, #1 to #6 represent different ring main units, 101 represents a substation, 102 represents a protection switch on the main feeder, 103 represents a protection switch on the incoming side of ring main unit #1, 104 represents a protection switch on the outgoing side of ring main unit #1, 105 represents a feeder, 106 represents a protection switch between the feeder and the load, 107 represents the load, 108 represents the first tie switch, 109 represents the second tie switch, 110 represents a tie line, 111 represents distributed energy resources, 112 represents a short circuit fault, 113 represents a protection switch on the incoming side of ring main unit #5, and 114 represents a protection switch on the outgoing side of ring main unit #6. Detailed Implementation

[0029] The following, in conjunction with the accompanying drawings and embodiments, provides a more detailed description of the distribution network cable short-circuit fault location method proposed in this application, which takes into account the power supply network topology.

[0030] Please see Figure 1 The diagram illustrates a flowchart of a method for locating short-circuit faults in distribution network cables, taking into account the power supply network topology, according to an embodiment of this application. The method includes the following steps:

[0031] Step 1: In the distribution network, using protection switches as topology nodes and the cable lines connecting each protection switch as topology edges, extract topology information that characterizes the physical connection relationship of the distribution network; and simultaneously collect electrical signals of each topology node during each monitoring period.

[0032] Step 2: For each topology node, perform time series prediction and comparison of electrical signals for each monitoring period, and extract electrical residual information of each topology node; use topology structure information to evaluate the correlation of electrical residual information between two topology nodes with a connection relationship, calculate the adjacent electrical similarity of the two topology nodes, and cluster all topology nodes accordingly.

[0033] Step 3: Analyze the average distribution level of electrical residual information and adjacent electrical similarity of other topological nodes that are connected to each topological node, calculate the electrical anomaly degree of each topological node, and select clusters and obtain candidate fault nodes accordingly.

[0034] Step 4: Based on the electrical signals of each candidate fault node, calculate the impedance angle and extract the impedance angle change of the candidate fault node between adjacent monitoring periods. Based on the impedance angle change, select the fault location point from the candidate fault node, evaluate the connection relationship between different fault location points and the sign change direction of the impedance angle change, and locate the short circuit fault of the cable line.

[0035] The cable topology of power distribution networks is complex and its operation is flexible and varied. When a short-circuit fault occurs in a cable, it can seriously affect power supply safety. Quickly locating and isolating the faulty section is a prerequisite for restoring power to non-faulty areas. Since the abrupt changes in electrical characteristics caused by a short-circuit fault propagate along the distribution network topology, to achieve accurate identification of the faulty section, it is necessary to first obtain the electrical data of each protection switch node and establish a topological adjacency matrix reflecting the actual physical connection relationships. Therefore, the specific process in step 1 is as follows:

[0036] For ease of understanding, the equivalent circuit diagram of a short circuit in a power distribution network cable provided in this embodiment is as follows: Figure 2 As shown, since a distribution network typically includes multiple substations and multiple ring main units, therefore, Figure 2 In this diagram, 101 represents the substation, #1 to #6 represent different ring main units (RNBs), and 102 represents the protection switch on the main feeder. Taking RNB #1 as an example, it has protection switches 103 and 104 inside. Feeder 105 connects RNB #1 to the load end 107, with protection switch 106 in between. The internal network structure of other RNBs is similar to that of RNB #1. In particular, the protection switches on the outgoing side of RNBs #3 and #4, located at the power supply end of substation 101, are used as the first tie switch 108 and the second tie switch 109, respectively, and are connected by tie line 110. When the distribution network cable... When the line is operating normally without faults, the first tie switch 108 and the second tie switch 109 are in the open state, the tie line 110 is interrupted, and the substation 101 operates independently to supply power. However, when there is a short circuit fault 112 in the cable line, the protection switches 113 and 114 in the fault area trip to cut off the fault. In order to ensure the normal power supply to the load end of the subsequent ring main unit #4 and #5, the first tie switch 108 and the second tie switch 109 are closed, the tie line 110 is connected, and the substation 101 supplies power to the ring main unit #4 and #5 through the tie line 110. In addition, 111 represents distributed energy.

[0037] Step 101: In the distribution network, an undirected graph is constructed with protection switches as topology nodes and cable lines connecting each protection switch as topology edges. The topology adjacency matrix of the undirected graph is then extracted and used as topology structure information.

[0038] It should be noted that the construction of undirected graphs and the extraction of topological adjacency matrices are well-known techniques and will not be elaborated here. The topological adjacency matrix is ​​N×N dimensional, where N represents the total number of topological nodes, and the rows and columns of the topological adjacency matrix represent the topological nodes.

[0039] Step 102: Using the intelligent network station terminal (DTU) configured in the ring network cabinet, synchronously collect the electrical signals flowing through each topology node during each monitoring period, wherein the electrical signals include voltage signals and current signals;

[0040] It should be noted that the signal acquisition frequency is 1kHz. Since the data sharing delay between DTUs does not exceed 100ms, in order to ensure that the electrical data acquired by the DTUs can be synchronously shared, the duration of the monitoring period is set to 100ms. As other implementation methods, implementers can set it according to the actual situation.

[0041] During normal operation of the distribution network, the electrical data of each topology node exhibits temporal continuity and stationarity. A time series prediction model can accurately fit the changing trends of this electrical data during normal operation. Furthermore, the fitting residuals are caused only by measurement noise, fluctuating stably around the zero mean and following a Gaussian distribution with a mean of 0. However, once a short-circuit fault occurs, transient electrical changes cannot be predicted using the model, leading to a sharp increase in the residual amplitude. Therefore, the specific process in step 2 is as follows:

[0042] Step 201: For each topology node, backtrack one power frequency cycle from each sampling point of the voltage signal, calculate the effective value of all voltage amplitudes within that power frequency cycle, and use it as the effective voltage value of that sampling point; similarly, for the current signal, calculate the effective current value of each sampling point.

[0043] Step 202: The effective voltage value and the effective current value are collectively referred to as the electrical effective value. For each topology node before each monitoring period, based on the electrical effective values ​​of all sampling points within the preset period, the electrical prediction value of each sampling point within each monitoring period is output using a time prediction model. The electrical residual between the electrical prediction value and the actual electrical effective value of the same sampling point is calculated, wherein the electrical residual includes voltage residual and current residual. For each topology node, the voltage residual and current residual of all sampling points within each monitoring period are concatenated to form an electrical residual sequence, which serves as the electrical residual information.

[0044] The preset duration is set to 10 minutes, but this can be adjusted by the implementer according to actual conditions in other implementation methods. Secondly, the time prediction model uses the ARIMA algorithm, which is a well-known technology and will not be described in detail here. Thirdly, the calculation process for the electrical residuals is as follows:

[0045] Calculate the absolute value of the difference between the predicted electrical value and the actual effective electrical value at the same sampling point, and normalize it to obtain the electrical residual. The normalization process is as follows: the ratio of the absolute value of the difference to the preset static noise level is used as the normalization result. The preset static noise level is equal to the product of the rated value of the electrical signal acquired by the DTU at the topology node and the accuracy level. The accuracy level is set to 0.5%, and the rated value refers to the maximum range value specified by the DTU acquisition device during design. It is set by consulting the device's usage data. In this embodiment, the rated value is set to 100A for current signals and 10kV for voltage signals. In other implementation methods, the implementer can set it according to the actual situation.

[0046] Secondly, the process of obtaining the electrical residual sequence is as follows: the voltage residuals of all sampling points of each topology node in each monitoring period are combined into a voltage residual sequence, the current residuals of all sampling points of each topology node in each monitoring period are combined into a current residual sequence, and the voltage residual sequence and the current residual sequence are spliced ​​together end to end to form the electrical residual sequence.

[0047] Secondly, in actual distribution network operation, an abnormal electrical residual at a single topology node does not necessarily indicate a short-circuit fault. It could also be due to isolated disturbances caused by the switching of large-capacity loads or localized electromagnetic interference. According to the transient characteristics of power systems, a true short-circuit fault will trigger drastic changes in current and voltage, and these fault characteristics will propagate along the cable topology upstream and downstream of the fault point, forming a clustered abnormal electrical disturbance field in physical space. Therefore, by performing cluster analysis on the topology nodes, the influence domain of clustered abnormal faults can be automatically separated from normal areas, narrowing the search range for subsequent precise fault location.

[0048] Step 203, the calculation process for the adjacent electrical similarity of any two topological nodes under each monitoring period is as follows:

[0049] If the corresponding element value of any two topological nodes in the topological adjacency matrix is ​​1, then the adjacency electrical similarity of any two topological nodes is the degree of correlation between the electrical residual sequences of these two topological nodes; conversely, if the corresponding element value in the topological adjacency matrix is ​​0, then the adjacency electrical similarity is assigned a value of 0.

[0050] The correlation between two topological nodes is calculated using the Gaussian radial basis function (RBF) kernel, which is a well-known technique. The specific formula is:

[0051] In the formula, For the first The electrical residual sequence corresponding to each topology node. For the first Electrical residual sequence of topological nodes, The preset bandwidth parameters, Represents Euclidean distance. Represents an exponential function with the natural constant as its base;

[0052] The process of obtaining the preset bandwidth parameter is as follows: calculate the Euclidean distance of the electrical residual sequence between any two topology nodes, and count the median of all Euclidean distances. If the median is within the preset parameter range, the median is directly used as the bandwidth parameter. If the median is less than the lower limit of the preset parameter range, the preset bandwidth parameter is set to the lower limit. If the median is greater than the upper limit of the preset parameter range, the preset bandwidth parameter is set to the upper limit. The preset parameter range is set to 0.1~10. As other implementation methods, implementers can set it according to the actual situation.

[0053] It should be noted that the greater the correlation, the closer the electrical residual sequences of the two topological nodes are in terms of numerical fluctuation characteristics. Since the drastic current and voltage changes caused by short-circuit faults will spread synchronously to adjacent nodes around the fault point, the greater the electrical similarity between adjacent nodes, the more likely that the two topological nodes are affected by the same short-circuit fault source.

[0054] Step 204: Construct a similarity matrix with the same dimension as the topological adjacency matrix, where for any element in the similarity matrix... , which is the first Row topology node and the first Analyze the electrical similarity between adjacent nodes in the column topology; based on the similarity matrix, use a spectral clustering algorithm to cluster all topology nodes and obtain multiple clusters;

[0055] The specific process of the spectral clustering algorithm is as follows: Based on the similarity matrix, construct the Laplacian matrix, where the Laplacian matrix... for , The degree matrix is ​​represented by its diagonal elements, which are the similarity matrix. The sum of all non-zero elements in each row of the matrix is ​​calculated; the Laplacian matrix is ​​decomposed into eigenvalues, and the eigenvectors corresponding to the first k smallest eigenvalues ​​are extracted to form an eigenma matrix. The eigenma matrix is ​​then subjected to traditional clustering (such as K-Means) to obtain the clustering results; the value of k is automatically determined by the Bayesian information criterion. The Bayesian information criterion and the spectral clustering algorithm are well-known techniques and will not be elaborated here.

[0056] Furthermore, while spectral clustering can group adjacent topological nodes with similar electrical residual variation trends into the same cluster, the clustering results only reflect the electrical similarity between topological nodes and cannot directly quantify whether a cluster belongs to the fault category. For topological nodes belonging to the category of true short-circuit faults, not only is it necessary to be highly similar to the surrounding adjacent topological nodes in electrical characteristics, but it is also required that the electrical residual amplitudes of these topological nodes increase synchronously. Therefore, in step 3, based on the electrical residuals and adjacent electrical similarity, the fault confidence is calculated to screen candidate fault nodes, specifically as follows:

[0057] Step 301: For the topological adjacency matrix, select the remaining topological nodes that have connections with each topological node and define them as adjacent nodes; calculate the mean of the electrical residual sequence of each adjacent node as the average residual; for each topological node and each of its adjacent nodes, calculate the product of the adjacent electrical similarity between the two and the average residual of the adjacent node, and sum the product of each topological node and all its adjacent nodes as the electrical anomaly degree of the topological node.

[0058] The existence of a connection means that the corresponding element value of two topological nodes in the topological adjacency matrix is ​​1; the formula for calculating the electrical anomaly degree is:

[0059] In the formula, Indicates the first Electrical anomaly degree of each topology node For the first The average residual of each adjacent node, For the first The topology node and the first Electrical similarity between adjacent nodes For the first Each topology node corresponds to the number of all its adjacent nodes;

[0060] It should be noted that the larger the average residual, the more the electrical data of the adjacent node deviates significantly from the predicted value, reflecting a strong transient electrical mutation in the area, which is very likely to have been directly impacted by a short-circuit fault. The higher the electrical similarity between adjacent nodes, the more consistent the topological node and its adjacent nodes are in terms of the shape and fluctuation trend in the residual space. The greater the electrical anomaly, the more the other topological nodes adjacent to this topological node not only have violent electrical residual mutations, but also that these mutations are highly coordinated with this topological node, reflecting that this topological node is located in a real short-circuit fault disturbance center with spatial clustering characteristics, and its credibility as belonging to the fault influence area is extremely high.

[0061] Step 302: Calculate the mean of the electrical anomaly degree of all topology nodes in each cluster as the fault confidence degree; select clusters with fault confidence degrees greater than the preset discrimination threshold, and define all topology nodes contained in the selected clusters as candidate fault nodes.

[0062] Specifically, the process of obtaining the preset discrimination threshold is as follows: obtain the electrical signals of each topology node when the distribution network is fault-free during historical operation, calculate the fault confidence of each cluster according to the above process, and take the maximum fault confidence of all clusters as the preset discrimination threshold.

[0063] It should be noted that the higher the fault confidence, the higher the overall electrical anomaly level within the cluster. This means that not only are the residual fluctuation trends of each topological node within the cluster highly consistent, but they have also all experienced drastic electrical amplitude mutations. This reflects that the cluster physically covers a core area that is strongly impacted by a short-circuit fault, and eliminates the possibility of single-point false faults. As a whole, it has extremely high reliability as a section affected by a real fault.

[0064] Furthermore, when the distribution network cable is in normal operation, the equivalent complex impedance measured by the protection switch is mainly composed of the line impedance and load impedance of the downstream load. Since load switching is a steady-state or slowly changing process, its equivalent impedance angle changes smoothly, basically fluctuating around the zero mean. However, once a short-circuit fault occurs at a point in the line, the fault point will be considered as an extremely low impedance grounding or phase-to-phase path. This sudden change will completely alter the original power flow distribution of the network, and the equivalent impedance of the line will change. Moreover, the impedance angle changes in opposite directions upstream and downstream of the fault point. Specifically, the protection switch closest upstream of the fault point will have an increased proportion of resistance in its measured equivalent impedance due to the short-circuit path, resulting in a significant decrease in the impedance angle. Conversely, the protection switch closest downstream of the fault point will have a significantly increased impedance angle because the direction of the system short-circuit current reverses as it flows through the protection switch. Therefore, in step 4, the specific process is as follows:

[0065] Step 401: For each candidate fault node, extract the voltage phasor and current phasor from the voltage and current signals of each monitoring period, calculate the complex impedance, perform a four-quadrant arctangent function operation on the complex impedance, and extract the impedance angle.

[0066] Among these methods, the discrete Fourier transform is used to extract the voltage and current phasors. The extraction of voltage and current phasors, as well as the calculation of complex impedance, are well-known techniques and will not be elaborated upon here. The complex impedance is the ratio of the voltage phasor to the current phasor. Secondly, the impedance angle... The calculation process is as follows: In the formula, Indicates extraction of complex impedance The imaginary part, Indicates extraction of complex impedance The real part, Let represent the arctangent function in the four quadrants. The arctangent function in the four quadrants is a well-known technique and will not be described in detail here.

[0067] It should be noted that when the real part of the complex impedance is 0, the impedance angle is directly set to equal to... .

[0068] Step 402: The difference between the impedance angle of each candidate fault node in each monitoring period and the impedance angle in the previous monitoring period is taken as the impedance angle change.

[0069] It should be noted that when the absolute value of the impedance angle change approaches zero, it indicates that the equivalent impedance angle of the candidate fault node remains stable between two adjacent monitoring periods without significant change. This point is either in normal operation or, although selected as a candidate fault node, its impedance angle is not substantially affected by the short-circuit fault. When the impedance angle change is negative and its absolute value is large, it indicates that the impedance angle of the candidate fault node has decreased significantly compared to normal operation. This reflects that the closer the point is to the fault point, the more severely affected it is by the fault, and the more likely it is to be the nearest upstream protection switch of the fault point. When the impedance angle change is positive and its absolute value is large, it indicates that the impedance angle of the candidate fault node has increased significantly compared to normal operation. This reflects that the closer the point is to the fault point, the more likely it is to be the nearest downstream protection switch of the fault point.

[0070] Step 403: Select candidate fault nodes whose absolute value of impedance angle change is greater than the preset fault threshold and mark them as fault location points;

[0071] The process of obtaining the preset fault threshold is as follows: obtain the electrical signals of each topology node when the distribution network is fault-free during historical operation, calculate the impedance angle change of each topology node according to the above process, and select the maximum value from the absolute values ​​of the impedance angle changes of all topology nodes as the preset fault threshold.

[0072] Step 404: Based on the topological adjacency matrix, two fault location points that are connected are defined as a pair of adjacent location points. If the signs of the impedance angle changes of the two fault location points in a pair of adjacent location points are opposite, it is determined that there is a short circuit fault in the cable line between the pair of adjacent location points; otherwise, no short circuit fault has occurred in the distribution network.

[0073] It should be noted that if the corresponding element value of two fault location points in the topological adjacency matrix is ​​1, it indicates that the two fault location points are connected and belong to a pair of adjacent location points; secondly, if there are fewer than 2 fault location points or no adjacent location point pair, it is determined that no short circuit fault has occurred in the distribution network.

[0074] In this embodiment, combined with Figure 2 The diagram shows the equivalent circuit of a short circuit in a distribution network cable. When a short circuit fault 112 exists in the cable line, the fault location points selected through the above process are protection switches 113 and 114. Since the corresponding element value of protection switches 113 and 114 in the topological adjacency matrix is ​​1, the two form a pair of adjacent location points. Furthermore, the impedance angle change of the upstream protection switch 114 is negative, and the impedance angle change of the downstream protection switch 113 is positive, with opposite signs. Based on this characteristic, the short circuit fault point 112 is accurately located on the cable line between protection switches 113 and 114. After the location is completed, a trip command can be issued to protection switches 113 and 114 to cut off the short circuit fault section. At the same time, the first tie switch 108 and the second tie switch 109 are closed, and the normal power supply to the non-faulted area affected by the fault, namely the load ends of ring main unit #4 and #5, is quickly restored through the tie line 110.

Claims

1. A method for locating short-circuit faults in distribution network cables that takes into account the topology of the power supply network, characterized in that, The method includes the following steps: In the distribution network, protection switches are used as topology nodes and cable lines connecting each protection switch are used as topology edges. Topology information that characterizes the physical connection relationship of the distribution network is extracted; electrical signals of each topology node are collected synchronously during each monitoring period. For each topology node, time series prediction and comparison of electrical signals for each monitoring period are performed to extract electrical residual information of each topology node; the correlation of electrical residual information between two topology nodes with a connection relationship is evaluated using topology structure information, the adjacent electrical similarity of the two topology nodes is calculated, and all topology nodes are clustered accordingly. The average distribution level of electrical residual information and adjacent electrical similarity of other topological nodes that are connected to each topological node are analyzed. The electrical anomaly degree of each topological node is calculated, and clusters are selected and candidate fault nodes are obtained accordingly. Based on the electrical signals of each candidate fault node, the impedance angle is calculated and the change in impedance angle between adjacent monitoring periods is extracted. Based on the change in impedance angle, the fault location point is selected from the candidate fault nodes. The connection relationship between different fault location points and the sign change direction of the change in impedance angle are evaluated to locate the short circuit fault in the cable line.

2. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 1, characterized in that, The electrical signals include voltage signals and current signals.

3. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 2, characterized in that, The extraction of electrical residual information for each topology node includes: For each topology node, starting from each sampling point of the voltage signal, backtrack one power frequency cycle and calculate the effective value of all voltage amplitudes within that power frequency cycle, which is then used as the effective voltage value of that sampling point; similarly, for the current signal, calculate the effective current value of each sampling point. The effective values ​​of voltage and current are collectively referred to as electrical effective values. For each topology node, before each monitoring period, based on the electrical effective values ​​of all sampling points within the preset period, the electrical prediction value of each sampling point within each monitoring period is output using a time prediction model. The electrical residual between the electrical prediction value and the actual electrical effective value of the same sampling point is calculated, wherein the electrical residual includes voltage residual and current residual. For each topology node, the voltage residuals and current residuals of all sampling points in each monitoring period are spliced ​​together to form an electrical residual sequence, which serves as electrical residual information.

4. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 3, characterized in that, The calculation process of the adjacent electrical similarity is as follows: for any two topological nodes, if the two topological nodes have a connection relationship in the topological structure information, then the adjacent electrical similarity of the two topological nodes is the degree of correlation of the electrical residual sequence between the two topological nodes; otherwise, the adjacent electrical similarity is assigned a value of 0.

5. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 1, characterized in that, The clustering of all topological nodes includes: constructing an N×N dimensional similarity matrix S, where N represents the number of all topological nodes, and the rows and columns of matrix S represent each topological node. For any element in the similarity matrix... , which is the first Row topology node and the first The electrical similarity between adjacent nodes in the column topology is calculated. Based on the similarity matrix, a spectral clustering algorithm is used to cluster all topology nodes to obtain multiple clusters.

6. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 4, characterized in that, The calculation of the electrical anomaly degree of each topology node includes: selecting other topology nodes that are connected to each topology node based on the topology information, defining them as adjacent nodes; calculating the mean of the electrical residual sequence of each adjacent node as the average residual; calculating the product of the adjacent electrical similarity between each topology node and each of its adjacent nodes and the average residual of that adjacent node, and summing the product of each topology node and all its adjacent nodes as the electrical anomaly degree of that topology node.

7. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 1, characterized in that, The process of filtering clusters and obtaining candidate fault nodes includes: calculating the mean of the electrical anomaly degree of all topological nodes in each cluster as the fault confidence degree; selecting clusters with fault confidence degrees greater than a preset discrimination threshold, and defining all topological nodes contained in the filtered clusters as candidate fault nodes.

8. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 2, characterized in that, The calculation of the impedance angle and extraction of the impedance angle change of candidate fault nodes between adjacent monitoring periods includes: For each candidate fault node, voltage phasors and current phasors are extracted from the voltage and current signals of each monitoring period, and complex impedance is calculated. The complex impedance is then subjected to a four-quadrant arctangent function operation to extract the impedance angle. The difference between the impedance angle of each candidate fault node in each monitoring period and the impedance angle in the previous monitoring period is taken as the impedance angle change.

9. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 1, characterized in that, The process of selecting fault location points includes: selecting candidate fault nodes whose absolute value of impedance angle change is greater than a preset fault threshold and marking them as fault location points.

10. The method for locating short-circuit faults in distribution network cables considering the power supply network topology as described in claim 1, characterized in that, The method of assessing the connectivity between different fault location points and the sign of the impedance angle change to locate short-circuit faults in cable lines includes: based on topology information, defining two fault location points that are connected as a pair of adjacent location points; if the signs of the impedance angle changes of the two fault location points in a pair of adjacent location points are opposite, then it is determined that there is a short-circuit fault in the cable line between the pair of adjacent location points; otherwise, no short-circuit fault has occurred in the distribution network.

Images