Method and system for locating fault sections in DC distribution networks based on multi-source information fusion

Abstract

The present invention relates to the technical field of distribution network fault location, and in particular to a method and system for locating a fault section in a DC distribution network based on multi-source information fusion. The method comprises analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected faulty component using a normal distribution model to obtain the electrical quantity fault degree; correcting the node state of a relay protection Bayesian network model based on local switch quantity action information and calculating the switch quantity fault degree through reverse reasoning; using the electrical quantity fault degree and the switch quantity fault degree as independent evidence bodies, calculating the credible interval projection distance weight and information entropy weight using an improved Dempster-Schafer evidence theory identification framework that introduces a full set subset, and weightedly fusing all evidence bodies to obtain the fused fault probability of each suspected faulty component; and locating the DC distribution network fault section based on the fused fault probability. The present invention achieves rapid and accurate location of the DC distribution network fault section by fusing high-frequency transient electrical quantity characteristics with switch quantity action information.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network fault location, and in particular to a method and system for locating a fault section in a DC distribution network based on multi-source information fusion. Background Art

[0002] As DC distribution networks develop towards diversified and multi-layered network structures, the number of distributed photovoltaic, energy storage, and flexible loads connected to them is increasing, and the complexity of system faults has increased significantly. When a short-circuit fault occurs in a DC distribution system, the fault current rises rapidly, and the fault mechanism is complex and changeable, posing a serious threat to the safe and stable operation of the system. To accurately locate and isolate the fault, current methods for locating fault sections in DC distribution networks mainly rely on three types of fault information: those based on electrical fault information, those based on switching fault information, and those based on the fusion of multi-source fault information. However, all of these methods have obvious limitations.

[0003] The location methods based on electrical fault information can be divided into two types: single-end and multi-end measurement. For example, the difference between the positive and negative transient voltage derivatives is used to identify the fault. Although this location method has strong anti-interference ability, it has high requirements for sampling synchronization and communication system. In practical applications, it is easy to cause location failure due to synchronization errors. The method based on switching fault information (such as relay protection action signals) realizes location by building fault diagnosis models such as expert systems or Bayesian networks. However, when Supervisory Control and Data Acquisition (SCADA) is used, the fault diagnosis model is constructed. When a SCADA (Supervisory Control and Acquisition) system experiences signal omissions or false alarms, the reliability of the switching quantity information drops sharply, resulting in insufficient differentiation between the faulty component and other components, or even completely incorrect positioning. The positioning method based on multi-source fault information fusion combines the advantages of electrical quantity and switching quantity information, but existing research still has problems: the fused information sources are mostly derived from the same source, the differences between information sources are low, and the fusion calculation process is cumbersome. At the same time, the positioning method based on multi-source fault information fusion lacks an effective correction mechanism for erroneous switching quantity information, resulting in a fuzzy fault probability distribution. In addition, the existing information fusion algorithm for power grid fault positioning is prone to problems such as information loss or fusion failure in special scenarios, which affects the accuracy and universality of the positioning results and is difficult to adapt to complex and changing fault scenarios.

[0004] In summary, the existing DC distribution network fault location technology has many shortcomings. These problems seriously restrict the accuracy and reliability of fault location, and it is difficult to meet the increasingly complex DC distribution network fault location needs. Therefore, how to improve the accuracy and reliability of fault section location in DC distribution networks is still an urgent problem to be solved. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for locating a fault section in a DC distribution network based on multi-source information fusion.

[0006] In a first aspect, the present invention provides a method for locating a fault section in a DC distribution network based on multi-source information fusion, the method comprising the following steps:

[0007] According to the high-frequency transient voltage signal generated at the moment of bipolar short-circuit fault in DC distribution network, the high-frequency fault current sparse vector is reconstructed.

[0008] Quantitatively analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude;

[0009] The high-frequency current amplitude estimation value of each suspected faulty component is normalized by cubic method to obtain the electrical quantity fault degree of each suspected faulty component;

[0010] The node states of the relay protection Bayesian network model are modified based on the pre-identified local switch action information, and the switch fault degree of each suspected faulty component is calculated through reverse reasoning.

[0011] The electrical quantity fault degree and the switch quantity fault degree are taken as independent evidence bodies, and the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory are respectively calculated using the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set;

[0012] The credible interval projection distance weight and information entropy weight of each evidence body are fused by the combined weighting method, and all evidence bodies are weighted and fused using the fused weights to obtain the fused fault probability of each suspected faulty component.

[0013] The fault section of the DC distribution network is located according to the fused fault probability, and the circuit breaker is driven to isolate the fault.

[0014] In a further embodiment, the step of reconstructing a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short circuit fault in the DC distribution network includes:

[0015] At the moment of a bipolar short circuit fault in the DC distribution network, a node high-frequency impedance matrix is ​​constructed based on the node self-impedance data and the mutual impedance data between nodes in the DC distribution network.

[0016] The high-frequency transient voltage signal at the fault measuring point within a preset time window before and after the occurrence of the bipolar short-circuit fault is synchronously collected by the voltage measuring device;

[0017] The faulty busbar node is equivalent to a high-frequency current source, the current of the non-busbar fault node is set to zero, and a busbar fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal;

[0018] The fault line is equivalent to a high-frequency current source of adjacent fault measuring points, the current between non-adjacent fault measuring points is set to zero, and a line fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal;

[0019] Integrating the bus fault sparse node equation and the line fault sparse node equation into an underdetermined sparse node equation group;

[0020] The underdetermined sparse node equations are transformed into a sparse signal reconstruction problem, and a Bayesian compressed sensing reconstruction algorithm is used to iteratively solve the problem to obtain a high-frequency fault current sparse vector.

[0021] In a further embodiment, the bus fault sparse node equation is used to describe the relationship between the voltage and current of each DC distribution network node when a bus fault occurs;

[0022] The line fault sparse node equation is used to describe the relationship between the voltage and current of each DC distribution network node when a line fault occurs.

[0023] In a further embodiment, the step of using a normal distribution model to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector to obtain an estimated value of the high-frequency current amplitude includes:

[0024] According to the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector, multiple groups of independent normal distribution models are constructed according to the adjacent relationship of the suspected fault components;

[0025] The maximum value of the high-frequency current amplitude in adjacent suspected fault components is taken as the peak value of the corresponding normal distribution model, and the attenuation parameter of each group of normal distribution models is solved by the maximum likelihood estimation criterion;

[0026] Based on the attenuation parameter, a normal distribution model is used to calculate an estimated value of the high-frequency current amplitude at a line midpoint between each pair of adjacent suspected faulty components.

[0027] In a further embodiment, the attenuation parameters include a standard deviation for controlling the attenuation rate and an electrical distance difference between two adjacent suspected faulty components under normal distribution.

[0028] In a further embodiment, the step of performing cubic normalization on the high-frequency current amplitude estimate of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component includes:

[0029] Performing a cubic operation on the estimated value of the high-frequency current amplitude of each suspected faulty component to obtain the cubic amplitude of the high-frequency current of each suspected faulty component;

[0030] The cubic amplitude of the high-frequency current is normalized to obtain the electrical quantity fault degree of each suspected fault component.

[0031] In a further embodiment, the step of correcting the node state of the relay protection Bayesian network model according to the pre-identified local switch quantity action information and calculating the switch quantity fault degree of each suspected faulty component by reverse reasoning includes:

[0032] According to the sequential action logic of the DC distribution network relay protection, a three-level sampling time window is set. The protection level of the tripped circuit breaker is determined by detecting the high-frequency transient voltage signal within each sampling time window, and the local circuit breaker action status of the protection level where the tripped circuit breaker is located is obtained.

[0033] Identifying local switch action information according to the local circuit breaker action state, and using the local switch action information to correct the switch information uploaded by the data acquisition and monitoring system to obtain switch action correction information;

[0034] According to the relay protection action logic of the DC distribution network, a relay protection Bayesian network model of the DC distribution network is constructed. The corresponding relay protection node states and circuit breaker node states in the relay protection Bayesian network model are updated using the switch action correction information to obtain a relay protection Bayesian network correction model.

[0035] Based on the suspected fault components determined from the perspective of electrical quantities, the relay protection Bayesian network modified model is used to perform reverse probability reasoning to calculate the posterior probability of fault of each suspected fault component;

[0036] The fault posterior probability of each suspected faulty component is cubically normalized to obtain the switching fault degree of each suspected faulty component.

[0037] In a further implementation scheme, the three-level sampling time window includes a first-level sampling time window, a second-level sampling time window and a third-level sampling time window; wherein, the first-level sampling time window corresponds to the main protection action period, the second-level sampling time window corresponds to the near backup protection action period, and the third-level sampling time window corresponds to the far backup protection action period.

[0038] In a further embodiment, the steps of respectively calculating the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory by using the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set include:

[0039] Defining the credible interval and maximum uncertainty interval of each suspected fault component in the evidence body according to the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set, and expressing the credible interval and the maximum uncertainty interval in the form of interval numbers;

[0040] Based on the credible interval and maximum uncertainty interval of each suspected faulty component expressed in the form of interval numbers, calculate the projected distance between the credible interval and the maximum uncertainty interval of each suspected faulty component in the evidence body;

[0041] Measuring the uncertainty of the evidence body using the projection distance as an indicator to obtain an uncertainty measurement value of the evidence body;

[0042] According to the uncertainty measurement value of the evidence body and the preset evidence adjustment coefficient, the credible interval projection distance weight of the evidence body from the perspective of evidence theory is obtained;

[0043] The probability of the subset of the entire evidence set is evenly assigned to all suspected faulty components to obtain the component probability distribution and calculate the information entropy of the component probability distribution;

[0044] According to the information entropy of the component probability distribution and the preset probability adjustment coefficient, the information entropy weight of the evidence body from the perspective of probability theory is calculated.

[0045] In a second aspect, the present invention provides a DC distribution network fault section location system based on multi-source information fusion, the system comprising:

[0046] A signal reconstruction module is used to reconstruct a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network;

[0047] an amplitude estimation module, configured to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude;

[0048] The electrical quantity analysis module is used to perform cubic normalization on the high-frequency current amplitude estimation value of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component;

[0049] The switch quantity analysis module is used to modify the node status of the relay protection Bayesian network model based on the pre-identified local switch quantity action information, and calculate the switch quantity fault degree of each suspected fault component through reverse reasoning;

[0050] a dual-weight analysis module, for treating the electrical quantity fault degree and the switch quantity fault degree as independent bodies of evidence, and using an improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set to respectively calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory;

[0051] The probability fusion module is used to fuse the credible interval projection distance weight and information entropy weight of each evidence body through the combined weighting method, and use the fused weights to weight all evidence bodies to obtain the fused fault probability of each suspected faulty component;

[0052] The fault location module is used to locate the fault section of the DC distribution network according to the fused fault probability and drive the circuit breaker to isolate the fault.

[0053] The present invention provides a method and system for locating a fault section in a DC distribution network based on multi-source information fusion. The method reconstructs a high-frequency fault current sparse vector based on a high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network; uses a normal distribution model to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector to obtain a high-frequency current amplitude estimation; cubically normalizes the high-frequency current amplitude estimation of each suspected fault component to obtain the electrical quantity fault degree of each suspected fault component; and corrects the relay protection Bayesian network model based on pre-identified local switch quantity action information. Based on the node status of the type, the switch fault degree of each suspected faulty component is calculated through reverse reasoning. The electrical fault degree and switch fault degree are used as independent evidence bodies. The improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set is used to calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory. The credible interval projection distance weight and information entropy weight of each evidence body are fused through a combined weighting method, and all evidence bodies are weighted and fused using the fused weights to obtain the fused fault probability of each suspected faulty component. The faulty section of the DC distribution network is located based on the fused fault probability, and the circuit breaker is actuated to isolate the fault. Compared with existing technologies, this method achieves rapid and accurate positioning of the DC distribution network fault section by fusing high-frequency transient electrical quantity characteristics with switch action information and combining it with improved evidence theory. It also effectively improves the fault tolerance capability of fault location and ensures the safe and stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 1 is a flow chart of a method for locating a fault section in a DC distribution network based on multi-source information fusion according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of a multi-terminal flexible DC distribution network architecture provided by an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the timing logic of the DC distribution network relay protection provided by an embodiment of the present invention;

[0057] Figure 4 This is an equivalent schematic diagram of a high-frequency current source in the event of a bus fault provided by an embodiment of the present invention;

[0058] Figure 5 This is an equivalent schematic diagram of a high-frequency current source when a line fault occurs, provided by an embodiment of the present invention;

[0059] Figure 6 Schematic diagram of the nonlinear attenuation characteristics of the high-frequency fault current amplitude provided by an embodiment of the present invention;

[0060] Figure 7 Schematic diagram of a normal distribution model with different peak values ​​provided by an embodiment of the present invention;

[0061] Figure 8 Schematic diagram of a Bayesian network model for relay protection of a DC distribution network provided by an embodiment of the present invention;

[0062] Figure 9 This is a schematic diagram of transient voltage change trends at a fault location provided by an embodiment of the present invention;

[0063] Figure 10 1 is a schematic diagram of a fault voltage spectrum analysis at a circuit breaker operation point provided by an embodiment of the present invention;

[0064] Figure 11 1 is a schematic diagram of high-frequency voltage signal detection results provided by an embodiment of the present invention;

[0065] Figure 12 Schematic diagram of a three-level sampling time window based on relay protection sequential logic provided by an embodiment of the present invention;

[0066] Figure 13 This is a schematic diagram of the overall process of switch fault degree assessment provided by an embodiment of the present invention;

[0067] Figure 14 This is a schematic diagram of a fault location process based on an improved Dempster-Schafer evidence theory identification framework provided by an embodiment of the present invention;

[0068] Figure 15 This is a three-dimensional schematic diagram of a high-frequency fault current sparse vector reconstructed using a Bayesian compressed sensing reconstruction algorithm provided by an embodiment of the present invention;

[0069] Figure 16 This is a block diagram of a DC distribution network fault section location system based on multi-source information fusion provided by an embodiment of the present invention.

[0070] Explanation of the accompanying symbols: 101, signal reconstruction module; 102, amplitude estimation module; 103, electrical quantity analysis module; 104, switch quantity analysis module; 105, dual-weight analysis module; 106, probability fusion module; 107, fault location module. DETAILED DESCRIPTION

[0071] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and are not to be construed as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of protection of the present invention. Many changes may be made to the present invention without departing from the spirit and scope of the present invention.

[0072] Figure 1 The embodiment of the present invention provides a method for locating a fault section of a DC distribution network based on multi-source information fusion. Figure 1 As shown, the method includes the following steps:

[0073] S1. Reconstruct a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network.

[0074] In some embodiments, the step of reconstructing a high-frequency fault current sparse vector based on a high-frequency transient voltage signal generated at the moment a bipolar short circuit fault occurs in the DC distribution network includes:

[0075] At the moment of a bipolar short circuit fault in the DC distribution network, a node high-frequency impedance matrix is ​​constructed based on the node self-impedance data and the mutual impedance data between nodes in the DC distribution network.

[0076] The high-frequency transient voltage signal at the fault measuring point within a preset time window before and after the occurrence of the bipolar short-circuit fault is synchronously collected by the voltage measuring device;

[0077] The faulty busbar node is equivalent to a high-frequency current source, the current of non-busbar fault nodes is set to zero, and the busbar fault sparse node equation is established based on the node high-frequency impedance matrix and high-frequency transient voltage signal;

[0078] The fault line is equivalent to a high-frequency current source adjacent to the fault measuring point, the current between non-adjacent fault measuring points is set to zero, and a line fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal;

[0079] The bus fault sparse node equations and line fault sparse node equations are integrated into an underdetermined sparse node equation system;

[0080] The underdetermined sparse node equations are transformed into a sparse signal reconstruction problem, and the Bayesian compressed sensing reconstruction algorithm is used to iteratively solve the problem to obtain the high-frequency fault current sparse vector.

[0081] Specifically, this embodiment is based on The multi-terminal flexible DC distribution network architecture is taken as the research object. Figure 2Figure 1 is a schematic diagram of the architecture of a multi-terminal flexible DC distribution network. The multi-terminal flexible DC distribution network uses modular multilevel converters (MMCs) to achieve flexible interconnection between AC and DC power. PV power sources and DC loads are connected to the distribution network through DC-DC converters (DC-DC) or direct current to alternating current (DC-AC) inverters. To ensure the safe operation of the system, each line and bus port is equipped with a DC circuit breaker (DCCB). The DC circuit breaker is used to quickly isolate the faulty line or bus in the event of a fault. PV power sources achieve maximum power output through a maximum power point tracking (MPPT) control strategy. The DC distribution network protection configuration can quickly isolate faulty components from the power system to prevent further damage to the distribution network. Faulty components include busbars and lines. The protection configuration of each protection area is shown in Table 1. Table 1 is as follows:

[0082] Table 1

[0083]

[0084] This embodiment focuses on the protection analysis of lines and buses in a DC distribution network. Figure 2 Middle Line Take the two-end protection as an example to illustrate its action logic. Figure 2 The multi-terminal flexible DC distribution network architecture shown in the figure includes busbars B1 to B17. Up, left protection Lm and right protection Rm is the main protection, responsible for protecting only its own line; the left nearby is the backup protection Lp and right near backup protection Rp is the near backup protection. When the main protection fails to operate for some reason, the near backup protection will start and drive the DC circuit breaker. and DC circuit breakers Trip to isolate the fault, and left remote backup protection Ls and right far back protection Rs is the remote backup protection. When both the main protection and the local backup protection fail to operate, the remote backup protection will operate, usually driving the DC circuit breaker on the adjacent line. and DC circuit breakers Tripping realizes complete isolation of the fault line. In this embodiment, m represents the main protection, p represents the near backup protection, s represents the far backup protection, L represents the line left end outlet protection, and R represents the line right end outlet protection.

[0085] The busbar protection is also equipped with main protection and backup protection. When the busbar main protection (such as the left main protection B3Lm of busbar B3 and the right main protection B3Rm of busbar B3) is activated, the circuit breaker will be driven. and circuit breakers If the main protection fails to operate, the busbar backup protection (the left backup protection B3Lp of busbar B3 and the right backup protection B3Rp of busbar B3) will operate, driving the DC circuit breaker. and DC circuit breakers Trip to isolate the faulty busbar, Figure 3 This is a schematic diagram of the timing logic of the DC distribution network relay protection provided by the embodiment of the present invention. The action timing of each protection and circuit breaker follows Figure 3 The DC distribution network relay protection timing logic is shown in Figure 1, where: is the initial moment of the fault, that is, the moment when the fault occurs; 、 and The action time of the main protection, the local backup protection and the remote backup protection are respectively: the action time of the main protection is the moment when the main protection detects a fault and starts tripping; the action time of the local backup protection is the moment when the local backup protection starts tripping when the main protection refuses to operate; the action time of the remote backup protection is the moment when the remote backup protection starts tripping when both the main protection and the local backup protection refuse to operate; 、 and The operating moments of the circuit breakers driven by the main protection, the local backup protection, and the remote backup protection, respectively. The operating moment of the circuit breaker driven by the main protection is the moment when the circuit breaker trips in response to the main protection signal, the operating moment of the circuit breaker driven by the local backup protection is the moment when the circuit breaker trips in response to the local backup protection signal, and the operating moment of the circuit breaker driven by the remote backup protection is the moment when the circuit breaker trips in response to the remote backup protection signal. These usually involve circuit breakers on adjacent lines. 、 and The setting action time limits of the main protection, local backup protection and remote backup protection compared to the fault moment are respectively. Among them, the setting action time limit of the main protection compared to the fault moment is the time required for the main protection to detect the fault and initiate tripping; 、 and They are the delay of circuit breaker action after each protection action.

[0086] When a bipolar short-circuit fault occurs in a DC distribution network, the voltage at the fault point suddenly changes, exhibiting a step characteristic. The fault transient signal contains full-frequency domain information, especially in the high-frequency domain. The voltage transient waveform at the fault point contains high-frequency components. The impedance characteristics of the converter equipment in the high-frequency domain are linear and are not affected by the converter control strategy. Therefore, the fault point can be regarded as an additional high-frequency voltage source, and its high-frequency current flows to the entire DC distribution network. The mathematical representation of the node high-frequency impedance matrix is:

[0087]

[0088] Where, is the node high-frequency impedance matrix, the dimension of the node high-frequency impedance matrix is ​​N×N, which represents the impedance characteristics between each node at the angular frequency w; is the node high-frequency admittance matrix, which has the dimension of N×N and is the inverse matrix of the node high-frequency impedance matrix. The superscript -1 is the matrix inversion operator. w is the angular frequency, which represents the frequency characteristics of the high-frequency signal. N is the number of nodes in the DC distribution network. is the element of the node high frequency impedance matrix at frequency w. When i=j, represents the self-impedance of node i; when hour, Represents the mutual impedance between nodes i and j; subscript N1 represents an N-dimensional column vector; subscript NN represents an N-order square matrix.

[0089] Combined with the node high-frequency impedance matrix, this embodiment can further derive the node high-frequency voltage equation. The mathematical expression of the node high-frequency voltage equation is:

[0090]

[0091] Where, is the node high-frequency voltage column vector, the dimension of the node high-frequency voltage column vector is N×1, which contains the voltage phasor of each node at frequency w; is the node high-frequency current column vector, the dimension of the node high-frequency current column vector is N×1, which contains the current phasor of each node at frequency w.

[0092] Short-circuit faults can be divided into bus faults and line faults according to the fault location. When a fault occurs at node i in the DC distribution network, the fault current at the node i can be equivalent to a high-frequency current source. Figure 4 : This is an equivalent schematic diagram of a high-frequency current source during a bus fault according to an embodiment of the present invention. In this case, the high-frequency voltage of the entire network fault is generated only by the current source at node i. In the injected current column vector, only the high-frequency injected current at node i is non-zero. Specifically, it is expressed as follows:

[0093]

[0094] In the formula, is the equivalent high-frequency current source of node i at frequency w.

[0095] When a fault occurs on the line between node i and node j, the equivalent high-frequency current source is located between the two nodes and is no longer distributed on the nodes. To simplify the analysis, in this embodiment, the high-frequency current source between the nodes is equivalently transformed to the two adjacent nodes closest to it. Figure 5 is the schematic diagram of the equivalent high-frequency current source during line fault provided by the embodiment of the present invention. At this time, the high-frequency fault voltage of the entire network is generated by node i and node j, and only the high-frequency injection currents of node i and node j in the node injection current column vector are not zero, which is specifically expressed as:

[0096]

[0097] In the formula, is the high-frequency current source equivalently transformed to node i during line fault at frequency w; is the high-frequency current source equivalently transformed to node j during line fault, where the subscripts i and j represent node numbers; in Figure 5 among them, is the fault location coefficient, , which is used to characterize the relative position of the fault point in the line; is the high-frequency current source before equivalent transformation on line ; is the high-frequency equivalent impedance of the converter at node i; is the high-frequency equivalent impedance of the converter at node j; is the grading parameter of the per-unit threshold interval, which is used for breaker tripping judgment; is the amplitude of the high-frequency current after equivalent distribution during line fault; is the high-frequency current source equivalently distributed at node i during fault; is the high-frequency current source equivalently distributed at node j during fault.

[0098] Based on the above formulas, it can be seen that during the bus or line fault in the DC distribution network, the high-frequency fault current shows sparsity. Based on this, in this embodiment, voltage measurement devices are configured at S (S < N) nodes, the high-frequency fault voltage amplitude and the impedance values of the corresponding rows in the node impedance matrix are extracted, and then the sparse node equations during bus and line faults are constructed. The sparse node equation for bus fault is used to describe the relationship between the voltages and currents of each node in the DC distribution network during bus fault, and the sparse node equation for line fault is used to describe the relationship between the voltages and currents of each node in the DC distribution network during line fault. The mathematical representation form of the sparse node equation for bus fault is:

[0099]

[0100] The mathematical expression of the line fault sparse node equation is:

[0101]

[0102] Where, is the column vector of the high-frequency voltage at the measuring point, is an S × 1 vector containing the high-frequency voltage information of S measuring points in the distribution network; The impedance sub-matrix of the measurement node, which represents the high-frequency impedance matrix of the slave node The S-row and N-column submatrix extracted from corresponds to the impedance relationship between the S measurement nodes and all N nodes; S represents the number of voltage measurement nodes.

[0103] According to the relationship between the rank and row of the bus fault sparse node equation and the line fault sparse node equation, since the rank of the bus fault sparse node equation and the line fault sparse node equation is less than or equal to N, the bus fault sparse node equation and the line fault sparse node equation are both underdetermined equations and have infinite solutions. Therefore, this embodiment adopts the Bayesian compressed sensing (BCS) reconstruction algorithm to solve the bus fault sparse node equation and the line fault sparse node equation to obtain the high-frequency current sparse vector .

[0104] S2. Quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude.

[0105] In some embodiments, the step of quantitatively analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated high-frequency current amplitude includes:

[0106] According to the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector, multiple groups of independent normal distribution models are constructed according to the adjacent relationship of the suspected fault components;

[0107] The maximum value of the high-frequency current amplitude in adjacent suspected fault components is taken as the peak value of the corresponding normal distribution model, and the attenuation parameter of each group of normal distribution models is solved by the maximum likelihood estimation criterion;

[0108] Based on the attenuation parameter, a normal distribution model is used to calculate an estimated value of the high-frequency current amplitude at a line midpoint between each pair of adjacent suspected faulty components.

[0109] Specifically, in the actual operation of the DC distribution network, due to the influence of interference factors such as transition resistance and noise, it is difficult to distinguish between line faults and busbar faults by relying solely on the reconstructed high-frequency current sparse vector. Under ideal conditions, the high-frequency current sparse vector obtained by the reconstruction algorithm only contains the high-frequency current amplitude of the faulty component. However, in the actual operating environment, due to the influence of interference factors such as transition resistance and noise, the accuracy of the reconstruction result is often challenged, resulting in the high-frequency current amplitude of non-faulty components being mixed into the high-frequency current sparse vector. This uncertainty makes it difficult to effectively distinguish between line faults and busbar faults based solely on the reconstruction result, and accurate fault location cannot be achieved. Figure 6 This is a schematic diagram of the nonlinear attenuation characteristics of the high-frequency fault current amplitude. For example, taking the suspected fault bus Bi as an example, the high-frequency current amplitude at the suspected fault bus Bi is The high-frequency current amplitudes at the adjacent buses Bg, Bh, Bj and Bk of the suspected faulty bus Bi are respectively , high-frequency current amplitude , high-frequency current amplitude and high-frequency current amplitude , 、 、 and All less than the peak , although in the reconstruction results, the high-frequency current amplitudes of some non-fault buses are It appears as a local maximum, but this is not enough to distinguish between bus faults and line faults. In addition, the high-frequency current amplitude of the faulty component may also deviate due to interference factors and may even be lower than the amplitude of some non-faulty components.

[0110] Based on the above analysis, although the electrical fault information can preliminarily define the fault range, it cannot accurately locate the fault section. To solve this problem, this embodiment proposes a high-frequency current amplitude estimation method based on the normal distribution model, which converts the reconstructed high-frequency current sparse vector The high-frequency current amplitude of each suspected faulty component is used as the fault probability characterization index from the perspective of electrical quantity to determine the fault degree of electrical quantity, and the information fusion method is used to realize the comprehensive evaluation of multi-source fault degree, so as to accurately locate the faulty component. According to the sparse node equations when the bus and line are faulty, the reconstruction results are Only the high-frequency current amplitude of the suspected fault busbar can be directly characterized, while the high-frequency current amplitude of each suspected fault line presents an implicit characteristic distribution. In order to ensure the completeness of the subsequent multi-source fault information fusion positioning, this embodiment explicitly solves these implicit features to obtain the high-frequency current amplitude of each suspected fault line. However, since the high-frequency current amplitude of each suspected fault element presents a nonlinear attenuation characteristic, it is impossible to directly construct a linear equation to solve the high-frequency current amplitude of each suspected fault line. Therefore, this embodiment introduces the normal distribution model in probability statistics theory, and constructs a normal distribution model to quantify the nonlinear attenuation trend of the high-frequency current amplitude and realize the explicit solution of the high-frequency current amplitude. Considering that if only one normal distribution model is constructed to characterize the high-frequency current amplitude of all suspected fault buses, it is difficult to ensure that all high-frequency current amplitudes are distributed on the normal distribution model, thereby introducing a large error and increasing the uncertainty of fusion positioning, therefore, this embodiment constructs multiple groups of normal distribution models based on the high-frequency current amplitude of the suspected fault busbar in the high-frequency sparse vector. The specific mathematical expression of the constructed normal distribution model is:

[0111]

[0112] Where, is the normal distribution model constructed for the suspected faulty bus Bi; is the standard deviation of the normal distribution, which characterizes the amplitude decay rate; x is a random variable of the normal distribution, which represents the spatial position variable; e is the base of the exponential function.

[0113] Based on the constructed multiple groups of normal distribution models, this embodiment uses the high-frequency current amplitude sequence of the suspected fault bus set as the input, takes the larger value of the high-frequency current amplitude of the two adjacent suspected fault buses as the peak value of the constructed normal distribution model, and solves the attenuation parameters of the normal distribution model based on the maximum likelihood estimation criterion. The attenuation parameters include the standard deviation used to control the attenuation rate. The electrical distance difference between two adjacent suspected fault components under normal distribution , attenuation parameter and The specific calculation formula is:

[0114]

[0115] Where, is the standard deviation of the normal distribution model between busbar Bi and busbar Bj, which is solved by maximum likelihood estimation; is the high-frequency current amplitude at busbar Bi; represents the distance difference between busbar Bi and busbar Bj; is the high-frequency current amplitude at bus Bj.

[0116] This embodiment combines and The combined estimation result can be used to further solve the line high-frequency current amplitude. In the calculation process, this embodiment takes the middle position of the line An estimate is made to obtain the high-frequency current amplitude estimate at the midpoint of the line between each pair of adjacent suspected faulty components. The specific calculation formula for the high-frequency current amplitude estimate is:

[0117]

[0118] Where, For the line The high-frequency current amplitude at the midpoint; It is the equivalent distance between busbar Bi and busbar Bj under normal distribution, which is used to reflect the electrical distance of the busbars.

[0119] By repeating the above steps, this embodiment can achieve a comprehensive estimation of the high-frequency current amplitude of each suspected fault line. The above steps convert the implicit line amplitude distribution into an explicit parameter by quantifying the nonlinear attenuation characteristics of the amplitude. The high-frequency current amplitude of the suspected faulty component is used as a fault probability characterization indicator from the perspective of electrical quantity, and the electrical quantity fault degree is determined. Then, the information fusion method is used to achieve the fusion of multi-source fault degrees and accurately locate the faulty component.

[0120] In order to verify the completeness of the proposed method at the theoretical level, this embodiment further analyzes whether the above calculation process meets the objective law of the normal distribution function. Figure 7 This is a schematic diagram of a normal distribution model under different peak values ​​provided by an embodiment of the present invention. In the constructed normal distribution model, there is a clear quantitative relationship between the parameters. When the high-frequency current amplitude of the bus Bi is When it increases, due to the influence of impedance parameters and interference factors in the distribution network, May appear as increasing or unchanged related characteristics, Showing negative correlation characteristics, Figure 7 The quantitative relationship between the model parameters is specifically manifested as follows: Increase to When the high-frequency current amplitude of the adjacent busbar Bj is Increase simultaneously to ,lead to Reduced to 、 Increase to At the same time, it can be seen from the above attenuation parameter calculation formula that and showed a significant negative correlation. Based on the above analysis, and is positively correlated with The proportional relationship satisfies the parameter constraints of the normal distribution function, and can still effectively solve the amplitude of each suspected fault line when the parameters change, verifying the theoretical completeness of the evaluation method. is the estimated value of the fault high-frequency current amplitude after the bus Bi is increased, which is the adjusted value after considering the interference factor; is the estimated value of the fault high-frequency current amplitude after the bus Bj is increased, which is the adjusted value after considering the interference factor; For the line The estimated value of the fault high-frequency current amplitude after the midpoint is increased, which is the adjusted value after considering the interference factors; For the corrected line Normal distribution model of; is the standard deviation of the normal distribution model between busbar Bi and busbar Bj; For the line The normal distribution model of is used to describe the attenuation characteristics of the amplitude along the line; The corrected spatial coordinates considering interference factors reflect the adjustment amount of the fault location; is the normal distribution distance at the busbar Bi, which is used to locate the relative position of the fault point; is the normal distribution distance at busbar Bj, which is used to locate the relative position of the fault point.

[0121] S3. Perform cubic normalization on the estimated high-frequency current amplitude of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component.

[0122] In some embodiments, the step of performing cubic normalization on the high-frequency current amplitude estimate of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component includes:

[0123] Performing a cubic operation on the estimated value of the high-frequency current amplitude of each suspected faulty component to obtain the cubic amplitude of the high-frequency current of each suspected faulty component;

[0124] The cubic amplitude of the high-frequency current is normalized to obtain the electrical quantity fault degree of each suspected fault component.

[0125] Specifically, to achieve accurate fault location based on multi-source fault information fusion, this embodiment needs to convert two different types of quantities, electrical quantities and switching quantities, into a unified dimension for comprehensive analysis. This embodiment uses the high-frequency current amplitude of the component as a characterization indicator of the electrical quantity fault degree. To ensure the comparability of the electrical quantity fault degrees of different components, this embodiment normalizes the high-frequency current amplitude and other related data of each component to obtain the electrical quantity fault degree corresponding to each suspected faulty component. The calculation formula for the electrical quantity fault degree is:

[0126]

[0127] Where, For the The electrical fault degree of each suspected faulty component; For the The cube amplitude of the high-frequency current of the suspected faulty component, that is, The cube of the high-frequency current amplitude of the suspected faulty component; is the cube amplitude of the high-frequency current of bus Bi; For the line The high-frequency current cube amplitude; n is the total number of suspected fault components, where ; The index of the suspected faulty component.

[0128] Through normalization processing, this embodiment can accurately reflect the relative importance of the electrical quantity fault degree of each component in the fault situation, effectively ensure the probabilistic completeness of the obtained electrical quantity fault degree, and provide an accurate and reliable data basis for the subsequent fusion of multi-source fault information and fault location.

[0129] S4. Modify the node state of the relay protection Bayesian network model based on the pre-identified local switch quantity action information, and calculate the switch quantity fault degree of each suspected fault component through reverse reasoning.

[0130] In some embodiments, the step of correcting the node state of the relay protection Bayesian network model based on the pre-identified local switch value action information and calculating the switch value fault degree of each suspected faulty component by reverse reasoning includes:

[0131] According to the sequential action logic of the DC distribution network relay protection, a three-level sampling time window is set. The protection level of the tripped circuit breaker is determined by detecting the high-frequency transient voltage signal within each sampling time window, and the local circuit breaker action status of the protection level where the tripped circuit breaker is located is obtained.

[0132] Identify local switch action information according to the local circuit breaker action state, and use the local switch action information to correct the switch information uploaded by the data acquisition and monitoring system to obtain switch action correction information;

[0133] According to the relay protection action logic of the DC distribution network, a relay protection Bayesian network model of the DC distribution network is constructed. The corresponding relay protection node states and circuit breaker node states in the relay protection Bayesian network model are updated using the switch action correction information to obtain a relay protection Bayesian network correction model.

[0134] Based on the suspected fault components determined from the perspective of electrical quantities, the relay protection Bayesian network modified model is used to perform reverse probability reasoning to calculate the posterior probability of fault of each suspected fault component;

[0135] The fault posterior probability of each suspected faulty component is cubically normalized to obtain the switching fault degree of each suspected faulty component.

[0136] Specifically, when a component in a DC distribution network fails, the relay protection system will drive the circuit breaker to trip, thereby isolating the fault area and ensuring the safe and stable operation of the power grid. Based on this action logic, this embodiment constructs a Bayesian network model for relay protection of the DC distribution network. Figure 8 Schematic diagram of a Bayesian network model for relay protection of a DC distribution network provided by an embodiment of the present invention. In the Bayesian network model, network nodes are interconnected based on the action logic of the relay protection. The connection order is: element, primary protection, local backup protection, local circuit breaker, remote backup protection, and remote circuit breaker. The refusal to operate state of a network node is represented by 0, and the normal operation state is represented by 1. This embodiment sets the prior probability of faults for the DC line and busbar based on actual DC distribution network engineering conditions. The prior probability of DC line faults is 0.1187, and the prior probability of DC busbar faults is 0.1123.

[0137] The status of network nodes in the relay protection Bayesian network model is usually determined by the switch action information collected by systems such as SCADA. However, when collecting switch information, SCADA and other systems may have omissions and false alarms, which will affect the accuracy of subsequent multi-source fault information fusion and positioning. To address this problem, this embodiment accurately identifies local switch action information and corrects omissions and false alarms, thereby ensuring the accuracy of subsequent fusion information. Specifically, this embodiment will explain the method for obtaining local switch action information from two aspects: circuit breaker action status identification and protection device action status identification:

[0138] This embodiment accurately identifies the high-frequency voltage signal generated at the moment of circuit breaker tripping to achieve precise identification of the circuit breaker's operating status. First, taking the two-stage change trend of the transient voltage at the fault location as the starting point, the generation mechanism of the high-frequency voltage signal (hereinafter referred to as the high-frequency voltage signal) during the circuit breaker tripping process is specifically analyzed. Figure 9 This is a schematic diagram of transient voltage variation trends at a fault location provided by an embodiment of the present invention. For the time failure, Always protect your actions. The moment the circuit breaker trips and isolates the fault, At this moment, the voltage at the fault point changes from the normal operating voltage value Sharply decreases to a voltage value ;exist At this moment, the voltage is It plummeted to 0, and The voltage variation trend at the fault location at all times shows characteristics similar to a step signal, and the special mutation form of the step signal makes it have rich full-frequency domain information. The mathematical expression of the unit step signal is:

[0139]

[0140] Where, is a unit step function, which is used to characterize the voltage mutation characteristics; t is the time variable of the fault process.

[0141] The mathematical expression of the Fourier transform result of the unit step signal is:

[0142]

[0143] Where, is the Fourier transform of the step signal, which reflects the full frequency domain characteristics; is pi; is the Dirac function, which represents the DC component and the impact when the angular frequency w is zero; the j in jw represents the imaginary unit, and the w in jw represents the angular frequency.

[0144] According to the above analysis, we can see that A stable high-frequency voltage signal can be detected at the fault location at all times. Based on this, this embodiment further analyzes Whether the high-frequency voltage signal can be detected at all times, At the moment, the frequency spectrum characteristics of the voltage at the circuit breaker tripping point are as follows: Figure 10 As shown by Figure 10 It can be seen that when the circuit breaker trips, the inter-electrode voltage has a large amplitude change in the low-frequency band, a small spectrum density, strong nonlinearity, and poor stability; while in the high-frequency band, the amplitude change is small, the spectrum density is large, its envelope is approximately linear, and the stability is good. This shows that high-frequency voltage signals can also be detected at the moment of circuit breaker tripping. Figure 11 The embodiment of the present invention provides 0- Schematic diagram of high-frequency voltage signal detection results within the time period.

[0145] In 0- During the period, Moment and High-frequency voltage signals can be detected at all times, but The peak value of the high-frequency voltage signal at the moment is much greater than The reason for the signal peak detected at the moment is that, compared with a bipolar short-circuit fault, the voltage fluctuation caused by the circuit breaker tripping is smaller, so that the amplitude of the high-frequency voltage signal is also smaller. Based on the above analysis, this embodiment combines the timing logic of the relay protection to construct a three-level sampling time window mechanism. The three-level sampling time window mechanism sets sampling time windows during each level of protection, wherein the three-level sampling time windows include a first-level sampling time window, a second-level sampling time window, and a third-level sampling time window; the first-level sampling time window corresponds to the main protection action period, the second-level sampling time window corresponds to the near backup protection action period, and the third-level sampling time window corresponds to the far backup protection action period. This embodiment can determine the protection period in which the tripped circuit breaker is located by detecting the high-frequency voltage signal in different sampling time windows. Figure 12 This is a schematic diagram of a three-level sampling time window based on relay protection timing logic. If a signal is detected only within a single sampling time window 1, sampling time window 2, and sampling time window 3, it means that the circuit breaker tripping only occurs during each level of protection, and no circuit breaker refuses to trip during the fault isolation process. Conversely, if a signal is detected within multiple sampling time windows, it means that the circuit breaker refuses to operate during the fault isolation process (this embodiment does not consider circuit breaker misoperation). It should be noted that since the busbar protection only has main protection and backup protection, when isolating the busbar fault, if the circuit breaker refuses to operate, the signal can be detected in sampling time window 1 and sampling time window 2, but no signal can be detected in sampling time window 3. In this embodiment, the sampling frequency is preferably set between 1 kHz and 1.5 kHz.

[0146] Based on the known protection period of the tripped circuit breaker, this embodiment further determines the line to which the tripped circuit breaker belongs. The high-frequency voltage amplitude signal generated at the moment the circuit breaker trips is small. At the same time, due to the presence of the high-frequency boundary, the high-frequency voltage flowing through the lower-level line is reduced compared to the line at the same level where the fault occurred. Therefore, this embodiment stipulates that the measuring point can detect the high-frequency voltage signals of the line in which it is located, the adjacent lines, and the separated lines. To facilitate analysis, this embodiment sets a three-level per-unit threshold criterion as the basis for determining the line to which the tripped circuit breaker belongs. At the same time, the maximum value of the high-frequency voltage amplitude is used as the reference value, and it is normalized to 1. The mathematical representation of the three-level per-unit threshold interval constructed in this embodiment is:

[0147]

[0148] Where, is the first-level threshold interval; is the secondary threshold interval; It is the three-level threshold interval.

[0149] When the per-unit value of the high-frequency voltage amplitude is within three different intervals, namely, the first-level threshold interval α, the second-level threshold interval β, and the third-level threshold interval λ, it indicates that a circuit breaker has tripped on the line where the measuring point is located, the adjacent line, and the separated line. When it is within the first-level threshold interval α, it indicates that the line where the measuring point is located has tripped; when it is within the second-level threshold interval β, it indicates that the adjacent line has tripped; and when it is within the third-level threshold interval λ, it indicates that the separated line has tripped. In addition, if the per-unit value is lower than 0.1 pu, it is regarded as an error caused by fluctuations within the system, and it can be determined that no circuit breaker has tripped. Where pu is the per-unit value unit, this embodiment can accurately identify the operating status of the circuit breaker and the line to which it belongs through the above method, providing reliable information support for subsequent fault location.

[0150] After identifying the circuit breaker action state, this embodiment further analyzes the specific protection type that drives the circuit breaker to trip. In this embodiment, the same circuit breaker can be driven by different protections. Therefore, this embodiment first summarizes and organizes the various protection drive types of the line to which the tripped circuit breaker belongs; secondly, combined with the detection results of the sampling time window and the relay protection timing logic, the summarized protection drive types are preliminarily screened to narrow the range of protection drive types; finally, combined with the action logic of the relay protection, the actual action protection type is reversely deduced. For the sake of clarity, this embodiment uses the identification of the line and lines The relay protection type of the tripped circuit breaker is described in detail as an example. and lines The high-frequency voltage signal is detected in the sampling time window 1 and the sampling time window 3 respectively. In this embodiment, the line and lines The above protection drive mode types are summarized, and the line and lines The summary results of the protection drive types of the upper circuit breaker are shown in Table 2. It can be seen from Table 2 that for the line Since the circuit breaker tripped during the main protection period, it may be caused by the line , busbar B2 and busbar B3 main protection to drive; for line Since the circuit breaker tripped during the remote backup protection period, it may be caused by the line and lines At the same time, this embodiment takes into account the situation that the circuit breaker refuses to operate during the isolation fault process, and no high-frequency voltage signal is detected in the sampling time window 2, so it can be inferred that the line The tripping of the upper circuit breaker is not driven by the related protection of the busbar, thus excluding the main protection of busbars B2 and B3, and confirming the line The circuit breaker that operates on the line The main protection driver, in addition, if the line The tripped circuit breaker is connected to the line Remote backup protection drive should be on the line High-frequency signals are detected on the line, but this situation does not occur in this scenario. Based on this, this embodiment excludes the line Remote backup protection can determine the line The circuit breaker that tripped is caused by the line Through the above analysis, it can be seen that the method proposed in this embodiment can effectively identify the action status of the protection and combine the identification results of the circuit breaker and the protection action status to form local switch action information, as shown in Table 2:

[0151] Table 2

[0152]

[0153] This embodiment constructs a corresponding relay protection Bayesian network model for each suspected faulty component determined from the perspective of electrical quantities, and then compares the local switch action information with the switch action information uploaded by the SCADA system. Through this comparison process, the false alarm information in the switch action information uploaded by the SCADA system is corrected and the missed alarm information is supplemented, and then the erroneous network node status in each relay protection Bayesian network model is corrected to ensure that the status of each network node in the relay protection Bayesian network model is accurate and reliable. On this basis, this embodiment combines the corrected correct component status information to perform reverse reasoning on the relay protection Bayesian network model, thereby obtaining the fault posterior probability of each suspected faulty component, and finally uses the cubic normalization processing of the fault posterior probability of each suspected faulty component to obtain the switch fault degree. Figure 13 : is a schematic diagram of the overall process of evaluating the switching fault degree provided by an embodiment of the present invention. The calculation formula of the switching fault degree is:

[0154]

[0155] Where, For the The switching fault degree of each suspected faulty component; The first The cube of the original fault posterior probability of the suspected faulty component; is the cube of the posterior probability of fault of bus Bi; For the line The cube of the posterior probability of failure.

[0156] In this embodiment, the switching quantity fault degree focuses more on the inference of fault possibility based on the probability model, while the electrical quantity fault degree is directly derived from the physical measurement of the electrical quantity. The electrical quantity fault degree reflects the electrical characteristics at the time of the fault. Both represent the fault information of the component from different perspectives and are used for the calculation of the switching quantity and electrical quantity fault degrees respectively.

[0157] S5. Taking the electrical quantity fault degree and the switch quantity fault degree as independent evidence bodies, the improved Dempster-Schafer evidence theory identification framework that introduces a full set subset is used to calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory.

[0158] In some embodiments, the step of using the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set to respectively calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory includes:

[0159] Defining the credible interval and maximum uncertainty interval of each suspected fault component in the evidence body according to the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set, and expressing the credible interval and the maximum uncertainty interval in the form of interval numbers;

[0160] Based on the credible interval and maximum uncertainty interval of each suspected faulty component expressed in the form of interval numbers, calculate the projected distance between the credible interval and the maximum uncertainty interval of each suspected faulty component in the evidence body;

[0161] Measuring the uncertainty of the evidence body using the projection distance as an indicator to obtain an uncertainty measurement value of the evidence body;

[0162] According to the uncertainty measurement value of the evidence body and the preset evidence adjustment coefficient, the credible interval projection distance weight of the evidence body from the perspective of evidence theory is obtained;

[0163] The probability of the subset of the entire evidence set is evenly assigned to all suspected faulty components to obtain the component probability distribution and calculate the information entropy of the component probability distribution;

[0164] According to the information entropy of the component probability distribution and the preset probability adjustment coefficient, the information entropy weight of the evidence body from the perspective of probability theory is calculated.

[0165] Specifically, the Dempster-Shafer evidence theory (DS evidence theory) can effectively handle the uncertainty between evidence bodies and present evidence information in a more intuitive and comprehensive manner. With this special advantage, this evidence theory has been widely used in the field of information fusion. In evidence theory, the identification framework is a finite non-empty set composed of different elements. In this embodiment, the identification framework is defined as the set of all suspected faulty elements in the DC distribution network. The specific expression of the identification framework is:

[0166]

[0167] Where, To identify the framework, that is, the set of all suspected faulty components; To identify the element in the frame, it represents the nth suspected faulty component (such as a busbar or line); n is the total number of suspected faulty components.

[0168] In a given recognition framework, if there exists a function m that satisfies the specified constraints, then the function m is called the basic probability assignment function of the evidence theory. The mathematical expression of the constraints specified by the function m is:

[0169]

[0170] Where, Empty set The basic probability assignment of ; Assign a function to the basic probability; To identify a subset in the framework, the subset may contain one or more suspected faulty components; For subset is a proper subset of the recognition framework; is an empty set, which does not contain any faulty components.

[0171] when hour, is called a focal element of the recognition frame, where To identify suspected faulty components The confidence distribution of the basic probability assignment function m can be expressed as:

[0172]

[0173]

[0174]

[0175] Where, To identify the framework Next, based on the subset of evidence m credible interval of ; To identify the framework Next, based on the subset of evidence m Trust function; To identify the framework Next, based on the subset of evidence m Likelihood function of ; To identify the framework The i-th subset in ; Assigning functions to the basic probabilities of subsets of focal elements; is a subset of the focal element.

[0176] The core of DS evidence theory lies in the Dempster combination rule, which reflects the fusion process between various evidence bodies. The specific expression of the combination rule is:

[0177]

[0178] Where, All are bodies of evidence; For the body of evidence Subset The basic probability assignment of ; For the body of evidence Subset The basic probability assignment of ; For the body of evidence Subset The basic probability assignment is: k is the conflict factor between evidence bodies, which is used to measure the degree of conflict between different evidence bodies. When k=1, it indicates that the evidence bodies are contradictory, and the combination rule cannot be fused. It represents the intersection of evidence bodies, which reflects the consistency of different evidences.

[0179] In order to avoid the failure of the combination rule due to an excessively large conflict coefficient, this embodiment improves the traditional DS evidence theory from the perspective of modifying the evidence body. On the basis of adding the full set to the identification framework, the fault degree weight is allocated from the perspectives of evidence theory and probability theory, and the weighted fault degree is fused. It should be noted that the conflict between evidence bodies mainly stems from the situation where the focal elements in the evidence body have no intersection. Based on this, this embodiment reduces the conflict between evidence bodies by supplementing the subsets in the identification framework and increasing the intersection of the focal elements in the evidence body, while ensuring that it does not affect the final decision result. Since the full set and all focal elements have intersections, this embodiment introduces the full set into the identification framework so that each evidence body contains the full set. At this time, the expression of the improved Dempster-Schafer evidence theory identification framework that introduces the full set subset is:

[0180]

[0181]

[0182] Where, To introduce the improved Dempster-Schafer evidence theory identification framework for full set subsets, it includes the original identification framework All subsets and complete works of itself; is the full set, which represents the set of all possible faulty components in the identification framework.

[0183] In this embodiment, the electrical quantity fault degree and the switch quantity fault degree are respectively used as evidence and body of evidence , specifically expressed as:

[0184]

[0185] Where, is the coefficient of the entire set of electrical quantity evidence; Electrical quantity evidence Assign basic probabilities to the entire set; Switching evidence Assign basic probabilities to the entire set; is the coefficient of the entire set of switch evidence.

[0186] From the perspective of evidence theory, this embodiment proposes a method for measuring the uncertainty of an evidence body. This method uses the projected distance between the credible interval of the evidence body and the maximum uncertainty interval as an indicator to measure the uncertainty of the evidence body. First, this embodiment constructs the credible interval of the suspected faulty component in the evidence body based on the credible interval of the basic probability assignment function m. At the same time, the credible interval and the maximum uncertainty interval are treated as interval numbers. The interval number is defined as follows:

[0187]

[0188] Where, is the number of credible intervals; is the minimum value of the credible interval; is the maximum value of the credible interval; x is the center value of the credible interval number; R is the radius of the credible interval number, which is used to measure the uncertainty range of the interval.

[0189] Next, this embodiment calculates the projection distance between the number of credible intervals of each suspected fault and the maximum number of uncertainty intervals in the evidence body. The calculation formula for the projection distance is as follows:

[0190]

[0191] Where, is the projected distance between the number of credible intervals and the maximum number of uncertainty intervals; is the upper bound of the maximum uncertainty interval; is the lower bound of the maximum uncertainty interval; b is the maximum uncertainty interval; is the Euclidean norm.

[0192] Then, this embodiment uses the projection distance as an indicator to measure the uncertainty of the evidence body, and obtains the uncertainty measurement value of the evidence body. The specific expression of the uncertainty measurement value is:

[0193]

[0194] Where, is the uncertainty of each body of evidence. The larger its value, the greater the uncertainty of the evidence. MUI is the maximum uncertainty interval. .

[0195] On this basis, this embodiment assigns evidence weights according to the uncertainty measurement value of the evidence body and the preset evidence adjustment coefficient, and obtains the credible interval projection distance weight of the evidence body from the perspective of evidence theory. The mathematical expression of the credible interval projection distance weight is:

[0196]

[0197] Where, For the body of evidence The credible interval projection distance weight, that is, the evidence theory weight; is the preset evidence adjustment coefficient; For the body of evidence Uncertainty measure of is the body of evidence; M is the total number of bodies of evidence.

[0198] From the mathematical expression of the credible interval projection distance weight, we can see that and Inversely proportional, that is, the greater the uncertainty of the evidence, the smaller its weight.

[0199] Next, this embodiment elaborates on the weight distribution from the perspective of probability theory. Probability theory can be regarded as a further extension of evidence theory. By converting the evidence under evidence theory into probabilities, the corresponding probabilities are obtained, and then the information entropy of the probabilities is calculated, which indirectly realizes the uncertainty measurement of the evidence body from the perspective of probability theory. The information entropy is defined as:

[0200]

[0201] Where, is the information entropy of the evidence body m; For the body of evidence Subset The basic probability assignment of .

[0202] In this embodiment, the weight of the evidence body is assigned according to the information entropy ratio. The information entropy weight assignment formula is:

[0203]

[0204] Where, is the information entropy weight, i.e. the probability theory weight; is the preset probability adjustment coefficient; For the body of evidence Information entropy.

[0205] From the information entropy weight distribution formula, we can see that the smaller the information entropy of the evidence body, the less uncertainty information it carries and the greater the weight it occupies.

[0206] S6. The credible interval projection distance weight and information entropy weight of each evidence body are fused through the combined weighting method, and all evidence bodies are weighted and fused using the fused weights to obtain the fused fault probability of each suspected faulty component.

[0207] In order to achieve the fusion of multi-source fault information, this embodiment proposes a method for combining the weights of evidence bodies. Specifically, the weight distribution is completed from the perspective of evidence theory and probability theory to obtain the credible interval projection distance weight and information entropy weight Finally, this embodiment combines the credible interval projection distance weight and information entropy weight of each evidence body through a combined weighting method to determine the final weight distribution of each evidence body, ensuring that the weight can comprehensively reflect the credibility and uncertainty of the evidence body, thereby obtaining the fused weight. The fused weight not only reflects the relative importance of the evidence body in the fault degree assessment of electrical quantities and switching quantities, but also balances the weight contributions from two different theoretical perspectives. The mathematical expression of the fused weight is:

[0208]

[0209] Where, For the body of evidence The weight after fusion.

[0210] This embodiment performs weighted averaging on the evidence bodies based on the fused weights to obtain the fused fault probability of each suspected faulty component. The weighted averaging process fully considers the final weight of each evidence body, ensuring that the contribution to the fusion result is proportional to the credibility of the evidence body. The fused fault probability obtained by weighted averaging can more accurately reflect the comprehensive fault probability of each suspected faulty component. This embodiment makes decisions based on the fused fault probability of each suspected faulty component, where the component with the highest fused fault probability is the faulty component. This embodiment uses dual weight optimization to consider both the degree of conflict between evidence and the certainty of probability distribution, significantly improving the reliability of fault location. The mathematical representation of the weighted averaging of the evidence bodies is:

[0211]

[0212] Where, is the fusion failure probability; For the body of evidence Subset The basic probability assignment of .

[0213] S7. Locate the fault section of the DC distribution network according to the fused fault probability, and drive the circuit breaker to isolate the fault.

[0214] This embodiment extracts high-frequency fault voltage components 3 ms before and after a fault based on the line parameters of the DC distribution network. The extracted high-frequency fault voltage components are then processed using a BSC reconstruction algorithm to obtain corresponding results to identify suspected faulty components. Furthermore, this embodiment uses an electrical quantity assessment method to calculate the electrical quantity fault degree. Based on the precise identification of high-frequency voltage signals and combined with the sequential action logic of relay protection, the operating status of circuit breakers and protections is accurately identified. Furthermore, erroneous information in systems such as SCADA is corrected, and reverse reasoning is performed on each Bayesian network after the component status information is corrected to calculate the switching quantity fault degree. Furthermore, this embodiment improves on the traditional DS evidence theory by supplementing the identification framework subset to calculate weights from the perspectives of evidence theory and probability theory. The weighted evidence is then fused based on the Dempster rule, and the component corresponding to the maximum fused fault probability is determined as the faulty section, thereby achieving accurate and rapid location of the faulty section in the DC distribution network.

[0215] In order to verify the correctness and effectiveness of the fault section location method proposed in this embodiment, this embodiment built an IEEE33-node DC distribution network simulation model based on PSCAD / EMTDC, and carried out research by taking line fault and busbar fault as examples. For the simulation analysis of line bipolar short circuit fault, this embodiment set the line A bipolar short circuit fault occurs, and the line Main protection drive circuit breaker Action, remote backup protection drive circuit The left end circuit breaker trips to isolate the fault. Figure 14 This is a schematic diagram of a fault location process based on an improved Dempster-Schafer evidence theory identification framework provided by an embodiment of the present invention. In this embodiment, a high-frequency current sparse vector is obtained by reconstructing the BCS reconstruction algorithm. Figure 15 This is a three-dimensional schematic diagram of the high-frequency fault current sparse vector reconstructed by the Bayesian compressed sensing reconstruction algorithm provided by the embodiment of the present invention. It can be seen from the diagram that bus B13, bus B14, bus B15 and bus B16 have large high-frequency current amplitudes, and thus the suspected fault components are judged to be bus B13, bus B14, bus B15, bus B16, line ,line and lines , the fault line set In the fault area, after determining the fault range, the electrical quantity fault degree is obtained based on the above fault degree evaluation method, as shown in Table 3:

[0216] Table 3

[0217]

[0218] As shown in Table 3, the discrimination degree of each suspected fault component in the electrical fault degree is low, and the line The fault degree of the non-faulty component B14 is 22.91%, which is lower than the fault degree of the non-faulty component B14. The fault degrees are close, which means that if only a single electrical quantity fault information is used for fault location, positioning errors will occur. To avoid this problem, this embodiment adjusts the electrical quantity fault information from the perspective of switch quantity. This embodiment needs to determine whether there is omission and false alarm information in the relay protection action sequence information collected by the SCADA system, etc., and assign the corrected switch quantity information to the Bayesian network. Under the premise of knowing the suspected fault component, the line and lines The upper circuit breaker protection drive mode is shown in Table 4:

[0219] Table 4

[0220]

[0221] Table 4 shows the circuit and lines The specific protection driving mode of the upper circuit breaker is due to the line The tripping of the upper circuit breaker occurs during the main protection period, which can eliminate the backup protection of the line and busbar. After analysis, it can be found that the line The tripped circuit breaker is connected to the line The main protection drive, combined with Table 4 analysis, shows that the local switch action information is the line ,breaker and circuit breakers Action, circuit breaker Refusal to operate, circuit breaker The action is a false alarm and misses the The refusal to operate information is assigned to the corrected switch quantity information in the Bayesian network. On this basis, the fault component is calculated. The switching fault degree is 83.95%, which is much higher than other non-faulty components. Compared with the electrical fault degree, the faulty component distinction in the switching fault degree is higher and the reliability is stronger. In order to ensure the objectivity of the positioning result and further improve the positioning accuracy, the electrical fault degree and the switching fault degree are fused. The fusion results are shown in Table 5:

[0222] Table 5

[0223]

[0224] From the fusion results in Table 5, we can see that the faulty element is line The fusion positioning result is 94.70%. Compared with the single fault degree, the positioning accuracy is further improved, which avoids the positioning error problem and verifies the accuracy of the fault section positioning method proposed in this invention when a line fault occurs.

[0225] An embodiment of the present invention provides a method for locating a fault section in a DC distribution network based on multi-source information fusion. The method reconstructs a high-frequency fault current sparse vector based on a high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network; uses a normal distribution model to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector to obtain a high-frequency current amplitude estimation; cubically normalizes the high-frequency current amplitude estimation of each suspected fault component to obtain the electrical quantity fault degree of each suspected fault component; and corrects the relay protection Bayesian network model based on pre-identified local switch quantity action information. Based on the node status of the type, the switch fault degree of each suspected faulty component is calculated through reverse reasoning. The electrical fault degree and switch fault degree are used as independent evidence bodies. The improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set is used to calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory. The credible interval projection distance weight and information entropy weight of each evidence body are fused through a combined weighting method, and all evidence bodies are weighted and fused using the fused weights to obtain the fused fault probability of each suspected faulty component. The faulty section of the DC distribution network is located based on the fused fault probability, and the circuit breaker is actuated to isolate the fault. Compared with existing technologies, this method achieves rapid and accurate positioning of the DC distribution network fault section by fusing high-frequency transient electrical quantity characteristics with switch action information and combining it with improved evidence theory. It also effectively improves the fault tolerance capability of fault location and ensures the safe and stable operation of the power grid.

[0226] It should be noted that the size of the serial numbers of the above-mentioned processes does not 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 this application.

[0227] In one embodiment, Figure 16 As shown, an embodiment of the present invention provides a DC distribution network fault section location system based on multi-source information fusion, the system comprising:

[0228] The signal reconstruction module 101 is used to reconstruct a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network;

[0229] an amplitude estimation module 102 for quantitatively analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude;

[0230] The electrical quantity analysis module 103 is used to perform cubic normalization on the high-frequency current amplitude estimation value of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component;

[0231] The switch value analysis module 104 is used to modify the node state of the relay protection Bayesian network model based on the pre-identified local switch value action information and calculate the switch value fault degree of each suspected fault component through reverse reasoning;

[0232] A dual-weight analysis module 105 is configured to use the electrical quantity fault degree and the switch quantity fault degree as independent evidence bodies and to calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory respectively using an improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set;

[0233] The probability fusion module 106 is used to fuse the credible interval projection distance weight and information entropy weight of each evidence body through a combined weighting method, and use the fused weights to weight all evidence bodies to obtain the fused fault probability of each suspected faulty component;

[0234] The fault location module 107 is configured to locate the fault section of the DC distribution network according to the fused fault probability and drive the circuit breaker to isolate the fault.

[0235] Regarding the specific definition of a DC distribution network fault section positioning system based on multi-source information fusion, please refer to the above-mentioned definition of a DC distribution network fault section positioning method based on multi-source information fusion, which will not be repeated here. A person of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0236] The embodiment of the present invention provides a DC distribution network fault section positioning system based on multi-source information fusion, wherein the signal reconstruction module of the system reconstructs a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network; the amplitude estimation module uses a normal distribution model to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector to obtain a high-frequency current amplitude estimation value; the electrical quantity analysis module performs cubic normalization on the high-frequency current amplitude estimation value of each suspected fault component to obtain the electrical quantity fault degree of each suspected fault component; the switch quantity analysis module corrects the relay protection bayes according to the pre-identified local switch quantity action information The system uses the node states of the St. Petersburg network model to calculate the switch fault degree of each suspected faulty component through reverse reasoning. The dual-weight analysis module uses the electrical fault degree and switch fault degree as independent evidence bodies and uses the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set to calculate the credible interval projected distance weight from the evidence theory perspective and the information entropy weight from the probability theory perspective. The probability fusion module fuses the credible interval projected distance weight and information entropy weight of each evidence body through a combined weighting method and uses the fused weights to weight all evidence bodies to obtain the fused fault probability of each suspected faulty component. The fault location module locates the faulty section of the DC distribution network based on the fused fault probability and actuates the circuit breaker to isolate the fault. Compared with existing technologies, this system achieves rapid and accurate location of the DC distribution network fault section and actuates the circuit breaker to isolate the fault by fusing high-frequency transient electrical quantity characteristics with switch action information and combining it with improved evidence theory. This effectively improves the fault tolerance capability of fault location and ensures the safe and stable operation of the power grid.

[0237] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection of the claims.

Claims

1. A method for locating fault sections in a DC distribution network based on multi-source information fusion, characterized in that: The following steps are involved: According to the high-frequency transient voltage signal generated at the moment of bipolar short-circuit fault in DC distribution network, the high-frequency fault current sparse vector is reconstructed. Quantitatively analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude; The high-frequency current amplitude estimation value of each suspected faulty component is normalized by cubic method to obtain the electrical quantity fault degree of each suspected faulty component; The node states of the relay protection Bayesian network model are modified based on the pre-identified local switch action information, and the switch fault degree of each suspected faulty component is calculated through reverse reasoning. The electrical quantity fault degree and the switch quantity fault degree are taken as independent evidence bodies, and the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory are respectively calculated using the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set; The credible interval projection distance weight and information entropy weight of each evidence body are fused by the combined weighting method, and all evidence bodies are weighted and fused using the fused weights to obtain the fused fault probability of each suspected faulty component. Locating a fault section of the DC distribution network according to the fused fault probability, and driving a circuit breaker to isolate the fault; The step of reconstructing a high-frequency fault current sparse vector based on a high-frequency transient voltage signal generated at the moment a bipolar short-circuit fault occurs in the DC distribution network includes: At the moment of a bipolar short circuit fault in the DC distribution network, a node high-frequency impedance matrix is ​​constructed based on the node self-impedance data and the mutual impedance data between nodes in the DC distribution network. The high-frequency transient voltage signal at the fault measuring point within a preset time window before and after the occurrence of the bipolar short-circuit fault is synchronously collected by the voltage measuring device; The faulty busbar node is equivalent to a high-frequency current source, the current of the non-busbar fault node is set to zero, and a busbar fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal; The fault line is equivalent to a high-frequency current source of adjacent fault measuring points, the current between non-adjacent fault measuring points is set to zero, and a line fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal; Integrating the bus fault sparse node equation and the line fault sparse node equation into an underdetermined sparse node equation group; The underdetermined sparse node equations are transformed into a sparse signal reconstruction problem, and a Bayesian compressed sensing reconstruction algorithm is used to iteratively solve the problem to obtain a high-frequency fault current sparse vector.

2. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 1, characterized in that: The bus fault sparse node equation is used to describe the relationship between the voltage and current of each DC distribution network node when a bus fault occurs; The line fault sparse node equation is used to describe the relationship between the voltage and current of each DC distribution network node when a line fault occurs.

3. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 1, characterized in that: The step of quantitatively analyzing the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude includes: According to the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector, multiple groups of independent normal distribution models are constructed according to the adjacent relationship of the suspected fault components; The maximum value of the high-frequency current amplitude in adjacent suspected fault components is taken as the peak value of the corresponding normal distribution model, and the attenuation parameter of each group of normal distribution models is solved by the maximum likelihood estimation criterion; Based on the attenuation parameter, a normal distribution model is used to calculate an estimated value of the high-frequency current amplitude at a line midpoint between each pair of adjacent suspected faulty components.

4. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 3, characterized in that: The attenuation parameters include a standard deviation for controlling the attenuation rate and an electrical distance difference between two adjacent suspected faulty components under normal distribution.

5. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 1, characterized in that: The step of performing cubic normalization on the high-frequency current amplitude estimation value of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component includes: Performing a cubic operation on the high-frequency current amplitude estimation value of each suspected faulty component to obtain the high-frequency current cube amplitude of each suspected faulty component; The cubic amplitude of the high-frequency current is normalized to obtain the electrical quantity fault degree of each suspected fault component.

6. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 1, characterized in that: The step of correcting the node state of the relay protection Bayesian network model according to the pre-identified local switch quantity action information and calculating the switch quantity fault degree of each suspected fault component by reverse reasoning includes: According to the sequential action logic of the DC distribution network relay protection, a three-level sampling time window is set. The protection level of the tripped circuit breaker is determined by detecting the high-frequency transient voltage signal within each sampling time window, and the local circuit breaker action status of the protection level where the tripped circuit breaker is located is obtained. Identifying local switch action information according to the local circuit breaker action state, and using the local switch action information to correct the switch information uploaded by the data acquisition and monitoring system to obtain switch action correction information; According to the relay protection action logic of the DC distribution network, a relay protection Bayesian network model of the DC distribution network is constructed. The corresponding relay protection node states and circuit breaker node states in the relay protection Bayesian network model are updated using the switch action correction information to obtain a relay protection Bayesian network correction model. Based on the suspected fault components determined from the perspective of electrical quantities, the relay protection Bayesian network modified model is used to perform reverse probability reasoning to calculate the posterior probability of fault of each suspected fault component; The fault posterior probability of each suspected faulty component is cubically normalized to obtain the switching fault degree of each suspected faulty component.

7. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 6, characterized in that: The three-level sampling time window includes a first-level sampling time window, a second-level sampling time window and a third-level sampling time window; wherein, the first-level sampling time window corresponds to the main protection action period, the second-level sampling time window corresponds to the near backup protection action period, and the third-level sampling time window corresponds to the far backup protection action period.

8. The method for locating a fault section in a DC distribution network based on multi-source information fusion according to claim 1, characterized in that: The steps of respectively calculating the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory by using the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the full set include: Defining the credible interval and maximum uncertainty interval of each suspected fault component in the evidence body according to the improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set, and expressing the credible interval and the maximum uncertainty interval in the form of interval numbers; Based on the credible interval and maximum uncertainty interval of each suspected faulty component expressed in the form of interval numbers, calculate the projected distance between the credible interval and the maximum uncertainty interval of each suspected faulty component in the evidence body; Measuring the uncertainty of the evidence body using the projection distance as an indicator to obtain an uncertainty measurement value of the evidence body; According to the uncertainty measurement value of the evidence body and the preset evidence adjustment coefficient, the credible interval projection distance weight of the evidence body from the perspective of evidence theory is obtained; The probability of the subset of the entire evidence set is evenly assigned to all suspected faulty components to obtain the component probability distribution and calculate the information entropy of the component probability distribution; According to the information entropy of the component probability distribution and the preset probability adjustment coefficient, the information entropy weight of the evidence body from the perspective of probability theory is calculated.

9. A DC distribution network fault section location system based on multi-source information fusion, characterized in that: The system comprises: A signal reconstruction module is used to reconstruct a high-frequency fault current sparse vector based on the high-frequency transient voltage signal generated at the moment of a bipolar short-circuit fault in the DC distribution network; an amplitude estimation module, configured to quantitatively analyze the nonlinear attenuation trend of the high-frequency current amplitude of each suspected fault component in the high-frequency fault current sparse vector using a normal distribution model to obtain an estimated value of the high-frequency current amplitude; The electrical quantity analysis module is used to perform cubic normalization on the high-frequency current amplitude estimation value of each suspected faulty component to obtain the electrical quantity fault degree of each suspected faulty component; The switch quantity analysis module is used to modify the node status of the relay protection Bayesian network model based on the pre-identified local switch quantity action information, and calculate the switch quantity fault degree of each suspected fault component through reverse reasoning; a dual-weight analysis module, for treating the electrical quantity fault degree and the switch quantity fault degree as independent bodies of evidence, and using an improved Dempster-Schafer evidence theory identification framework that introduces a subset of the entire set to respectively calculate the credible interval projection distance weight from the perspective of evidence theory and the information entropy weight from the perspective of probability theory; The probability fusion module is used to fuse the credible interval projection distance weight and information entropy weight of each evidence body through the combined weighting method, and use the fused weights to weight all evidence bodies to obtain the fused fault probability of each suspected faulty component; a fault location module, configured to locate a fault section of the DC distribution network according to the fused fault probability and drive a circuit breaker to isolate the fault; The signal reconstruction module is specifically used to: At the moment of a bipolar short circuit fault in the DC distribution network, a node high-frequency impedance matrix is ​​constructed based on the node self-impedance data and the mutual impedance data between nodes in the DC distribution network. The high-frequency transient voltage signal at the fault measuring point within a preset time window before and after the occurrence of the bipolar short-circuit fault is synchronously collected by the voltage measuring device; The faulty busbar node is equivalent to a high-frequency current source, the current of the non-busbar fault node is set to zero, and a busbar fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal; The fault line is equivalent to a high-frequency current source of adjacent fault measuring points, the current between non-adjacent fault measuring points is set to zero, and a line fault sparse node equation is established based on the node high-frequency impedance matrix and the high-frequency transient voltage signal; Integrating the bus fault sparse node equation and the line fault sparse node equation into an underdetermined sparse node equation group; The underdetermined sparse node equations are transformed into a sparse signal reconstruction problem, and a Bayesian compressed sensing reconstruction algorithm is used to iteratively solve the problem to obtain a high-frequency fault current sparse vector.

Images