Power distribution network fault positioning method

Abstract

The invention relates to the technical field of power distribution network faults, and discloses a power distribution network fault positioning method, which comprises the following steps: acquiring voltage drop vectors before and after a power distribution network fault and the line length of each line, and when the ratio of the maximum line length to the minimum line length is greater than or equal to a ratio threshold value, determining that the power distribution network fault exists; virtual nodes are inserted based on the minimum line length to perform segmentation reconstruction, a node impedance matrix containing real nodes and the virtual nodes is constructed, the voltage drop vector is used as an observation vector, a submatrix of the node impedance matrix is used as a sensing matrix, and a node fault current source vector is determined by using a sparse Bayesian learning algorithm so as to determine a fault positioning result. According to the method, the adaptive complex topology is reconstructed by inserting the virtual nodes into the segments, and high-precision positioning is realized under a small amount of measurement equipment in combination with the sparse Bayesian learning algorithm, so that the problems that the traditional method is highly dependent on the measurement equipment and is easily influenced by topology complexity and distributed power supply fluctuation are effectively solved; and the accuracy, robustness and the like of fault positioning are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution network fault technology, and in particular to a method for locating power distribution network faults. Background Technology

[0002] Distribution network accidents account for up to 90% of all power grid accidents, with single-phase grounding faults in medium-voltage distribution networks accounting for 80%. Rapid and accurate fault location is the core of fault isolation and promoting the self-healing of distribution networks. However, the complex topology of distribution networks, with their multiple branches, radial patterns, and interspersed long and short lines, coupled with the current situation where insufficient measurement equipment makes it difficult to achieve full coverage, and the uncertainty and volatility brought about by distributed power source access, pose serious challenges to fault location.

[0003] In my country's medium- and low-voltage distribution networks, ineffective grounding is commonly used. During a fault, the fault current only forms a loop through the line-to-ground capacitance, resulting in weak and ambiguous fault current characteristics, making conventional fault location methods difficult to identify effectively. Existing technologies include: impedance methods suffer from fault dead zones and are sensitive to load; traveling wave methods are significantly affected by branch lines; artificial intelligence methods face the contradiction between small-sample learning and model generalization; state estimation methods have high computational complexity and poor real-time performance and economy; and compressed sensing-based methods are prone to misjudgment when the fault point is close to a node or when the impedance difference between adjacent lines is too large. With the development of new distribution systems, the number of branches and nodes in the distribution network continues to increase, and the topology becomes more complex, while the number of measuring devices remains insufficient. Therefore, there is an urgent need for a method to achieve high-precision fault location with a limited number of measuring devices. Summary of the Invention

[0004] Based on this, it is necessary to propose a fault location method for distribution networks to address the above problems. By inserting virtual nodes and reconstructing segments to adapt to complex topologies, and combining sparse Bayesian learning algorithms, high-precision location can be achieved with a small number of measuring devices. This effectively overcomes the problems of high dependence on measuring devices and susceptibility to topological complexity and distributed power source fluctuations in traditional methods. It significantly improves the accuracy and robustness of fault location, and ensures the self-healing ability and power supply reliability of the distribution network in complex operating environments.

[0005] To achieve the above objectives, the present invention provides a method for locating faults in a distribution network in a first aspect, the method comprising: Obtain the voltage drop vector before and after the distribution network fault occurs, as well as the line length of each line in the distribution network; The maximum and minimum line lengths of each line in the distribution network are determined. If the ratio between the maximum and minimum line lengths is greater than or equal to a ratio threshold, the distribution network is reconstructed by inserting virtual nodes based on the minimum line length and the line lengths of each line in the distribution network to obtain the reconstructed distribution network. Determine the node impedance matrix of the nodes in the reconstructed distribution network, wherein the nodes include real nodes and virtual nodes; The voltage drop vector is used as the observation vector, the submatrix of the node impedance matrix is ​​used as the sensing matrix, and the sparse Bayesian learning algorithm is used to determine the node fault current source vector based on the observation vector and the sensing matrix. Based on the node fault current source vector, the fault location result of the distribution network is determined.

[0006] Optionally, the sparse Bayesian learning algorithm is a feature-adaptive sparse Bayesian learning algorithm. The step of using the sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the sensing matrix includes: Obtain the line characteristic factor vector of the nodes in the reconstructed distribution network; The sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance are initialized to obtain the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the first iteration. Based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix, the posterior covariance matrix and posterior mean vector in the t-th iteration are determined, where the initial value of t is 1. Based on the posterior mean vector and posterior covariance vector of the t-th iteration, the sparse hypervector of the t-th iteration is updated to obtain the sparse hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the node line feature factor vector, the adaptive hyperparameter vector of the t-th iteration is updated to obtain the adaptive hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the observation vector and the perception matrix, the noise variance of the (t+1)-th iteration is determined. If t=1, let t=t+1, then return to the step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix. If t is not equal to 1, then based on the noise variance, posterior mean vector and posterior covariance matrix in the t-th iteration, as well as the node line feature factor vector and the perception matrix, the weight factors of all nodes of the reconstructed distribution network in the t-th iteration are obtained. Based on the sparse hyperparameter vector, adaptive hyperparameter vector, and weight factors of all nodes in the t-th iteration, and the sparse hyperparameter vector and adaptive hyperparameter vector in the (t-1)-th iteration, determine the weighted relative change in the t-th iteration. If the weighted relative change in the t-th iteration is less than or equal to the convergence threshold, or if t equals the maximum number of iterations, then the posterior mean vector in the t-th iteration will be used as the node fault current source vector. Otherwise, let t=t+1, and return to the step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix.

[0007] Optionally, the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance at the (t+1)th iteration are obtained using the following formula, as well as the weighted relative change at the tth iteration: ; in, Let i be the i-th element in the sparse hyperparameter vector of the (t+1)-th iteration. and These are the first preset prior hyperparameter and the second preset prior hyperparameter, respectively. Let i be the i-th element in the posterior mean vector of the t-th iteration. Let i be the i-th diagonal element in the posterior covariance matrix of the t-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the (t+1)-th iteration. and These are the third and fourth preset prior hyperparameters, respectively. Let i be the i-th element in the node line feature factor vector. Let Variance be the noise variance in the (t+1)th iteration. Let be the observation vector. The perception matrix is... Let be the posterior mean vector of the t-th iteration. It is the L2 norm. This is a function to sum the diagonal elements of a matrix. It is the transpose symbol. Let be the posterior covariance matrix of the t-th iteration. The total number of elements in the observation vector. Let be the weighted relative change in the t-th iteration. This refers to the total number of nodes in the reconstructed distribution network. Let be the weight factor of the i-th node in the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the (t-1)-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the t-th iteration. Let be the i-th element in the adaptive hyperparameter vector during the (t-1)-th iteration.

[0008] Optionally, obtaining the weight factors for all nodes of the reconstructed distribution network in the t-th iteration based on the noise variance, posterior mean vector, and posterior covariance matrix of the t-th iteration, as well as the node line feature factor vector and the perception matrix, includes: Based on the noise variance and posterior mean vector in the t-th iteration, and the perception matrix, determine the signal-to-noise ratio of all nodes in the t-th iteration; Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, determine the failure probability of all nodes in the t-th iteration. The weighting factors for all nodes in the t-th iteration are determined based on the signal-to-noise ratio and fault probability of all nodes in the t-th iteration, as well as the node line feature factor vector.

[0009] Optionally, determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix, includes: Based on the sparse hyperparameter vector and adaptive hyperparameter vector in the t-th iteration, and the node line feature factor vector, determine the hyperparameter diagonal matrix in the t-th iteration; Based on the hyperparameter diagonal matrix and noise variance in the t-th iteration, and the perception matrix, determine the posterior covariance matrix in the t-th iteration. Based on the noise variance in the t-th iteration, the observation vector, and the perception matrix, the posterior mean vector in the t-th iteration is determined.

[0010] Optionally, obtaining the line characteristic factor vector of the nodes in the reconstructed distribution network includes: Based on the line lengths of all adjacent line segments of each node in the reconstructed distribution network, determine the line characteristic factor of each node; The line characteristic factors of each node are normalized to obtain the standard line characteristic factors of each node. Based on the standard line characteristic factors of each node, construct the node line characteristic factor vector.

[0011] Optionally, the step of segmenting and reconstructing the distribution network by inserting virtual nodes into each line based on the minimum line length and the line length of each line in the distribution network to obtain the reconstructed distribution network includes: Based on the length of each line in the distribution network and the minimum line length, determine the total number of virtual nodes that need to be inserted for each line in the distribution network. Based on the minimum line length interval, and according to the total number of virtual nodes required to be inserted for each line in the distribution network, the distribution network is reconstructed in segments by inserting virtual nodes, thus obtaining the reconstructed distribution network.

[0012] Optionally, the method further includes: If the ratio between the maximum line length and the minimum line length is less than the ratio threshold, the target line length is determined based on the line length of each line in the distribution network, and the segmented reconstruction of each line in the distribution network is performed by inserting virtual nodes based on the target line length to obtain the reconstructed distribution network.

[0013] Optionally, obtaining the voltage drop vector before and after the distribution network fault includes: Obtain the voltage amplitude of each measuring device in the power distribution network before and after the fault occurs; A first voltage amplitude vector is constructed based on the voltage amplitude of each measuring device in the power distribution network before the fault occurs. A second voltage amplitude vector is constructed based on the voltage amplitude of each measuring device in the power distribution network after a fault occurs. The voltage drop vector is determined based on the first voltage amplitude vector and the second voltage amplitude vector.

[0014] Optionally, determining the fault location result of the distribution network based on the node fault current source vector includes: When the number of non-zero elements in the node fault current source vector is not equal to 1, determine the largest non-zero element and the second largest non-zero element in the node fault current source vector, and take the line segment between the node corresponding to the largest non-zero element and the node corresponding to the second largest non-zero element as the fault location result. When the number of non-zero elements in the node fault current source vector is equal to 1, the adjacent line segment of the node corresponding to the non-zero element in the node fault current source vector is taken as the fault location result.

[0015] To achieve the above objectives, the present invention provides a power distribution network fault location device in a second aspect, the device comprising: The acquisition module is used to acquire the voltage drop vector before and after the distribution network fault occurs, as well as the line length of each line in the distribution network; The determination and reconstruction module is used to determine the maximum and minimum line lengths of each line in the distribution network, and when the ratio between the maximum and minimum line lengths is greater than or equal to a ratio threshold, the module performs segmented reconstruction of each line in the distribution network by inserting virtual nodes based on the minimum line length and the line lengths of each line in the distribution network, thereby obtaining the reconstructed distribution network. The first determining module is used to determine the node impedance matrix of the nodes in the reconstructed distribution network, wherein the nodes include real nodes and virtual nodes; The determination module is used to take the voltage drop vector as the observation vector, take the submatrix of the node impedance matrix as the sensing matrix, and use a sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the sensing matrix. The second determining module is used to determine the fault location result of the distribution network based on the node fault current source vector.

[0016] To achieve the above objectives, the present invention provides, in a third aspect, a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the power distribution network fault location method as described in any one of the first aspects.

[0017] To achieve the above objectives, the present invention provides a computer device in a fourth aspect, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the power distribution network fault location method as described in any one of the first aspects.

[0018] The present invention has the following beneficial effects: The above method obtains the voltage drop vector before and after a distribution network fault, as well as the line length of each line in the distribution network. Then, it determines the maximum and minimum line lengths of each line in the distribution network. When the ratio between the maximum and minimum line lengths is greater than or equal to a threshold value, it performs segmented reconstruction of each line in the distribution network by inserting virtual nodes, based on the minimum line length and the line lengths of each line in the distribution network, to obtain the reconstructed distribution network. Then, it determines the node impedance matrix of the nodes in the reconstructed distribution network, where nodes include both real and virtual nodes. Finally, it calculates the voltage... Using the reduced vector as the observation vector and the submatrix of the node impedance matrix as the sensing matrix, a sparse Bayesian learning algorithm is used to determine the node fault current source vector based on the observation vector and the sensing matrix. Finally, the fault location result of the distribution network is determined based on the node fault current source vector. That is, by inserting virtual nodes and reconstructing segments to adapt to complex topologies, combined with the sparse Bayesian learning algorithm, high-precision location is achieved with a small number of measuring devices. This effectively overcomes the problems of high dependence on measuring devices and susceptibility to topological complexity and distributed power source fluctuations in traditional methods, significantly improving the accuracy and robustness of fault location, and ensuring the self-healing capability and power supply reliability of the distribution network in complex operating environments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] in: Figure 1 This is a schematic diagram of a power distribution network fault location method according to an embodiment of this application; Figure 2 This is a schematic diagram of a power distribution network fault location device in an embodiment of this application; Figure 3 This is a diagram showing the internal structure of a computer device in some embodiments. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Distribution network accidents account for up to 90% of all power grid accidents, with single-phase grounding faults in medium-voltage distribution networks accounting for 80%. Rapid and accurate fault location is the core of fault isolation and promoting the self-healing of distribution networks. However, the complex topology of distribution networks, with their multiple branches, radial patterns, and interspersed long and short lines, coupled with the current situation where insufficient measurement equipment makes it difficult to achieve full coverage, and the uncertainty and volatility brought about by distributed power source access, pose serious challenges to fault location.

[0023] In my country's medium- and low-voltage distribution networks, ineffective grounding is commonly used. During a fault, the fault current only forms a loop through the line-to-ground capacitance, resulting in weak and ambiguous fault current characteristics, making conventional fault location methods difficult to identify effectively. Existing technologies include: impedance methods suffer from fault dead zones and are sensitive to load; traveling wave methods are significantly affected by branch lines; artificial intelligence methods face the contradiction between small-sample learning and model generalization; state estimation methods have high computational complexity and poor real-time performance and economy; and compressed sensing-based methods are prone to misjudgment when the fault point is close to a node or when the impedance difference between adjacent lines is too large. With the development of new distribution systems, the number of branches and nodes in the distribution network continues to increase, and the topology becomes more complex, while the number of measuring devices remains insufficient. Therefore, there is an urgent need for a method to achieve high-precision fault location with a limited number of measuring devices.

[0024] To address the aforementioned issues, this application proposes a fault location method for distribution networks. By inserting virtual nodes and reconstructing segments to adapt to complex topologies, and combining this with a sparse Bayesian learning algorithm, high-precision fault location is achieved with a limited number of measuring devices. This effectively overcomes the problems of high dependence on measuring devices and susceptibility to topological complexity and distributed power source fluctuations in traditional methods. It significantly improves the accuracy and robustness of fault location, ensuring the self-healing capability and power supply reliability of the distribution network under complex operating environments. The specific implementation principle will be described in detail in the following embodiments.

[0025] This application provides a method for locating faults in a power distribution network in its first aspect.

[0026] Please see Figure 1 This is a schematic diagram of a power distribution network fault location method according to an embodiment of this application. The method includes: Step 110: Obtain the voltage drop vector before and after the distribution network fault, as well as the line length of each line in the distribution network.

[0027] Regarding the method of obtaining the voltage drop vector, in some embodiments, the difference between the voltage amplitudes of various measuring devices in the distribution network before and after the fault occurs can be obtained to determine the voltage drop vector.

[0028] It should be noted that in the distribution network's line topology, there is a line connecting two adjacent real nodes. However, as the distribution network's line topology becomes more complex and the number of real nodes increases, it becomes impossible to configure a measuring device for every real node. Therefore, each measuring device is only installed in some key real nodes of the distribution network.

[0029] Regarding the selection of installation locations for measuring equipment, in some embodiments, the various measuring devices can be installed at actual nodes such as the beginning of the distribution network, the end of the line, and the distributed power supply access point.

[0030] Step 120: Determine the maximum and minimum line lengths of each line in the distribution network. If the ratio between the maximum and minimum line lengths is greater than or equal to a ratio threshold, perform segmented reconstruction of each line in the distribution network by inserting virtual nodes based on the minimum line length and the line lengths of each line in the distribution network to obtain the reconstructed distribution network.

[0031] The ratio threshold can be obtained and preset by the operator based on extensive experience, experiments or statistics, or it can be set by the operator according to actual needs.

[0032] Regarding the value of the ratio threshold, in some embodiments, this application preferably sets the ratio threshold to 10.

[0033] Step 130: Determine the node impedance matrix of the nodes in the reconstructed distribution network, where the nodes include real nodes and virtual nodes.

[0034] Regarding the determination of the node impedance matrix, in some embodiments, existing methods for determining the node impedance matrix can be used, which will not be elaborated here.

[0035] Step 140: Use the voltage drop vector as the observation vector, the submatrix of the node impedance matrix as the sensing matrix, and use the sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the sensing matrix.

[0036] Among them, the sparse Bayesian learning algorithm is the SBL algorithm, which is the Sparse Bayesian Learning algorithm.

[0037] In some embodiments, the M rows of elements in the node impedance matrix can be taken as submatrices of the node impedance matrix, where M is the total number of elements in the voltage drop vector or observation vector.

[0038] Regarding the determination of the node fault current source vector, in some embodiments, a sparse Bayesian learning algorithm can be used to iteratively update the posterior mean vector, sparse hyperparameter vector, and noise variance based on the observation vector and the perception matrix, until the quotient between the square root of the sparse hyperparameter vector and the noise variance is less than or equal to the iteration threshold, or the number of iterations is equal to the preset maximum number of iterations. The posterior mean vector is then used as the node fault current source vector. The iteration threshold and the preset maximum number of iterations can both be obtained and preset by the operator based on extensive experience, experiments, or statistics. Of course, they can also be set by the operator according to actual needs.

[0039] Regarding the selection of sparse Bayesian learning algorithms, in some embodiments, existing sparse Bayesian learning algorithms or improved sparse Bayesian learning algorithms can be used, which will not be elaborated here.

[0040] Step 150: Determine the fault location result of the distribution network based on the node fault current source vector.

[0041] It should be noted that when a distribution network fault occurs, the voltage of all nodes changes, and the drop in node voltage can be regarded as being caused by the virtual injected current source at both ends of the faulty line segment. Therefore, in some embodiments, the fault location result of the distribution network can be determined based on the solved node fault current source vector.

[0042] In this embodiment, the complex topology is adapted by segmented reconstruction through the insertion of virtual nodes, and high-precision positioning is achieved with a small number of measurement devices by combining sparse Bayesian learning algorithm. This effectively overcomes the problems of high dependence on measurement devices and susceptibility to topological complexity and distributed power source fluctuations in traditional methods, significantly improving the accuracy and robustness of fault location, and ensuring the self-healing capability and power supply reliability of the distribution network in complex operating environments.

[0043] In addition to the aforementioned beneficial effects, this distribution network fault location method also has the following advantages: Reduced equipment costs: Traditional methods rely heavily on measuring equipment, requiring a large number of devices to achieve relatively accurate fault location, which undoubtedly increases the costs of equipment procurement, installation, and maintenance. The method in this application, through virtual node insertion, segmented reconstruction, and sparse Bayesian learning algorithms, achieves high-precision fault location with a small number of measuring devices, significantly reducing the number of measuring devices required and thus effectively lowering equipment costs; Improved fault handling efficiency: Rapid and accurate fault location is the core of fault isolation and promoting distribution network self-healing. The method in this application can quickly determine the fault location even with complex topologies and a small number of measuring devices, reducing fault investigation time and allowing maintenance personnel to take faster measures to isolate the fault and restore power to non-faulty areas, thereby improving the overall fault handling efficiency of the distribution network, reducing outage time and scope, and enhancing user satisfaction; Adaptation to distribution network development trends: With the development of new distribution systems, the number of branches and nodes in the distribution network continues to increase, and the topology becomes more complex. The increasing complexity and uncertainty brought about by distributed power source integration have led to a significant increase in volatility. This application's method is specifically designed to address these complexities. It adapts to complex topologies through segmented reconstruction using virtual node insertion and overcomes the impact of distributed power source fluctuations using a sparse Bayesian learning algorithm. This approach effectively adapts to future development trends in distribution networks, providing strong support for the stable operation of new distribution systems. Enhanced system scalability: Based on virtual node insertion and sparse Bayesian learning algorithms, this method offers good flexibility and scalability. When the distribution network scale expands further or the topology changes, only the virtual node insertion strategy and algorithm parameters need to be adjusted according to the new situation, without requiring large-scale modifications to the overall method. This allows for continued high-precision fault location, reducing the difficulty and cost of system upgrades and expansions. Optimized resource allocation: By reducing the configuration requirements of measurement equipment, limited resources can be more rationally allocated to other key aspects of the distribution network, such as equipment maintenance and line upgrades. This helps optimize the overall resource allocation of the distribution network, improve resource utilization efficiency, and further enhance the overall performance and operational quality of the distribution network.

[0044] In one feasible implementation, the sparse Bayesian learning algorithm in the above embodiments is a feature-adaptive sparse Bayesian learning algorithm.

[0045] Step 140 in the above embodiment utilizes a sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the perception matrix. This includes: obtaining the line characteristic factor vector of the node in the reconstructed distribution network; initializing the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance to obtain the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the first iteration; and determining the posterior covariance matrix and posterior covariance matrix in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line characteristic factor vector, observation vector, and perception matrix. The mean vector is initialized to 1. Based on the posterior mean vector and posterior covariance vector of the t-th iteration, the sparse hyperparameter vector of the t-th iteration is updated to obtain the sparse hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the node line feature factor vector, the adaptive hyperparameter vector of the t-th iteration is updated to obtain the adaptive hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the observation vector and perception matrix, the noise variance of the (t+1)-th iteration is determined. If t=1, Let t = t + 1, then return to the step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, observation vector, and perception matrix; if t is not equal to 1, then obtain the weight factors of all nodes in the reconstructed distribution network in the t-th iteration based on the noise variance, posterior mean vector, and posterior covariance matrix in the t-th iteration, as well as the node line feature factor vector and perception matrix; based on the sparse hyperparameter vector, adaptive hyperparameter vector, and weight factors of all nodes in the t-th iteration... The weighted relative change in the t-th iteration is determined by taking the weighted factor, the sparse hyperparameter vector and the adaptive hyperparameter vector in the (t-1)-th iteration, and t. If the weighted relative change in the t-th iteration is less than or equal to the convergence threshold, or t equals the maximum number of iterations, then the posterior mean vector in the t-th iteration is used as the node fault current source vector. Otherwise, t = t + 1 is set, and the process returns to the step of determining the posterior covariance matrix and the posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, the adaptive hyperparameter vector and the noise variance in the t-th iteration, as well as the node line characteristic factor vector, the observation vector and the perception matrix.

[0046] The convergence threshold and the maximum number of iterations can both be preset by the operator based on extensive experience, experiments, or statistics. Alternatively, they can be set by the operator according to actual needs.

[0047] It should be noted that the feature-adaptive sparse Bayesian learning algorithm is the improved sparse Bayesian learning algorithm of this application.

[0048] Regarding the method of obtaining the line characteristic factor vector, in some embodiments, the line characteristic factor vector can be determined based on the line length of each line in the reconstructed distribution network.

[0049] Regarding the initialization methods of sparse hyperparameter vectors, adaptive hyperparameter vectors, and noise variance, in some embodiments, the sparse hyperparameter vectors, adaptive hyperparameter vectors, and noise variance can be initialized to their respective default values, or they can be initialized to random values, or they can be initialized according to the observation vector and / or the perception matrix to obtain the sparse hyperparameter vectors, adaptive hyperparameter vectors, and noise variance in the first iteration.

[0050] In this embodiment, the accuracy and adaptability of fault location are further improved by using the feature-adaptive sparse Bayesian learning algorithm, which enhances the performance of the algorithm in complex power distribution network environments.

[0051] Understandably, this improves positioning accuracy: by introducing line feature factor vectors, the algorithm can more accurately capture the unique characteristics of different lines. During the iteration process, it updates the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance based on these characteristics, making the final determined node fault current source vector more accurate, thus significantly improving the accuracy of distribution network fault location and reducing misjudgments. It also enhances adaptability: the algorithm can adaptively adjust parameters according to the actual situation of each line in the distribution network. During the iteration process, it dynamically updates the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance based on the line feature factor vector and the posterior mean vector and posterior covariance matrix of each iteration. By adapting to hyperparameter vectors and noise variance, the algorithm can better adapt to the complex and ever-changing topology of the distribution network and the uncertainties and fluctuations brought about by the access of distributed power sources, achieving relatively accurate fault location in different scenarios. The optimized iteration process, by determining the weight factors and weighted relative changes of all nodes in each iteration, allows the algorithm to more scientifically determine whether the iteration has reached the convergence condition. Iteration stops when the weighted relative change is less than or equal to the convergence threshold, or when the maximum number of iterations is reached, avoiding unnecessary calculations, improving the algorithm's computational efficiency, and ensuring the reliability of the results. This enables faster and more accurate fault location even with a limited number of measuring devices.

[0052] In one feasible implementation, the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance at the (t+1)th iteration are obtained using the following formulas, as well as the weighted relative change at the tth iteration: ; in, Let i be the i-th element in the sparse hyperparameter vector of the (t+1)-th iteration. and These are the first preset prior hyperparameter and the second preset prior hyperparameter, respectively. Let i be the i-th element in the posterior mean vector of the t-th iteration. Let i be the i-th diagonal element in the posterior covariance matrix of the t-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the (t+1)-th iteration. and These are the third and fourth preset prior hyperparameters, respectively. This is the i-th element in the node line feature factor vector. Let Variance be the noise variance in the (t+1)th iteration. For the observation vector, For the perception matrix, Let be the posterior mean vector of the t-th iteration. It is the L2 norm. This is a function to sum the diagonal elements of a matrix. It is the transpose symbol. Let be the posterior covariance matrix of the t-th iteration. The total number of elements in the observation vector. Let be the weighted relative change in the t-th iteration. This represents the total number of nodes in the reconstructed distribution network. Let be the weight factor of the i-th node in the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the (t-1)-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the t-th iteration. Let be the i-th element in the adaptive hyperparameter vector during the (t-1)-th iteration.

[0053] It should be noted that the first, second, third, and fourth preset prior hyperparameters can all be obtained and preset by the operator based on extensive experience, experiments, or statistics. Of course, they can also be set by the operator according to actual needs.

[0054] In the embodiments of this application, sparse hyperparameter vectors, adaptive hyperparameter vectors, and noise variance iterative updates and weighted relative changes are determined through specific formulas, which significantly improves fault location accuracy, enhances algorithm adaptability, and optimizes the iteration process.

[0055] Understandably, improving fault location accuracy involves incorporating elements such as the posterior mean vector and posterior covariance matrix into the formula. When calculating the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the (t+1)th iteration, it fully considers the information changes during each iteration. By accurately capturing these changes, it can more accurately determine the node fault current source vector, thereby significantly improving the accuracy of distribution network fault location and effectively reducing misjudgments. Furthermore, enhancing algorithm adaptability involves considering the node line feature factor vector in the formula. This vector reflects the unique characteristics of each line in the distribution network. During iteration, the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance are updated specifically based on these characteristics, making the algorithm more adaptable. The algorithm can adaptively adjust to the actual conditions of each line in the distribution network. This allows it to better adapt to the complex and ever-changing topology of the distribution network and the uncertainties and fluctuations brought about by the access of distributed power sources. It can achieve relatively accurate fault location in different scenarios. The optimization of the iteration process is achieved by determining the weight factors and weighted relative changes of all nodes in each iteration through formulas. This provides a scientific basis for judging whether the iteration has reached the convergence condition. The iteration stops when the weighted relative change is less than or equal to the convergence threshold or when the maximum number of iterations is reached. This avoids unnecessary calculations, improves the computational efficiency of the algorithm, and ensures the reliability of the results. This enables faster and more accurate fault location with a small number of measuring devices.

[0056] In one feasible implementation, the process of obtaining the weight factors of all nodes in the reconstructed distribution network in the t-th iteration based on the noise variance, posterior mean vector, and posterior covariance matrix in the t-th iteration, as well as the node line feature factor vector and the sensing matrix, includes: determining the signal-to-noise ratio of all nodes in the t-th iteration based on the noise variance, posterior mean vector, and sensing matrix in the t-th iteration; determining the fault probability of all nodes in the t-th iteration based on the posterior mean vector and posterior covariance matrix in the t-th iteration; and determining the weight factors of all nodes in the t-th iteration based on the signal-to-noise ratio and fault probability of all nodes in the t-th iteration, as well as the node line feature factor vector.

[0057] In some embodiments, the signal-to-noise ratio, failure probability, and weighting factor of all nodes can be determined using the following formula: ; in, Let be the signal-to-noise ratio of the i-th node in the t-th iteration. For the element in the i-th column of the perception matrix, It is the transpose symbol. Let be the posterior mean vector of the t-th iteration. Let Variance be the noise variance in the t-th iteration. Let be the failure probability of the i-th node in the t-th iteration. For natural numbers, Let i be the i-th element in the posterior mean vector of the t-th iteration. Let i be the i-th diagonal element in the posterior covariance matrix of the t-th iteration. Let be the weight factor of the i-th node in the t-th iteration. This represents the total number of nodes in the reconstructed distribution network. It is the i-th element in the node line feature factor vector.

[0058] In this embodiment, by scientifically determining the weight factors of all nodes, the fault location accuracy is significantly improved, the algorithm adaptability is enhanced, and the iteration process is optimized.

[0059] Understandably, improving fault location accuracy involves considering factors such as signal-to-noise ratio (SNR), fault probability, and node-line characteristic factor vectors when determining node weighting factors. SNR reflects the quality of the node signal, fault probability indicates the likelihood of a fault occurring at the node, and node-line characteristic factor vectors reflect the unique characteristics of the line. By combining these factors to determine the weighting factors, the fault current source vector at the node can be more accurately determined, thus significantly improving the accuracy of distribution network fault location and effectively reducing misjudgments. Furthermore, enhancing algorithm adaptability is crucial. The node-line characteristic factor vector reflects the unique characteristics of each line in the distribution network. Considering this vector when determining the weighting factors allows the algorithm to adapt to different network conditions. The algorithm adaptively adjusts to the actual conditions of each line in the power grid, enabling it to better adapt to the complex and ever-changing topology of the distribution network and the uncertainties and fluctuations brought about by the access of distributed power sources. This allows for relatively accurate fault location in different scenarios. The optimization of the iteration process: determining the node weight factors provides a basis for subsequent calculation of the weighted relative change. The weighted relative change allows for a more scientific determination of whether the iteration has reached the convergence condition. Iteration stops when the weighted relative change is less than or equal to the convergence threshold, or when the maximum number of iterations is reached. This avoids unnecessary calculations, improves the algorithm's computational efficiency, and ensures the reliability of the results, enabling faster and more accurate fault location even with a limited number of measuring devices.

[0060] In one feasible implementation, determining the posterior covariance matrix and posterior mean vector in the above embodiment based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, observation vector, and perception matrix, includes: determining the hyperparameter diagonal matrix in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and node line feature factor vector in the t-th iteration; determining the posterior covariance matrix in the t-th iteration based on the hyperparameter diagonal matrix, noise variance, and perception matrix in the t-th iteration; and determining the posterior mean vector in the t-th iteration based on the noise variance, observation vector, and perception matrix in the t-th iteration.

[0061] Regarding the determination of the hyperparameter diagonal matrix, posterior covariance matrix, and posterior mean vector in the t-th iteration, in some embodiments, the following formula can be used to determine the hyperparameter diagonal matrix, posterior covariance matrix, and posterior mean vector in the t-th iteration: ; in, Let be the hyperparameter diagonal matrix in the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the t-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the t-th iteration. This is the i-th element in the node line feature factor vector. This represents the total number of nodes in the reconstructed distribution network. Let be the posterior covariance matrix of the t-th iteration. Let Variance be the noise variance in the t-th iteration. For the perception matrix, It is the transpose symbol. Let be the posterior mean vector of the t-th iteration. This is the observation vector.

[0062] In this embodiment, the hyperparameter diagonal matrix is ​​determined by comprehensively considering sparse hyperparameter vectors, adaptive hyperparameter vectors, and node line feature factor vectors. This improves the accuracy and adaptability of the calculation of the posterior covariance matrix and posterior mean vector, lays the foundation for the iterative optimization of the fault location algorithm, and enhances the performance of the algorithm in complex power distribution network environments.

[0063] Understandably, this improves computational accuracy: by comprehensively considering sparse hyperparameter vectors, adaptive hyperparameter vectors, and node line feature factor vectors to determine the hyperparameter diagonal matrix, the parameter characteristics of different lines can be more accurately characterized. This provides a more accurate basis for calculating the posterior covariance matrix and posterior mean vector, making the calculation results more consistent with the actual operation of the distribution network and effectively improving fault location accuracy. It also enhances algorithm adaptability: the node line feature factor vectors reflect the unique characteristics of each line. Incorporating these characteristics when determining the hyperparameter diagonal matrix allows the algorithm to adaptively adjust the parameter calculation method according to the actual situation of different lines, better adapting to the complex and ever-changing topology of the distribution network and the uncertainties and fluctuations brought about by distributed power source access. This ensures accurate calculation of posterior parameters in different scenarios. Finally, it optimizes the iteration process: accurately calculating the posterior covariance matrix and posterior mean vector is a key step in algorithm iteration. The results obtained based on the above methods provide a reliable basis for subsequent parameter updates and iterative judgments, ensuring that the iteration process proceeds in the correct direction, avoiding iteration deviations caused by inaccurate parameter calculations, improving algorithm efficiency, and ensuring rapid high-precision fault location with a limited number of measuring devices.

[0064] In one feasible implementation, obtaining the line characteristic factor vector of a node in the reconstructed distribution network in the above embodiments includes: determining the line characteristic factor of each node based on the line length of all adjacent line segments of each node in the reconstructed distribution network; normalizing the line characteristic factor of each node to obtain the standard line characteristic factor of each node; and constructing a node line characteristic factor vector based on the standard line characteristic factors of each node.

[0065] For the line characteristic factor of each node, and the method for determining the standard line characteristic factor of each node, in some embodiments, the line characteristic factor of each node and the standard line characteristic factor of each node can be determined using the following formula: ; in, Let be the line characteristic factor of the i-th node. Let be the total number of adjacent line segments of the i-th node. Let be the length of the j-th adjacent line segment of the i-th node. This refers to the i-th element in the standard line characteristic factor or the node line characteristic factor vector of the i-th node. It is the minimum line characteristic factor among all the line characteristic factors of all nodes. It is the largest line characteristic factor among all the line characteristic factors of all nodes.

[0066] In this embodiment, the line feature factors are determined by comprehensively considering the lengths of all adjacent line segments of each node, and a normalization process is performed to construct a line feature factor vector. This improves the fault location accuracy and algorithm adaptability, optimizes the line feature representation method, and provides a reliable foundation for subsequent calculations.

[0067] Understandably, improving fault location accuracy involves several key aspects: First, determining line characteristic factors by comprehensively considering the lengths of all adjacent line segments at each node and then normalizing them to construct a line characteristic factor vector. This accurately reflects the characteristics of each node's lines, leading to more accurate determination of the fault current source vector in subsequent calculations, reducing misjudgments, and improving location accuracy. Second, enhancing algorithm adaptability: The node line characteristic factor vector considers the actual conditions of each node's lines. During algorithm iteration, parameters are dynamically adjusted based on this vector, enabling the algorithm to better adapt to the complex topology of the distribution network and the uncertainties and fluctuations brought about by distributed power source access, achieving accurate fault location in different scenarios. Third, optimizing line feature representation: First, determining the line characteristic factors for each node, then normalizing them to obtain standard line characteristic factors, and finally constructing a line characteristic factor vector. This approach makes line feature representation more scientific and reasonable, providing a reliable foundation for subsequent calculations of hyperparameter diagonal matrices, posterior covariance matrices, and posterior mean vectors based on this line characteristic factor vector, ensuring the smooth progress of the algorithm iteration process.

[0068] In one feasible implementation, step 120 in the above embodiment, which involves segmenting and reconstructing each line in the distribution network by inserting virtual nodes based on the minimum line length and the line length of each line in the distribution network, to obtain the reconstructed distribution network, includes: determining the total number of virtual nodes to be inserted into each line in the distribution network based on the line length of each line in the distribution network and the minimum line length; and segmenting and reconstructing each line in the distribution network by inserting virtual nodes according to the length interval of the minimum line length and the total number of virtual nodes to be inserted into each line in the distribution network, to obtain the reconstructed distribution network.

[0069] In some embodiments, the total number of virtual nodes to be inserted for each line in a distribution network can be determined using a formula. Determine the total number of virtual nodes required for each line in the distribution network; where, Let be the total number of virtual nodes required to be inserted for the j-th line in the distribution network. This is the floor function. Let j be the length of the j-th line in the distribution network. This is the minimum line length.

[0070] In this embodiment, by determining the total number of virtual nodes to be inserted based on the length of each line and the minimum line length, and inserting them at fixed length intervals, the fault location accuracy and algorithm adaptability can be improved, effectively addressing complex power distribution network topologies.

[0071] Understandably, this improves fault location accuracy: by determining the total number of virtual nodes to be inserted based on the length of each line and the minimum line length, and inserting them at fixed length intervals, it can accurately adapt to different line lengths, making the reconstructed distribution network more adaptable to fault location. This provides an accurate foundation for subsequent calculations based on node impedance matrices to determine the node fault current source vector, reducing location errors caused by complex topology structures, thereby improving fault location accuracy. It also enhances algorithm adaptability: this segmented reconstruction method for inserting virtual nodes can be flexibly applied to distribution networks of different scales and topologies. Regardless of differences in line length, it can reasonably insert virtual nodes, enabling the algorithm to better adapt to the complex and ever-changing topology of the distribution network and the uncertainties and fluctuations brought about by distributed power source access, achieving relatively accurate fault location in different scenarios.

[0072] In one feasible implementation, the method in the above embodiments further includes: when the ratio between the maximum line length and the minimum line length is less than a ratio threshold, determining the target line length based on the line length of each line in the distribution network, and performing segmented reconstruction of each line in the distribution network by inserting virtual nodes based on the target line length, thereby obtaining the reconstructed distribution network.

[0073] Regarding the method for determining the target line length, in some embodiments, the average value between the line lengths of each line in the distribution network can be used as the target line length. Alternatively, the maximum and minimum line lengths of each line in the distribution network can be determined, and then the average value between the maximum and minimum line lengths can be used as the target line length.

[0074] Of course, in other embodiments, a preset line length can be used as the target line length; wherein, the preset line length can be obtained and preset by the operator based on a large amount of experience, experimentation or statistics, or it can be set by the operator according to actual needs.

[0075] Regarding the determination method of the reconstructed distribution network, in some embodiments, the total number of virtual nodes to be inserted for each line in the distribution network can be determined based on the line length of each line in the distribution network and the target line length. Then, according to the length interval of the target line length, the virtual node insertion of each line in the distribution network is segmented and reconstructed according to the total number of virtual nodes to be inserted for each line in the distribution network, so as to obtain the reconstructed distribution network.

[0076] In this embodiment, when the ratio of the maximum line length to the minimum line length is less than a ratio threshold, the target line length is reasonably determined, and virtual node insertion and segmentation reconstruction are performed on each line accordingly. This improves the fault location accuracy and enhances the algorithm's adaptability, effectively addressing complex power distribution network topologies.

[0077] Understandably, this improves fault location accuracy: when the ratio of the maximum line length to the minimum line length is less than a threshold, by reasonably determining the target line length and then performing segmented reconstruction by inserting virtual nodes into each line, it can accurately adapt to distribution networks with small differences in line length. This makes the reconstructed distribution network more suitable for fault location, providing an accurate basis for subsequent calculations based on node impedance matrices to determine the node fault current source vector, reducing location errors caused by complex topology, and thus improving fault location accuracy. It also enhances algorithm adaptability: this segmented reconstruction method for inserting virtual nodes can flexibly determine the target line length according to the line length of different distribution networks. Whether using the average line length, the average of the maximum and minimum line lengths, or a preset line length, it can reasonably insert virtual nodes, enabling the algorithm to better adapt to the complex and ever-changing topology of the distribution network and the uncertainty and volatility brought about by distributed power source access. This allows for relatively accurate fault location in different scenarios.

[0078] In one feasible implementation, step 110 in the above embodiment, obtaining the voltage drop vector before and after the distribution network fault, includes: obtaining the voltage amplitude of each measuring device in the distribution network before the fault and the voltage amplitude after the fault; constructing a first voltage amplitude vector based on the voltage amplitude of each measuring device in the distribution network before the fault; constructing a second voltage amplitude vector based on the voltage amplitude of each measuring device in the distribution network after the fault; and determining the voltage drop vector based on the first voltage amplitude vector and the second voltage amplitude vector.

[0079] Regarding the method of determining the voltage drop vector, in some embodiments, the absolute value of the difference between the i-th element in the first voltage amplitude vector and the i-th element in the second voltage amplitude vector can be used as the i-th element in the voltage drop vector, where i takes the value of an integer greater than 0 in sequence until i is greater than M.

[0080] In this embodiment of the application, by obtaining the voltage amplitude of each measuring device before and after the fault and constructing the first voltage amplitude vector and the second voltage amplitude vector to calculate the voltage drop vector, the fault location accuracy and reliability can be improved, and the adaptability of the algorithm to complex power distribution networks can be enhanced.

[0081] Understandably, this improves fault location accuracy: by acquiring the voltage amplitude of each measuring device before and after the fault and constructing a first voltage amplitude vector and a second voltage amplitude vector to calculate the voltage drop vector, the voltage changes before and after the fault can be accurately captured. This provides accurate basic data for subsequent calculations based on node impedance matrices to determine the node fault current source vector, thereby improving fault location accuracy. It also enhances algorithm reliability: this method of acquiring the voltage drop vector is not overly restricted by the installation location and number of measuring devices. As long as a certain number of measuring devices acquire the voltage amplitude before and after the fault, the first voltage amplitude vector and the second voltage amplitude vector can be constructed to calculate the voltage drop vector. This allows the algorithm to run reliably under different measuring device configurations, enhancing the algorithm's adaptability to complex power distribution networks.

[0082] In one feasible implementation, step 150 in the above embodiment, which determines the fault location result of the distribution network based on the node fault current source vector, includes: when the number of non-zero elements in the node fault current source vector is not equal to 1, determining the largest and second largest non-zero elements in the node fault current source vector, and taking the line segment between the node corresponding to the largest non-zero element and the node corresponding to the second largest non-zero element as the fault location result; when the number of non-zero elements in the node fault current source vector is equal to 1, taking the adjacent line segment of the node corresponding to the non-zero element in the node fault current source vector as the fault location result.

[0083] In this embodiment of the application, by distinguishing between the different cases where the number of non-zero elements in the node fault current source vector is equal to 1 and not equal to 1, different fault location strategies are adopted, which can improve the accuracy and reliability of fault location and enhance the adaptability of the algorithm to complex fault scenarios.

[0084] Understandably, this method improves fault location accuracy by differentiating between cases where the number of non-zero elements in the node fault current source vector is equal to 1 and not equal to 1, and by employing different fault location strategies. This allows for more precise determination of the fault location, reducing misjudgments and omissions. Furthermore, it enhances algorithm reliability by considering potential scenarios in actual distribution networks. Through flexible selection of fault location strategies, the algorithm adapts to different scenarios, improving its robustness and reliability, and providing strong support for the stable operation of the distribution network.

[0085] In a second aspect, this application provides a power distribution network fault location device.

[0086] Please see Figure 2 This is a schematic diagram of a power distribution network fault location device according to an embodiment of this application. The device 210 includes: The acquisition module 211 is used to acquire the voltage drop vector before and after the distribution network fault occurs, as well as the line length of each line in the distribution network; The determination and reconstruction module 212 is used to determine the maximum and minimum line lengths of each line in the distribution network. When the ratio between the maximum and minimum line lengths is greater than or equal to a ratio threshold, the module performs segmented reconstruction of each line in the distribution network by inserting virtual nodes based on the minimum line length and the line lengths of each line in the distribution network, thereby obtaining the reconstructed distribution network. The first determining module 213 is used to determine the node impedance matrix of the nodes in the reconstructed distribution network, wherein the nodes include real nodes and virtual nodes; The determination module 214 is used to take the voltage drop vector as the observation vector, take the submatrix of the node impedance matrix as the sensing matrix, and use the sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the sensing matrix. The second determining module 215 is used to determine the fault location result of the distribution network based on the node fault current source vector.

[0087] In this embodiment, the relevant contents of the above-mentioned acquisition module 211, determination and reconstruction module 212, first determination module 213, assignment determination module 214 and second determination module 215 can be found in the following references. Figure 1 The contents of the illustrated embodiments will not be repeated here.

[0088] It should be noted that the device 210 of this application also includes other modules. It can be understood that the method of this application and the device 210 have a one-to-one correspondence. Therefore, the other modules of the device 210 of this application are the contents corresponding to the method of this application in the above embodiments.

[0089] In this embodiment, by inserting virtual nodes and reconstructing segments to adapt to complex topologies, and combining sparse Bayesian learning algorithms, high-precision positioning is achieved with a small number of measurement devices. This effectively overcomes the problems of traditional devices being highly dependent on measurement devices and easily affected by topological complexity and distributed power source fluctuations. It significantly improves the accuracy and robustness of fault location and ensures the self-healing capability and power supply reliability of the distribution network in complex operating environments.

[0090] In addition to the aforementioned beneficial effects, this power distribution network fault location device also has the following advantages: Reduced equipment costs: Traditional devices rely heavily on measuring equipment, requiring a large number of measuring devices to achieve relatively accurate fault location. This undoubtedly increases the costs of equipment procurement, installation, and maintenance. The device in this application, through virtual node insertion, segmented reconstruction, and sparse Bayesian learning algorithms, can achieve high-precision fault location with a small number of measuring devices, significantly reducing the number of measuring devices required and thus effectively reducing equipment costs; Improved fault handling efficiency: Rapid and accurate fault location is the core of fault isolation and promoting power distribution network self-healing. The device in this application can quickly determine the fault location even with complex topologies and a small number of measuring devices, reducing fault investigation time. This allows maintenance personnel to take measures to isolate the fault and restore power to non-faulty areas more quickly, thereby improving the overall fault handling efficiency of the power distribution network, reducing outage time and scope, and increasing user satisfaction; Adaptation to power distribution network development trends: With the development of new power distribution systems, the number of branches and nodes in the power distribution network continues to increase, and the topology becomes more... The increasing complexity and uncertainty brought about by distributed power source access have led to the design of this device specifically for these complex situations. It adapts to complex topologies through segmented reconstruction using virtual node insertion and overcomes the impact of distributed power source fluctuations using a sparse Bayesian learning algorithm. This allows it to well adapt to the future development trend of distribution networks and provides strong support for the stable operation of new distribution systems. Enhanced system scalability: Based on virtual node insertion and sparse Bayesian learning algorithms, this device offers good flexibility and scalability. When the distribution network scale expands further or the topology changes, only the virtual node insertion strategy and algorithm parameters need to be adjusted according to the new situation, without requiring large-scale modifications to the overall device, to continue achieving high-precision fault location. This reduces the difficulty and cost of system upgrades and expansions. Optimized resource allocation: By reducing the configuration requirements of measurement equipment, limited resources can be more rationally allocated to other key aspects of the distribution network, such as equipment maintenance and line upgrades. This helps to optimize the overall resource allocation of the distribution network, improve resource utilization efficiency, and further enhance the overall performance and operational quality of the distribution network.

[0091] In a third aspect, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform a power distribution network fault location method as described in any of the first aspects.

[0092] This application provides a computer device in a fourth aspect, including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform a power distribution network fault location method as described in any of the first aspects.

[0093] Figure 3The diagram illustrates the internal structure of a computer device in some embodiments. This computer device may specifically be a terminal, a server, or a gateway. Figure 3 As shown, the computer device includes a processor, memory, and network interface connected via a system bus.

[0094] The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When executed by a processor, this computer program causes the processor to perform the steps in the above method embodiments. The internal memory may also store a computer program, which, when executed by a processor, causes the processor to perform the steps in the above method embodiments. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods.

[0096] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0097] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0098] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

Claims

1. A method for locating faults in a power distribution network, characterized in that, The method includes: Obtain the voltage drop vector before and after the distribution network fault occurs, as well as the line length of each line in the distribution network; The maximum and minimum line lengths of each line in the distribution network are determined. If the ratio between the maximum and minimum line lengths is greater than or equal to a ratio threshold, the distribution network is reconstructed by inserting virtual nodes based on the minimum line length and the line lengths of each line in the distribution network to obtain the reconstructed distribution network. Determine the node impedance matrix of the nodes in the reconstructed distribution network, wherein the nodes include real nodes and virtual nodes; The voltage drop vector is used as the observation vector, the submatrix of the node impedance matrix is ​​used as the sensing matrix, and the sparse Bayesian learning algorithm is used to determine the node fault current source vector based on the observation vector and the sensing matrix. Based on the node fault current source vector, the fault location result of the distribution network is determined.

2. The method for locating faults in a power distribution network according to claim 1, characterized in that, The sparse Bayesian learning algorithm is a feature-adaptive sparse Bayesian learning algorithm. The step of using the sparse Bayesian learning algorithm to determine the node fault current source vector based on the observation vector and the sensing matrix includes: Obtain the line characteristic factor vector of the nodes in the reconstructed distribution network; The sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance are initialized to obtain the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the first iteration. Based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix, the posterior covariance matrix and posterior mean vector in the t-th iteration are determined, where the initial value of t is 1. Based on the posterior mean vector and posterior covariance vector of the t-th iteration, the sparse hypervector of the t-th iteration is updated to obtain the sparse hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the node line feature factor vector, the adaptive hyperparameter vector of the t-th iteration is updated to obtain the adaptive hyperparameter vector of the (t+1)-th iteration. Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, and the observation vector and the perception matrix, the noise variance of the (t+1)-th iteration is determined. If t=1, let t=t+1, then return to the step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix. If t is not equal to 1, then based on the noise variance, posterior mean vector and posterior covariance matrix in the t-th iteration, as well as the node line feature factor vector and the perception matrix, the weight factors of all nodes of the reconstructed distribution network in the t-th iteration are obtained. Based on the sparse hyperparameter vector, adaptive hyperparameter vector, and weight factors of all nodes in the t-th iteration, and the sparse hyperparameter vector and adaptive hyperparameter vector in the (t-1)-th iteration, determine the weighted relative change in the t-th iteration. If the weighted relative change in the t-th iteration is less than or equal to the convergence threshold, or if t equals the maximum number of iterations, then the posterior mean vector in the t-th iteration will be used as the node fault current source vector. Otherwise, let t=t+1, and return to the step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix.

3. The method for locating faults in a power distribution network according to claim 2, characterized in that, The sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance at the (t+1)th iteration are obtained using the following formulas, as well as the weighted relative change at the tth iteration: ； in, Let i be the i-th element in the sparse hyperparameter vector of the (t+1)-th iteration. and These are the first preset prior hyperparameter and the second preset prior hyperparameter, respectively. Let i be the i-th element in the posterior mean vector of the t-th iteration. Let i be the i-th diagonal element in the posterior covariance matrix of the t-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the (t+1)-th iteration. and These are the third and fourth preset prior hyperparameters, respectively. Let i be the i-th element in the node line feature factor vector. Let Variance be the noise variance in the (t+1)th iteration. Let be the observation vector. The perception matrix is... Let be the posterior mean vector of the t-th iteration. It is the L2 norm. This is a function to sum the diagonal elements of a matrix. It is the transpose symbol. Let be the posterior covariance matrix of the t-th iteration. The total number of elements in the observation vector. Let be the weighted relative change in the t-th iteration. This refers to the total number of nodes in the reconstructed distribution network. Let be the weight factor of the i-th node in the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the t-th iteration. Let i be the i-th element in the sparse hyperparameter vector of the (t-1)-th iteration. Let i be the i-th element in the adaptive hyperparameter vector of the t-th iteration. Let be the i-th element in the adaptive hyperparameter vector during the (t-1)-th iteration.

4. The method for locating faults in a power distribution network according to claim 2, characterized in that, The weight factors for all nodes of the reconstructed distribution network in the t-th iteration are obtained based on the noise variance, posterior mean vector, and posterior covariance matrix in the t-th iteration, as well as the node line feature factor vector and the perception matrix, including: Based on the noise variance and posterior mean vector in the t-th iteration, and the perception matrix, determine the signal-to-noise ratio of all nodes in the t-th iteration; Based on the posterior mean vector and posterior covariance matrix of the t-th iteration, determine the failure probability of all nodes in the t-th iteration. The weighting factors for all nodes in the t-th iteration are determined based on the signal-to-noise ratio and fault probability of all nodes in the t-th iteration, as well as the node line feature factor vector.

5. The method for locating faults in a power distribution network according to claim 2, characterized in that, The step of determining the posterior covariance matrix and posterior mean vector in the t-th iteration based on the sparse hyperparameter vector, adaptive hyperparameter vector, and noise variance in the t-th iteration, as well as the node line feature factor vector, the observation vector, and the perception matrix, includes: Based on the sparse hyperparameter vector and adaptive hyperparameter vector in the t-th iteration, and the node line feature factor vector, determine the hyperparameter diagonal matrix in the t-th iteration; Based on the hyperparameter diagonal matrix and noise variance in the t-th iteration, and the perception matrix, determine the posterior covariance matrix in the t-th iteration. Based on the noise variance in the t-th iteration, the observation vector, and the perception matrix, the posterior mean vector in the t-th iteration is determined.

6. The method for locating faults in a power distribution network according to claim 2, characterized in that, The step of obtaining the line feature factor vector of the nodes in the reconstructed distribution network includes: Based on the line lengths of all adjacent line segments of each node in the reconstructed distribution network, determine the line characteristic factor of each node; The line characteristic factors of each node are normalized to obtain the standard line characteristic factors of each node. Based on the standard line characteristic factors of each node, construct the node line characteristic factor vector.

7. The method for locating faults in a power distribution network according to claim 1, characterized in that, The step of segmenting and reconstructing the distribution network by inserting virtual nodes into each line based on the minimum line length and the line length of each line in the distribution network to obtain the reconstructed distribution network includes: Based on the length of each line in the distribution network and the minimum line length, determine the total number of virtual nodes that need to be inserted for each line in the distribution network. Based on the minimum line length interval, and according to the total number of virtual nodes required to be inserted for each line in the distribution network, the distribution network is reconstructed in segments by inserting virtual nodes, thus obtaining the reconstructed distribution network.

8. The method for locating faults in a power distribution network according to claim 1, characterized in that, The method further includes: If the ratio between the maximum line length and the minimum line length is less than the ratio threshold, the target line length is determined based on the line length of each line in the distribution network, and the segmented reconstruction of each line in the distribution network is performed by inserting virtual nodes based on the target line length to obtain the reconstructed distribution network.

9. The method for locating faults in a distribution network according to claim 1, characterized in that, The process of obtaining the voltage drop vector before and after a distribution network fault includes: Obtain the voltage amplitude of each measuring device in the power distribution network before and after the fault occurs; A first voltage amplitude vector is constructed based on the voltage amplitude of each measuring device in the power distribution network before the fault occurs. A second voltage amplitude vector is constructed based on the voltage amplitude of each measuring device in the power distribution network after a fault occurs. The voltage drop vector is determined based on the first voltage amplitude vector and the second voltage amplitude vector.

10. The method for locating faults in a power distribution network according to claim 1, characterized in that, The step of determining the fault location result of the distribution network based on the node fault current source vector includes: When the number of non-zero elements in the node fault current source vector is not equal to 1, determine the largest non-zero element and the second largest non-zero element in the node fault current source vector, and take the line segment between the node corresponding to the largest non-zero element and the node corresponding to the second largest non-zero element as the fault location result. When the number of non-zero elements in the node fault current source vector is equal to 1, the adjacent line segment of the node corresponding to the non-zero element in the node fault current source vector is taken as the fault location result.

Images