Power distribution network fault intelligent positioning and self-healing recombination method and system

Abstract

The invention provides a power distribution network fault intelligent positioning and self-healing recombination method and system. The method comprises the steps that S1, operation data of a power distribution network are acquired and preprocessed; s2, modeling the power distribution network topology into a weighted undirected graph, and constructing transient characteristics including node characteristics and edge characteristics; s3, on the basis of the transient characteristics and the power distribution network topology, the fault type is judged in real time by adopting a graph convolutional network, and a fault section is determined in combination with a traveling wave distance measurement and impedance method; s4, when entering an isolation fault state, automatically generating a power supply reconstruction scheme meeting current, voltage and switch constraints, and selecting an optimal scheme through multi-objective optimization; and S5, issuing a switching instruction according to the optimal scheme to complete topology reconstruction, and updating a fault identification and recombination strategy according to an actual execution result to improve positioning precision and recovery efficiency. According to the invention, a novel fault processing scheme integrating multi-source measurement, an efficient algorithm and online learning can be adopted, so that the power failure time is shortened, the power failure range is reduced, and the toughness of the power distribution network is improved.

Description

Technical Field

[0001] This invention relates to the field of distribution network automation and intelligent operation and maintenance technology, and in particular to a method and system for intelligent fault location and self-healing reconfiguration in distribution networks. Background Technology

[0002] In existing distribution networks, fault isolation and power restoration typically rely on fixed protection settings and manual intervention when a fault occurs. This approach has several drawbacks: limited location accuracy (using traditional impedance ranging or traveling wave ranging alone is susceptible to grounding methods and system parameter errors); slow recovery speed (requiring manual calculation and verification of reconfiguration schemes, which is insufficient to meet the high reliability requirements of large-scale distributed power generation); and a lack of self-learning capabilities (processing experience cannot be automatically accumulated, making it difficult to continuously optimize strategies based on historical operating data). Therefore, a novel fault handling solution that integrates multi-source measurement, efficient algorithms, and online learning is urgently needed to shorten outage time, reduce outage area, and improve the resilience of the distribution network. Summary of the Invention

[0003] The purpose of this invention is to propose a method and system for intelligent fault location and self-healing reorganization in power distribution networks, which further reduces the location error, further reduces non-power supply load, and further improves the fault identification accuracy.

[0004] To achieve the above objectives, the technical solution of the present invention is: a method for intelligent fault location and self-healing reconfiguration in a distribution network, the method comprising:

[0005] Step S1: Obtain and preprocess the operating data of the distribution network, including voltage, current, phase angle and switch status;

[0006] Step S2: Model the distribution network topology as a weighted undirected graph and construct transient features including node features and edge features;

[0007] Step S3: Based on transient characteristics and distribution network topology, a graph convolutional network is used to determine the fault type in real time, and the fault section is determined by combining traveling wave ranging and impedance method;

[0008] Step S4: When entering the isolated fault state, automatically generate a power supply reconfiguration scheme that meets the current, voltage and switching constraints, and select the optimal scheme through multi-objective optimization;

[0009] Step S5: Issue a switch command according to the optimal solution to complete the topology reconstruction, and update the fault identification and reassembly strategy according to the actual execution results to improve positioning accuracy and recovery efficiency.

[0010] Preferably, step S2 specifically comprises:

[0011] Based on the primary wiring diagram of the distribution network, a weighted undirected graph is established with busbars and sectionalizing switches as nodes and lines or transformers as edges. The edge weights are taken as the equivalent impedance of the lines.

[0012] The node feature vector includes the three-phase voltage amplitude, phase angle difference, and 25th harmonic amplitude;

[0013] The edge eigenvectors include the line impedance magnitude and power flow direction.

[0014] Preferably, the voltage, current, and phase angle are synchronously acquired through a distributed synchronous measurement terminal, and the switch status is uploaded in real time through a switch controller; the synchronous measurement terminal is a miniature synchronous phasor measurement unit.

[0015] Preferably, the preprocessing specifically involves: filtering, time-aligning, and anomaly removal of the running data at the edge nodes to generate cleaned time-series data; wherein, the filtering process includes adaptive Kalman filtering and wavelet denoising; and outliers are removed using residual testing.

[0016] Preferably, a graph convolutional network is used to determine the fault type in real time, specifically:

[0017] Based on the weighted undirected graph of the distribution network topology, the adjacency matrix is ​​preprocessed: a self-loop is added to the original adjacency matrix A to obtain the matrix. And according to the matrix Calculate the normalized matrix

[0018]

[0019] Where D is the node degree matrix;

[0020] The fault type probability distribution is obtained by constructing a graph convolutional network, and the fault type is determined based on the maximum probability; the graph convolutional network uses two layers of convolution:

[0021]

[0022] H (2) = AH (1) W (1) ,

[0023] Among them, H (1) and H (2) These are the outputs of the convolutions at layers 0 and 1, respectively; H (2) The fault type probability distribution is obtained through Softmax; X is the transient feature matrix; W (0) W (1) These are the trainable weight matrices for the convolutional layers 0 and 1, respectively; ReLU is a non-linear activation function.

[0024] Preferably, the fault type localization combines traveling wave ranging and impedance characteristics, and the specific method for determining the fault segment using traveling wave ranging and impedance is as follows:

[0025] The coarse measurement section of the fault is obtained by traveling wave ranging;

[0026] From the power source point to the starting point of the coarse measurement section, the shortest impedance path is determined based on minimizing the cumulative impedance.

[0027] The intersection of the coarsely measured section and the shortest path is taken as the final fault section.

[0028] Preferably, the automatic generation of a power supply reconfiguration scheme that meets current, voltage, and switching constraints specifically means that the power supply reconfiguration scheme simultaneously satisfies: the topology maintains a radial or weakly looped structure, the line current does not exceed its thermal stability limit, the node voltage is within the qualified range specified in the operating procedure, and the number of switching operations of a single feeder does not exceed its limit.

[0029] Preferably, the selection of the optimal solution through multi-objective optimization specifically involves: weighting multiple objectives, including non-power supply load, network loss increment, and voltage qualification rate, to select the optimal solution. The weight coefficients of the multi-objective optimization are adaptively set using fuzzy hierarchical analysis to ensure that non-power supply load, network loss increment, and voltage qualification rate are balanced under different operating scenarios.

[0030] Preferably, the method further includes recording the switching operations and load recovery information for each fault handling in a closed loop, performing statistical analysis on the historical records at a set period, and automatically adjusting the fault identification threshold and the priority of the reassembly scheme based on the statistical analysis results.

[0031] A distribution network fault intelligent location and self-healing reconfiguration system, wherein the system is implemented using any of the above-mentioned distribution network fault intelligent location and self-healing reconfiguration methods, comprising:

[0032] The data acquisition unit is used to synchronously acquire running data and complete data preprocessing;

[0033] The fault detection and location unit identifies fault types and determines fault sections based on transient characteristics and distribution network topology.

[0034] The candidate generation unit is reconstructed to automatically generate a set of power supply schemes that meet current, voltage, and switching constraints after the fault is isolated.

[0035] The optimization decision-making unit performs multi-objective optimization on the candidate schemes to obtain the optimal reorganization scheme;

[0036] The execution and learning unit issues switching commands to complete topology reconstruction and updates the fault identification and reconfiguration strategy based on the actual execution results;

[0037] Communication links are used to enable high-speed data exchange between various units.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] This invention deploys synchronous phasor measurement terminals at key nodes of the distribution network and performs high-precision filtering and time alignment at the edge. Combined with a graph convolutional network fault identification algorithm based on topology constraints, it can accurately determine the fault type within milliseconds and control the location error within a single line segment. Subsequently, multi-objective optimization is used to automatically generate and screen power reconfiguration schemes, achieving second-level isolation and topology reconfiguration, minimizing non-power supply loads and significantly reducing voltage-overriding nodes. Simultaneously, the closed-loop execution module statistically analyzes the results of each operation and dynamically adjusts the fault identification threshold and reconfiguration priority, enabling the system to maintain high location accuracy and rapid recovery capability even under scenarios involving distributed power source access or load structure changes, thereby significantly improving the power supply reliability and resilience of the distribution network. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the system architecture provided in an exemplary embodiment of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0042] This invention proposes a method for intelligent fault location and self-healing reconfiguration in power distribution networks, the method comprising:

[0043] Step S1: Obtain and preprocess the operating data of the distribution network, including voltage, current, phase angle and switch status;

[0044] Step S2: Model the distribution network topology as a weighted undirected graph and construct transient features including node features and edge features;

[0045] Step S3: Based on transient characteristics and distribution network topology, a graph convolutional network is used to determine the fault type in real time, and the fault section is determined by combining traveling wave ranging and impedance method;

[0046] The graph convolutional network is input with features including node voltage phasors, harmonic amplitudes, and edge impedances. It is trained by combining measured and simulated data to improve the accuracy of fault type identification, thereby reducing the false alarm rate.

[0047] Step S4: When entering the isolated fault state, automatically generate a power supply reconfiguration scheme that meets the current, voltage and switching constraints, and select the optimal scheme through multi-objective optimization;

[0048] Step S5: Issue a switch command according to the optimal solution to complete the topology reconstruction, and update the fault identification and reassembly strategy according to the actual execution results to improve positioning accuracy and recovery efficiency.

[0049] Furthermore, the voltage, current, phase angle, and switch status play specific roles in this method as follows: In step S2, the real-time primary topology is restored based on the switch status, and the voltage, current, and phase angle sequences are denoised and synchronized. Symmetrical components, harmonic / wavelet energy, phase angle abrupt changes, and traveling wave arrival time differences are extracted to form a node-level transient feature matrix. In step S3, this feature matrix and the topology are input into a graph convolutional network to determine the fault type, and traveling wave ranging and impedance measurement based on ΔZ≈ΔV / ΔI are cross-checked to accurately locate the fault section. In step S4, voltage measurement is used for voltage compliance constraints, current measurement is used for line thermal stability and power flow constraints, and switch status is used to limit the operable set, maintain the legality of radial / weak loop topology, and control the number of control actions, thereby generating and screening a power reconfiguration scheme through multi-objective optimization. In step S5, a closed-loop evaluation is carried out based on the actual closing / opening sequence and phasor recovery trajectory. The above observations are used as the basis for effect judgment and adaptive updating of thresholds and priorities to continuously improve fault location accuracy and power restoration efficiency.

[0050] In some embodiments, step S2 specifically includes:

[0051] Based on the primary wiring diagram of the distribution network, a weighted undirected graph is constructed with busbars and sectionalizing switches as nodes and lines or transformers as edges. The edge weights are taken as the equivalent impedance Z of the lines. ij ;

[0052] Node feature vector x i This includes the three-phase voltage amplitude, phase angle difference, and 25th harmonic amplitude;

[0053] Edge feature vector e ij This includes the line impedance modulus and power flow direction.

[0054] In some embodiments, distributed synchronous measurement terminals are used to acquire operational data such as voltage, current, phase angle, and switch status of the distribution network. Specifically, the synchronous measurement terminal can be an active distribution network synchronous vector measurement unit, which synchronously acquires voltage, current, and phase angle at a sampling rate of 10kHz; the switch controller uploads the circuit breaker status in real time.

[0055] The synchronous measurement terminal is a miniature synchronous phasor measurement unit. Specifically, the synchronous phasor measurement unit uses 24-bit ADS and GPS dual-mode timing, which improves the accuracy of traveling wave arrival time difference calculation and makes the positioning results more stable.

[0056] The voltage, current, and phase angle are synchronously acquired through a distributed synchronous measurement terminal, and the switch status is uploaded in real time through a switch controller; the synchronous measurement terminal is a miniature synchronous phasor measurement unit.

[0057] In some embodiments, the preprocessing specifically involves: filtering, time-aligning, and anomaly removal of the running data at edge nodes to generate cleaned time-series data; wherein, the filtering process includes adaptive Kalman filtering and wavelet denoising; outliers are removed using residual testing. This significantly filters out noise spikes, providing high-quality input for subsequent identification.

[0058] In some embodiments, graph convolutional networks are used to determine the fault type in real time, specifically:

[0059] Based on the weighted undirected graph of the distribution network topology, the adjacency matrix is ​​preprocessed: a self-loop is added to the original adjacency matrix A to obtain the matrix. And according to the matrix Calculate the normalized matrix

[0060]

[0061] Where D is the node degree matrix;

[0062] The fault type probability distribution is obtained by constructing a graph convolutional network, and the fault type is determined based on the maximum probability; the graph convolutional network uses two layers of convolution:

[0063]

[0064] H (2) = AH (1) W (1) ,

[0065] Among them, H (1) and H (2) These are the outputs of the convolutions at layers 0 and 1, respectively; H (2) The fault type probability distribution is obtained through Softmax; X is the transient feature matrix; W (0) W (1) These are the trainable weight matrices for the convolutional layers 0 and 1, respectively; ReLU is a non-linear activation function.

[0066] The graph convolutional network is trained using a dataset consisting of 600 real-world test samples and 1000 PSCAD simulation fault samples, covering types such as single-phase grounding, two-phase short circuit, and open circuit. The loss function uses weighted cross-entropy loss to address class imbalance. The optimizer is Adam with a learning rate of 0.001.

[0067] Take the maximum probability P in Softmax max If P max If the value is less than 0.6, a backup traditional impedance / traveling wave algorithm dual verification is triggered. This reduces false alarms and ensures engineering reliability.

[0068] In some embodiments, the fault type localization combines traveling wave ranging and impedance characteristics. Specifically, the fault segment is determined by combining traveling wave ranging and impedance methods as follows:

[0069] The coarse fault section is obtained by traveling wave ranging; the shortest impedance path is determined from the power source point to the starting point of the coarse fault section, based on minimizing the cumulative impedance; the intersection of the coarse fault section and the shortest path is taken as the final fault section; thus, the convergence time can be further shortened, avoiding getting trapped in local optima.

[0070] In some embodiments, the automatic generation of power supply reconfiguration schemes that meet current, voltage, and switching constraints specifically means that the power supply reconfiguration scheme simultaneously satisfies: the topology maintains a radial or weakly looped structure, the line current does not exceed its thermal stability limit, the node voltage is within the qualified range specified in the operating procedures, and the number of switching operations on a single feeder does not exceed its limit. Specifically, the topology is checked using depth-first traversal to ensure no loops; the number of switching operations on the same path is ≤2. Through the above method, all candidate schemes meet the national standard operating procedures, avoiding the risk of secondary tripping or overload.

[0071] In some embodiments, the selection of the optimal solution through multi-objective optimization specifically involves weighting multiple objectives, including non-power supply load, network loss increment, and voltage qualification rate, to select the optimal solution. The weight coefficients of the multi-objective optimization are adaptively set using fuzzy hierarchical analysis to ensure that non-power supply load, network loss increment, and voltage qualification rate are balanced under different operating scenarios. Specifically, the membership degrees of the three indicators are input, and the output weight vector is {0.5, 0.3, 0.2} (load, network loss, voltage). During peak periods, the weights are automatically adjusted to {0.6, 0.25, 0.15}. This allows the selection of the optimal solution under different operating scenarios to better align with the scheduling objectives.

[0072] In some embodiments, the method further includes closed-loop execution of recording the switching operations and load recovery information for each fault handling, and performing statistical analysis on the historical records at set intervals. Based on the statistical analysis results, the fault identification threshold and reassembly scheme priority are automatically adjusted. Specifically, the system scans historical fault records every 24 hours, calculates the average recovery time and the number of nodes exceeding the limit; if the average recovery time increases by more than 10%, the GCN threshold is automatically lowered and the network loss weight is increased. This adaptive threshold and weight adjustment keeps the positioning error and recovery time within a preset range, eliminating the need for manual parameter tuning during long-term operation.

[0073] refer to Figure 1 This invention proposes a distribution network fault intelligent location and self-healing reconfiguration system, which is implemented using any of the above-mentioned distribution network fault intelligent location and self-healing reconfiguration methods, including:

[0074] The data acquisition unit is used to synchronously acquire operational data and complete data preprocessing; synchronously acquire voltage, current, phase angle and switch status and complete filtering, time alignment and anomaly removal.

[0075] The fault detection and location unit identifies fault types and determines fault sections based on transient characteristics and distribution network topology.

[0076] The candidate generation unit is reconstructed to automatically generate a set of power supply schemes that meet current, voltage, and switching constraints after the fault is isolated.

[0077] The optimization decision-making unit performs multi-objective optimization on the candidate schemes to obtain the optimal reorganization scheme;

[0078] The execution and learning unit issues switching commands to complete topology reconstruction and updates the fault identification and reconfiguration strategy based on the actual execution results;

[0079] Communication links are used to enable high-speed data exchange between various units.

[0080] In some embodiments, the data unit employs a multi-core processor in conjunction with programmable logic;

[0081] The fault detection and localization unit supports cloud-edge collaborative model compression and online transfer learning to adapt to topology changes or distributed power supply access.

[0082] The optimization decision-making unit adopts rolling time-domain model predictive control, which automatically triggers secondary optimization for load or new energy output deviations.

[0083] The execution and learning unit writes the actual switching operation results into the distribution network operation knowledge graph and automatically updates the candidate solution library when the actual deviation from the plan exceeds the limit.

[0084] The communication link has a primary and backup channel hot backup function. When the primary channel experiences a continuous high packet loss rate, it automatically switches to the backup channel and seamlessly switches back after recovery.

[0085] This invention is not limited to the preferred embodiments described above. Specific parameters such as current, voltage, and time delay appearing in this specification are exemplary data and do not constitute a limitation of the claims.

[0086] For components that are functionally identical or similar, existing mature devices can be used for replacement, and the implementation method is not limited to the structure shown in this specification.

[0087] The present invention can also be implemented in the form of software, hardware, or a combination of software and hardware; wherein the software can be stored in a computer-readable medium and executed by a processor to perform the corresponding function.

[0088] Suitable computer-readable media include, but are not limited to, hard disks, flash memory, read-only memory (ROM), random access memory (RAM), and other media capable of storing program code.

[0089] The execution order of the steps described in the flowchart or logic block diagram can be adjusted or parallelized as needed, provided that it does not affect the implementation of the function.

[0090] The accompanying drawings referenced in this specification are for illustrative purposes only. Their dimensions, scale, or colors may be adjusted according to actual production needs. The reference numerals in the drawings should not be construed as limiting the scope of protection.

[0091] All the technical features disclosed in this invention can be combined in any way to form an interactive or collaborative structure, as long as the combination does not contradict each other or conflict with the technology.

[0092] Although embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope of protection claimed by the present invention.

Claims

1. A method for intelligent fault location and self-healing reconfiguration in a distribution network, characterized in that, The method includes: Step S1: Obtain and preprocess the operating data of the distribution network, including voltage, current, phase angle and switch status; Step S2: Model the distribution network topology as a weighted undirected graph and construct transient features including node features and edge features; Step S3: Based on transient characteristics and distribution network topology, a graph convolutional network is used to determine the fault type in real time, and the fault section is determined by combining traveling wave ranging and impedance method; Step S4: When entering the isolated fault state, automatically generate a power supply reconfiguration scheme that meets the current, voltage and switching constraints, and select the optimal scheme through multi-objective optimization; Step S5: Issue a switch command according to the optimal solution to complete the topology reconstruction, and update the fault identification and reassembly strategy according to the actual execution results to improve positioning accuracy and recovery efficiency.

2. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, Step S2 specifically involves: Based on the primary wiring diagram of the distribution network, a weighted undirected graph is established with busbars and sectionalizing switches as nodes and lines or transformers as edges. The edge weights are taken as the equivalent impedance of the lines. The node feature vector includes the three-phase voltage amplitude, phase angle difference, and 25th harmonic amplitude; The edge eigenvectors include the line impedance magnitude and power flow direction.

3. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The voltage, current, and phase angle are synchronously acquired through a distributed synchronous measurement terminal, and the switch status is uploaded in real time through a switch controller; the synchronous measurement terminal is a miniature synchronous phasor measurement unit.

4. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The preprocessing specifically involves filtering, time alignment, and anomaly removal of the running data at the edge nodes to generate cleaned time-series data; wherein, the filtering process includes adaptive Kalman filtering and wavelet denoising; and outliers are removed using residual testing.

5. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, A graph convolutional network is used to determine the fault type in real time, specifically: Based on the weighted undirected graph of the distribution network topology, the adjacency matrix is ​​preprocessed: a self-loop is added to the original adjacency matrix A to obtain the matrix. And according to the matrix Calculate the normalized matrix Where D is the node degree matrix; The probability distribution of fault types is obtained by constructing a graph convolutional network, and the fault type is determined based on the maximum probability; the graph convolutional network uses two layers of convolution: H (2) = AH (1) W (1) , Among them, H (1) and H (2) These are the outputs of the convolutions at layers 0 and 1, respectively; H (2) The fault type probability distribution is obtained through Softmax; X is the transient feature matrix; W (0) W (1) These are the trainable weight matrices for the convolutional layers 0 and 1, respectively; ReLU is a non-linear activation function.

6. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The fault type localization method combines traveling wave ranging and impedance characteristics, and specifically determines the fault segment using the traveling wave ranging and impedance method: The coarse measurement section of the fault is obtained by traveling wave ranging; From the power source point to the starting point of the coarse measurement section, the shortest impedance path is determined based on minimizing the cumulative impedance. The intersection of the coarsely measured section and the shortest path is taken as the final fault section.

7. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The automatic generation of a power supply reconfiguration scheme that meets current, voltage, and switching constraints specifically means that the power supply reconfiguration scheme simultaneously satisfies: the topology maintains a radial or weakly looped structure, the line current does not exceed its thermal stability limit, the node voltage is within the qualified range specified in the operating procedures, and the number of switching operations on a single feeder does not exceed its limit.

8. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The selection of the optimal solution through multi-objective optimization specifically involves weighting multiple objectives, including non-power supply load, network loss increment, and voltage qualification rate, to select the optimal solution. The weight coefficients of the multi-objective optimization are adaptively set using fuzzy hierarchical analysis to ensure that non-power supply load, network loss increment, and voltage qualification rate are balanced under different operating scenarios.

9. The intelligent fault location and self-healing reconfiguration method for distribution networks according to claim 1, characterized in that, The method also includes recording the switching operations and load recovery information for each fault handling in a closed loop, performing statistical analysis on the historical records at a set period, and automatically adjusting the fault identification threshold and the priority of the reassembly scheme based on the statistical analysis results.

10. A power distribution network fault intelligent location and self-healing reconfiguration system, characterized in that, The system is implemented using the intelligent fault location and self-healing reconfiguration method for distribution networks as described in any one of claims 1-9, including: The data acquisition unit is used to synchronously acquire running data and complete data preprocessing; The fault detection and location unit identifies fault types and determines fault sections based on transient characteristics and distribution network topology. The candidate generation unit is reconstructed to automatically generate a set of power supply schemes that meet current, voltage, and switching constraints after the fault is isolated. The optimization decision-making unit performs multi-objective optimization on the candidate schemes to obtain the optimal reorganization scheme; The execution and learning unit issues switching commands to complete topology reconstruction and updates the fault identification and reconfiguration strategy based on the actual execution results; Communication links are used to enable high-speed data exchange between various units.

Images