Power distribution network fault cooperative diagnosis method and system and related device

Abstract

Embodiments of the present application provide a power distribution network fault cooperative diagnosis method and system and related equipment. The technical solution comprises: acquiring real-time measurement data of each node of the power distribution network; generating an initial positioning probability vector based on the real-time measurement data; reporting the initial positioning probability vector, the original measurement sequence and the action protection message to the cloud master station as targeted data; receiving the authorization isolation instruction fed back by the cloud master station; executing fault isolation based on the authorization isolation instruction; generating a fault isolation confirmation message based on the fault isolation result, and feeding back the fault isolation message to the cloud master station. Embodiments of the present application preliminarily locate through the edge side feeder terminal, and the cloud master station evaluates in this way, establishing a closed-loop cooperative mechanism between the edge and the cloud. The edge requires less computing power, while taking into account "high fault tolerance" and "extremely fast solution".

Description

Technical Field

[0001] This application relates to the field of power distribution network fault diagnosis technology, and in particular to a collaborative fault diagnosis method, system and related equipment for power distribution networks. Background Technology

[0002] With the large-scale integration of distributed power sources into the distribution network, the traditional single-source radial network has evolved into a complex active distribution network with multiple sources and branches, which places higher demands on the accuracy and timeliness of fault location.

[0003] In existing technologies, edge-side feeder terminals mostly adopt a one-way communication mode of "direct reporting after sensing," with the cloud acting only as a passive data receiver. This lack of a mechanism for assessing the reliability of preliminary edge positioning results, and the inability of cloud knowledge to be fed back to the edge for parameter correction, leads to long-term stagnation of edge models and difficulty in adapting to complex and changing fault conditions. Therefore, how to improve positioning fault tolerance while ensuring limited computing power at the edge and construct a closed-loop architecture of cloud-edge bidirectional cognitive collaboration has become an urgent technical problem to be solved. Summary of the Invention

[0004] This application provides a method, system, and related equipment for collaborative fault diagnosis in power distribution networks to address existing problems in related technologies. The technical solution is as follows: In a first aspect, embodiments of this application provide a collaborative fault diagnosis method for a distribution network, applied to edge-side feeder terminals, comprising: Acquire real-time measurement data of each node in the power distribution network; An initial location probability vector is generated based on real-time measurement data. The initial location probability vector is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network. The initial positioning probability vector, the original measurement sequence, and the action protection message are reported as target data to the cloud master station. Receive the authorization isolation command from the cloud master station. The authorization isolation command is generated by the cloud master station after it evaluates the credibility of the initial positioning probability vector and determines that the evaluation is successful. Perform fault isolation based on authorized isolation commands; A fault isolation confirmation message is generated based on the fault isolation results, and the fault isolation message is sent back to the cloud main station.

[0005] In one implementation, generating an initial positioning probability vector based on real-time measurement data includes: An initial state feature vector is generated based on real-time measurement data. A topological adjacency matrix is ​​constructed based on the physical topological connection relationship of the distribution network. The topological adjacency matrix is ​​then algebraically multiplied with the initial state feature vector to calculate the node connectivity feature vector. Based on the overcurrent alarm status of each node and its corresponding connectivity feature value, the distorted state that violates the physical topology logic is adaptively flipped to generate a noiseless state feature vector. The initial weight matrix, which is pre-set locally, is called and forward multiplied with the noiseless state feature vector to obtain the initial localization probability vector.

[0006] In one implementation, it further includes: Receive the dynamic weight matrix sent by the cloud master station. The dynamic weight matrix is ​​the feedback parameter generated by the cloud master station for the interference section where the initial positioning probability vector is greater than the first threshold and is confirmed to be without fault. Perform a second forward dot product calculation between the dynamic weight matrix and the cached real-time state feature vector to generate a secondary localization probability vector; The actual fault section is determined based on the secondary positioning probability vector, a relay protection command is generated, and fault isolation is performed based on the relay protection command.

[0007] In one implementation, determining the actual faulty section based on the secondary location probability vector includes: The maximum value index is used to determine the segment index corresponding to the maximum probability value in the secondary positioning probability vector, and the segment is identified as the actual fault segment.

[0008] Secondly, embodiments of this application provide a collaborative fault diagnosis method for a distribution network, applied to a cloud-based master station, comprising: Receive target data reported by edge-side feeder terminals. The target data includes at least an initial location probability vector, which is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network. Obtain multimodal constraint information, which includes at least one of the following: real-time power grid topology status, current weather conditions, and historical equipment operation and maintenance records; The confidence score is obtained based on the initial positioning probability vector and multimodal constraint information; Based on the confidence score, the faulty section that has been confirmed to have failed is identified, and an authorized isolation command is sent to the edge-side feeder terminal.

[0009] In one implementation, it further includes: Identify the interference segment whose initial positioning probability vector is greater than the first threshold and is confirmed to be a fault-free area; Cognitive bias features are extracted based on the physical and logical conflict between multimodal constraint information and the initial localization probability vector, and a spatial penalty matrix and a state compensation vector are constructed. Based on the spatial penalty matrix and the state compensation vector, a dynamic correction is applied to the preset initial weight matrix to generate a dynamic weight matrix; The dynamic weight matrix is ​​sent to the edge-side feeder terminal.

[0010] In one implementation, it further includes: Receive fault isolation confirmation messages from the edge-side feeder terminal; Update the global physical connectivity status of the distribution network based on the isolated fault sections, and reconstruct the isolated topology; Based on the reconstructed topology matrix, and combined with the load demand of each non-fault outage section, the available capacity of distributed power sources, and the thermal stability margin of the lines, the optimal load transfer path is calculated. Generate collaborative isolation and adaptive power transfer scheduling strategies and push them to the distribution network SCADA system or dispatcher workstation.

[0011] Thirdly, embodiments of this application provide a distribution network fault collaborative diagnosis system, including an edge-side feeder terminal and a cloud master station; The edge-side feeder terminal is used to perform preliminary fault location, generate an initial location probability vector and report it to the cloud master station, and receive the dynamic weight matrix fed back by the cloud master station and perform secondary fault location and physical isolation. The cloud-based main station receives the initial positioning probability vector, performs a reliability assessment by combining multimodal constraints, selectively generates a dynamic weight matrix based on the assessment results and sends it to the edge-side feeder terminal, and performs global topology reconstruction and power restoration decisions based on the isolation results fed back by the edge-side feeder terminal.

[0012] Fourthly, embodiments of this application provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the above-described power distribution network fault collaborative diagnosis method.

[0013] Fifthly, embodiments of this application provide a computer-readable storage medium that stores computer instructions, which, when executed on a computer, allow the methods in any of the above-described embodiments to be performed.

[0014] This application embodiment acquires real-time measurement data and generates an initial positioning probability vector through the edge-side feeder terminal. Then, the target data, including the probability vector, the original measurement sequence, and the action protection message, is reported to the cloud for credibility assessment. Fault isolation is only executed and a confirmation message is sent back after receiving an authorized isolation command from the cloud. Thus, while completing the initial positioning on the edge side without relying on complex iterative algorithms, a closed-loop collaborative mechanism of "edge initial perception - cloud credibility assessment - authorized execution isolation - result feedback confirmation" is established. This effectively solves the technical problems in the prior art where the edge side is limited by computing power and cannot balance fast positioning and high fault tolerance, and the lack of credibility assessment and two-way interaction between the cloud and the edge leads to high system malfunction risk and insufficient positioning reliability. It realizes a collaborative closed loop of lightweight and fast edge perception and cloud cognitive decision-making.

[0015] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of this application will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0016] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0017] Figure 1 A schematic flowchart illustrating a collaborative fault diagnosis method for a power distribution network according to an embodiment of this application is shown. Figure 2 Show Figure 1 Flowchart of the sub-steps in step S220.

[0018] Figure 3 A schematic flowchart illustrating a collaborative fault diagnosis method for power distribution networks according to another embodiment of this application is shown.

[0019] Figure 4 A schematic flowchart illustrating a collaborative fault diagnosis method for a power distribution network according to yet another embodiment of this application is shown.

[0020] Figure 5 A schematic flowchart illustrating a collaborative fault diagnosis method for a distribution network according to another embodiment of this application is shown.

[0021] Figure 6 A schematic flowchart illustrating a collaborative fault diagnosis method for a distribution network according to another embodiment of this application is shown.

[0022] Figure 7This is a schematic structural block diagram of a power distribution network fault collaborative diagnosis system according to an embodiment of this application.

[0023] Figure 8 A block diagram of an electronic device used to implement the power distribution network fault collaborative diagnosis method according to embodiments of this application is shown. Detailed Implementation

[0024] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0025] This application provides a method for collaborative fault diagnosis in a distribution network. The system architecture includes edge-side feeder terminals and a cloud-based master station deployed in a dispatch center. The cloud-based master station has at least data analysis, storage, and processing functions to analyze and process information fed back from the edge-side feeder terminals. The edge-side feeder terminals are communicatively connected to each node in the distribution network and include at least a data acquisition module and a relay protection actuator.

[0026] This application first provides a method for collaborative fault diagnosis in a distribution network, applied to edge-side feeder terminals. Figure 1 A schematic flowchart illustrating a collaborative fault diagnosis method for a distribution network according to an embodiment of this application is shown. Figure 1 As shown, the collaborative fault diagnosis method for power distribution networks may include: S210 acquires real-time measurement data of each node in the distribution network.

[0027] In a distribution network, each node is equipped with a feeder terminal unit (FTU). The FTU collects electrical quantity data passing through that node in real time via current transformers. Real-time measurement data includes zero-sequence current for single-phase ground faults and phase current for phase-to-phase short circuits. After analog-to-digital conversion (A / D), the FTU's local microprocessor first filters out transient high-frequency interference signals using a low-pass filtering algorithm to extract the RMS value of the power frequency fundamental wave. Those skilled in the art should understand that for different types of distribution networks, real-time measurement data may also include electrical parameters such as three-phase voltage amplitude and three-phase current amplitude.

[0028] S220 generates an initial location probability vector based on real-time measurement data. The initial location probability vector is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network, and each element of the vector corresponds to the fault probability value of a potential fault section.

[0029] In this embodiment of the application, the edge side does not directly trigger trip isolation based on the initial positioning probability vector, but uses the calculated initial positioning probability vector as the preliminary judgment result to enter the next cloud-edge collaborative evaluation process.

[0030] S230 reports the initial positioning probability vector, the original measurement sequence, and the action protection message as target data to the cloud master station.

[0031] The original measurement sequence is the sequence of continuous effective current values ​​collected at each node.

[0032] Action protection messages include, but are not limited to: overcurrent alarm timestamp, node number, alarm type, protection action status, etc.

[0033] By packaging multi-dimensional data into targeted data for reporting, the cloud-based main station can obtain richer decision-making information than a single probability vector.

[0034] S240 receives the authorization isolation command from the cloud master station. The authorization isolation command is generated by the cloud master station after it evaluates the credibility of the initial positioning probability vector and determines that the evaluation is successful.

[0035] After receiving the target data reported by the FTU, the cloud master station performs a reliability assessment of the initial localization probability vector. If the reliability assessment passes, the cloud master station issues an authorized isolation command to the corresponding edge-side FTU. The authorized isolation command may include the index of the segment confirmed to have failed, the isolation operation permission identifier, and the operation timing requirements. After receiving the command, the FTU confirms that the preliminary localization results are reliable.

[0036] S250 performs fault isolation based on authorized isolation instructions.

[0037] After receiving the authorized isolation command from the cloud master station, the FTU generates a deterministic relay protection command. Based on the relay protection command, it triggers the associated circuit breaker to perform a tripping action, completing the physical isolation of the fault section. Specifically, the edge terminal sends a tripping control signal to the smart circuit breaker or load switch on both sides of the corresponding section to cut off the fault current path.

[0038] S260 generates a fault isolation confirmation message based on the fault isolation result and sends the fault isolation message back to the cloud master station.

[0039] After completing the circuit breaker tripping operation, the FTU collects the actual opening and closing status of the circuit breaker and generates a fault isolation confirmation message containing information such as the fault section number, tripping execution time, and circuit breaker status. This message is then synchronously transmitted back to the cloud master station via the dedicated network for subsequent topology reconstruction and recovery decisions.

[0040] This application embodiment acquires real-time measurement data and generates an initial positioning probability vector through the edge-side feeder terminal. Then, the target data, including the probability vector, the original measurement sequence, and the action protection message, is reported to the cloud for credibility assessment. Fault isolation is only executed and a confirmation message is sent back after receiving an authorized isolation command from the cloud. Thus, while completing the initial positioning on the edge side without relying on complex iterative algorithms, a closed-loop collaborative mechanism of "edge initial perception - cloud credibility assessment - authorized execution isolation - result feedback confirmation" is established. This effectively solves the technical problems in the prior art where the edge side is limited by computing power and cannot balance fast positioning and high fault tolerance, and the lack of credibility assessment and two-way interaction between the cloud and the edge leads to high system malfunction risk and insufficient positioning reliability. It realizes a collaborative closed loop of lightweight and fast edge perception and cloud cognitive decision-making.

[0041] In one implementation, such as Figure 2 As shown, step S220 includes: S221, based on real-time measurement data, generates an initial state feature vector.

[0042] For example, the edge-side feeder terminals deployed at each node of the distribution network collect the zero-sequence current passing through that node in real time through current transformers. After analog-to-digital conversion (A / D), the FTU local microprocessor first sets an overcurrent safety threshold. (This threshold can be dynamically adjusted according to the line's rated load), and the continuous analog quantity is converted into a discrete overcurrent alarm state through an indication function. ∈{0,1}.

[0043] At time t, the states of all nodes in the network are synchronously sampled and aligned to generate an initial state feature column vector: ; Where N is the total number of feeder terminal nodes in the distribution network. The conversion rule is: when the measured current value is greater than or equal to the overcurrent safety threshold, =1 indicates that the node is in an overcurrent state; otherwise... =0.

[0044] S222, construct a topological adjacency matrix based on the physical topological connection relationship of the distribution network, and perform algebraic multiplication of the topological adjacency matrix with the initial state feature vector to calculate the node connectivity feature vector.

[0045] Based on the current static connection relationships of the physical lines in the distribution network, a symmetric adjacency matrix representing the network spatial topology is constructed. .

[0046] Elements of adjacency matrix A The value selection rule is: if the node With nodes If there is a direct physical connection, then =1; otherwise =0.

[0047] The adjacency matrix A is compared with the initial state eigenvector. Perform algebraic chain multiplication to calculate the node connectivity eigenvectors. : ; Where, for any i∈{1,2,…,N}, the element , representing the node The number of all physically adjacent nodes in the "overcurrent (1)" state. This eigenvalue quantifies the abstract physical connectivity constraints, providing a direct mathematical basis for accurately identifying distorted sites that violate Kirchhoff's current law.

[0048] S223, based on the overcurrent alarm status of each node and its corresponding connectivity feature value, adaptively flip the distorted state that violates the physical topology logic to generate a noiseless state feature vector.

[0049] Based on the physical continuity of current flow—the short-circuit fault current in the power grid has spatial continuity, meaning the overcurrent signal must propagate continuously along connected lines, and the state of each node is strictly bound and constrained by the physical topology—an adaptive state reconstruction function is designed. It identifies and corrects "isolated 0s (false negatives)" or "isolated 1s (false positives)" that violate physical logic. The flipping rules are as follows: ; If the overcurrent alarm status of a certain node is 0 (i.e. =0), but its connectivity eigenvalue If the value is ≥2 (meaning the node is surrounded by two or more overcurrent neighbor nodes), then the node is determined to be a missed node caused by communication interference, and its state is flipped to 1. If the overcurrent alarm status of a certain node is 1 (i.e. =1), but its connectivity eigenvalue If the value is 0 (meaning there are no overcurrent neighbor nodes around the node), then the node is determined to be a false alarm node, and its state is flipped to 0.

[0050] The states of each node after adaptive flipping Arranged in order of the original node numbers, they are constructed as column vectors, and the output is a noiseless state feature vector that conforms to the physical topological logic. .

[0051] Through the above fault-tolerant cleaning mechanism, even under complex operating conditions where communication interference causes FTU alarm state distortion or missed alarms, the system can still obtain state feature vectors that conform to physical topology logic.

[0052] S224: Call the pre-set local initial weight matrix and perform a forward dot product calculation with the noiseless state feature vector to obtain the initial localization probability vector.

[0053] FTU calls the pre-set initial weight matrix locally. (K is the total number of potentially faulty isolation sections in the distribution network), and the cleaned noise-free state feature vector. Perform forward dot product calculation: ; Output That is, the initial localization probability vector, where This represents the probability of a fault occurring in the k-th segment. The calculation process is a purely algebraic analytical step with a time complexity of O(1). It does not require any iterative optimization calculations on the edge side and can be completed very quickly on the edge-side microprocessor, effectively overcoming the engineering constraints of limited computing power of the feeder terminal microprocessor.

[0054] The initial weight matrix W, pre-set locally, is obtained as follows: On the high-performance computing server of the cloud main station, M sets of typical distribution network fault samples are generated using historical operation records of the distribution network and / or a high-fidelity simulation model constructed using electromagnetic transient simulation software. To ensure the global generalization capability of the mapping model, the sample set needs to cover various complex operating conditions, specifically including: different fault types (single-phase grounding, two-phase short circuit, three-phase short circuit), different transition resistances (metallic short circuit to high-resistance grounding), different load levels, and short-circuit characteristics under different output fluctuation scenarios of high-proportion distributed power sources (such as photovoltaic and wind power).

[0055] Extract the input features after data cleaning to form a large-dimensional sample input matrix. Simultaneously, one-hot encoding is used to construct the corresponding real fault segment label matrix. Introducing the Tikhonov-regularized Moore-Ponros generalized inverse theory, the initial weight matrix W is obtained by solving the following analytical formula: ; in, It is the transpose matrix. λ is the identity matrix, and λ is the set minimum regularization penalty coefficient. This weight matrix is ​​pre-calculated by the cloud main station and then distributed to the FTU storage on each edge side.

[0056] In one embodiment, this application further provides an implementation method whereby the edge-side FTU receives a dynamic weight matrix fed back from the cloud and performs secondary fault location. The steps in this embodiment can be performed in parallel with the steps in Embodiment 1; specifically, it is an alternative branch path where the authorized isolation command is received in step S240 of Embodiment 1.

[0057] Specifically, such as Figure 3 As shown, it includes: S310 receives the dynamic weight matrix sent by the cloud master station. The dynamic weight matrix is ​​the feedback parameter generated by the cloud master station for interference sections whose probability values ​​in the initial positioning probability vector are greater than the first threshold and are confirmed to be fault-free.

[0058] The first threshold can be a threshold set based on experience, and segments that exceed the first threshold are identified as faulty segments at the edge.

[0059] In one example, a power distribution network is divided into 5 sections (K=5), and the initial location probability vector output by the edge side is: ; but: The probability of segment 2 is 0.85 (the highest, with an absolute advantage), and it is determined to be a faulty segment; The probabilities of segments 1, 3, 4, and 5 are 0.02, 0.05, 0.06, and 0.02, respectively, all of which are much less than 0.85, and are therefore determined to be non-faulty segments.

[0060] Dynamic weight matrix It is a feedback parameter generated by the cloud master station based on the physical and logical conflict between multimodal constraint information (including real-time power grid topology status, current weather conditions and equipment historical operation and maintenance records) and the initial positioning probability vector.

[0061] The cloud-based main station uses multimodal constraints and the initial localization probability vector. The physical and logical conflicts between them are analyzed to extract cognitive bias features, deconstruct cognitive biases, and construct a spatial penalty matrix. With state compensation vector .

[0062] Apply dynamic adjustments to the initial weight matrix W, and generate the dynamic weight matrix according to the following formula: ; in, Update the incremental matrix for parameters generated based on the state compensation vector.

[0063] After receiving the dynamic weight matrix, the FTU microprocessor immediately overwrites the static mapping parameters in its local memory, replacing the initial weight matrix W with the dynamic weight matrix. This mechanism enables online adaptive correction of model parameters, laying the parameter foundation for subsequent high-precision secondary analysis.

[0064] S320 performs a second forward dot product calculation on the dynamic weight matrix and the cached real-time state feature vector to generate a secondary localization probability vector.

[0065] The edge feeder terminal microprocessor receives and reloads the dynamic weight matrix sent from the cloud. Then, it is compared with the cached real-time distortion feature vector. Perform the second forward dot product calculation as follows: ; Output This is the high-precision secondary positioning probability vector after cloud-edge cognitive collaborative correction. Because... Having incorporated the spatial penalty and state compensation constraints of the current mutation conditions, this secondary inference process successfully filters out "pseudo-feature" disturbances caused by extreme interference. It is worth noting that the algorithm's time complexity remains constant at O(1), and the secondary localization does not increase the computational burden on the FTU.

[0066] S330 determines the actual fault section based on the secondary positioning probability vector, generates relay protection commands, and performs fault isolation based on the relay protection commands.

[0067] Specifically, the maximum probability value in the secondary positioning probability vector is determined by the maximum value index function, and the segment index is locked as the actual fault segment.

[0068] FTU uses the maximum index function to... Perform secondary target locking: ; The system ultimately pinpointed the actual faulty section with a high confidence level. The edge terminal then generates deterministic relays based on this information; The protection command triggers the associated circuit breaker to perform a tripping action, completing highly reliable physical isolation of the fault, and synchronously sending the final isolation confirmation message back to the cloud master station.

[0069] This application's embodiments introduce a cloud-edge cognitive collaborative feedback mechanism, enabling the edge side to obtain cloud compensation parameters under complex distortion conditions, perform high-precision secondary analytical positioning, and maintain a computational complexity of O(1), thus balancing positioning accuracy and real-time performance.

[0070] This application also provides a method for collaborative fault diagnosis in a distribution network, applied to a cloud-based master station, such as... Figure 4 As shown, the collaborative fault diagnosis method for the power distribution network includes: S401, Receives target data reported by the edge-side feeder terminal. The target data includes at least an initial location probability vector, which is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network.

[0071] S402, Obtain multimodal constraint information. Multimodal constraint information includes at least one of the following: real-time power grid topology status, current weather conditions, and historical equipment maintenance records.

[0072] Real-time power grid topology status: including the current on / off status of each switching device and line connection relationships; Current meteorological conditions: information on extreme weather events that may affect power grid operation, such as thunderstorms, typhoons, and heavy rain. Historical maintenance records for equipment include historical failure rates, maintenance records, and equipment aging status for each FTU.

[0073] As an alternative implementation method, multimodal constraint information can also include real-time output data of distributed power sources, load forecast data, etc.

[0074] S403, the confidence score is obtained based on the initial positioning probability vector and multimodal constraint information.

[0075] Specifically, the cloud-based main station extracts the initial positioning probability vector. The information entropy is calculated and mathematically aligned with multimodal constraints to establish a quantitative credibility evaluation function. ; in, Let K be the probability value of the i-th segment in the initial positioning probability vector, and K be the total number of segments. The first two terms are probability concentration characteristics based on information entropy. When the probability distribution is more concentrated (i.e., the probability value of a certain segment is significantly higher than that of other segments), the information entropy is smaller and the confidence score is higher. This is a penalty term for topology and environmental anomalies. When there are anomalies in the power grid topology (such as unplanned switch changes) or when the current weather conditions are severe, the value of this penalty term increases. This is a preset penalty coefficient.

[0076] The cloud-based main site calculates the credibility score based on the above formula. The score is between 0 and 1, and the higher the score, the more reliable the initial positioning result.

[0077] S404, based on the confidence score, identifies the faulty section where the fault has occurred and sends an authorized isolation command to the edge-side feeder terminal.

[0078] Specifically, the cloud-based main site will calculate the credibility score. Compare with the preset safety threshold γ: like If the probability is greater than or equal to γ, the preliminary analysis result indicated by the initial positioning probability vector is deemed reliable. The cloud-based master station determines the faulty segment based on the segment index corresponding to the maximum probability value in the initial positioning probability vector and issues an authorized isolation command to the corresponding edge-side FTU. Upon receiving the command, the FTU performs the fault isolation operation.

[0079] like If the value is less than γ, the initial static parameters are determined to be invalid under the current distortion conditions. The system rejects the preliminary positioning result and triggers the cognitive feedback mechanism.

[0080] The fault collaborative diagnosis method executed by the cloud master station provided in this application introduces a quantitative credibility evaluation function based on information entropy, which enables the cloud to scientifically verify the preliminary positioning results of the edge side in combination with multimodal constraints. This effectively overcomes the technical defect that the cloud only acts as a passive data receiver under the traditional "one-way open-loop" architecture and realizes the first step of the closed loop of "perception-evaluation".

[0081] In one implementation, based on the above steps, this application further provides a method whereby the cloud-based master station triggers a cognitive feedback mechanism when the trustworthiness assessment fails, generates a dynamic weight matrix, and distributes it to the edge side to achieve secondary locking of the faulty segment. Specifically, as shown... Figure 5 As shown, the method also includes: S405, Determine that the initial positioning probability vector is greater than the first threshold and is confirmed as an interference section without fault.

[0082] When the cloud-based main site determines the credibility score When γ < γ, the system triggers a cognitive feedback mechanism. The cloud-based main station first processes the initial localization probability vector. A step-by-step analysis is performed to identify sections with probability values ​​greater than a first preset threshold (e.g., 0.1, but this can be flexibly set by those skilled in the art based on the actual power grid scale) but confirmed as fault-free based on multimodal constraint information (especially power grid topology status and equipment historical records). These sections were initially misjudged as faulty by the edge-side localization algorithm, but were determined to be non-faulty by the cloud through global information. These sections are marked as "interference sections." The corresponding sections are the "pseudo-feature" sections in the initial localization.

[0083] S406 extracts cognitive bias features based on the physical and logical conflict between multimodal constraint information and the initial localization probability vector, and constructs a spatial penalty matrix and a state compensation vector.

[0084] The cloud-based main station uses multimodal constraint information (real-time power grid topology status, meteorological constraints, and historical equipment records) and an initial positioning probability vector. Physical-logical conflicts between components allow for the extraction of cognitive bias characteristics. For example, if a segment is topologically disconnected but is assigned a high probability of failure, or if a device shows recent maintenance in historical records but is marked as having a high probability of failure, these all constitute "physical-logical conflicts."

[0085] Based on the extracted cognitive bias features, the cloud-based main station constructs a spatial penalty matrix targeting specific interference areas. With state compensation vector Among them, the spatial penalty matrix Used to suppress the weights of the segment containing the interference segment, state compensation vector Used to provide positive compensation for the weights of segments containing non-interference segments.

[0086] S407, based on the spatial penalty matrix and the state compensation vector, applies dynamic correction to the preset initial weight matrix to generate a dynamic weight matrix.

[0087] The cloud-based main station applies dynamic cognitive feedback constraints to the initial weight matrix W based on feedback compensation parameters, and calculates the dynamic weight matrix for the current scenario through local correction formulas. : ; In the formula, ΔW is the state compensation vector. The generated parameter update increment matrix is ​​used. This correction process abandons the time-consuming global historical full matrix recalculation, and instead starts a fast feedback compensation engine based on local residuals, which only performs local corrections in the weight region corresponding to the interference segment, resulting in extremely high computational efficiency.

[0088] S408 sends the dynamic weight matrix to the edge-side feeder terminal.

[0089] The cloud-based main site calculates the dynamic weight matrix. Then, it is used as a cognitive feedback control signal and sent to the corresponding edge-side FTU via the dedicated power grid. Upon receiving the feedback parameters, the edge FTU microprocessor quickly overwrites the static mapping parameters (i.e., the original initial weight matrix W) in its local memory. Write it into memory for subsequent secondary fault location.

[0090] This application embodiment effectively breaks the deadlock of parameter rigidity under traditional unidirectional open-loop control by constructing a complete cognitive feedback link of "cognitive bias deconstruction - feedback compensation matrix construction - weight dynamic correction - edge parameter reloading", realizing online adaptive dynamic correction of mapping model parameters, enabling the edge side to obtain cloud compensation parameters under complex distortion conditions.

[0091] In one implementation, such as Figure 6As shown, it also includes: S409, Receive fault isolation confirmation message from edge-side feeder terminal; The cloud-based master station receives fault isolation confirmation messages from the edge-side FTU via a dedicated network communication module. These confirmation messages must contain at least the actual faulty segment number. The message includes confirmation information such as the trip execution time and the actual open / closed status of the associated circuit breaker. Upon receiving this message, the cognitive collaborative decision-making engine is triggered at the cloud master station.

[0092] S410 updates the global physical connectivity status of the distribution network based on the isolated fault sections and reconstructs the isolated topology.

[0093] The cloud-based main station first determines the fault segment number from the confirmation message. The system updates the global physical connectivity status of the distribution network in real time within the system. Specifically, it updates the faulty sections... The system removes nodes from the current topology, updates the connectivity relationships between nodes, and reconstructs the isolated active distribution network spatial topology matrix. This topology reconstruction process needs to consider the access location and status of distributed generation sources to ensure that the reconstructed topology accurately reflects the actual physical connectivity of the current power grid.

[0094] S411, based on the reconstructed topology matrix, combined with the load demand of each non-faulty power outage section, the available capacity of distributed power sources, and the thermal stability margin of the lines, calculates the optimal load transfer path.

[0095] Based on the reconstructed global topology, the cloud-based main station optimizes power restoration by considering the following multi-source constraints: Real-time load demand of each non-faulty power outage section—that is, the power supply capacity required by each section that is forced to shut down due to fault isolation but is not itself faulty; Available capacity of distributed generation (DG) – including the current actual output, adjustability margin and grid connection capability of each distributed generation; Line thermal stability margin – the difference between the maximum current carrying capacity of the line where each tie switch is located and the current load.

[0096] The cloud-based master station aims to "maximize the restored power load" and "minimize network losses". Under the premise of meeting the above constraints, it adopts an optimization algorithm based on power flow constraints (such as, but not limited to, forward-backward substitution method, Newton-Raphson method, etc., which can be flexibly selected by those skilled in the art according to the actual power grid scale) to automatically calculate the optimal load transfer path that meets the system's safe operation constraints, i.e., the specific sequence of tie switches that need to be closed.

[0097] S412 generates collaborative isolation and adaptive power transfer scheduling strategies and pushes them to the distribution network SCADA system or dispatcher workstation.

[0098] The cloud-based master station generates the optimal load transfer path into a standardized "Cooperative Isolation and Adaptive Load Transfer Scheduling Strategy" file. This scheduling strategy includes, but is not limited to, key information such as: a list of tie switches that need to be closed, the action sequence of each switch, the expected load transfer capacity, and the estimated system power flow distribution after power restoration.

[0099] Subsequently, the cloud-based master station pushes the dispatching strategy to the distribution network SCADA system or dispatcher workstation via a standard communication interface, for dispatchers to review and confirm, or for the SCADA system to execute automatically. This completely establishes a closed-loop control workflow for the distribution network, from "high-speed perception at the lower level" to "reliable cloud-based assessment and parameter feedback," and finally to "highly reliable collaborative recovery."

[0100] This embodiment builds a power restoration decision mechanism based on global topology reconstruction and multi-source constraint optimization in the cloud, enabling the system to not only accurately execute the underlying fault isolation, but also automatically generate reasonable load transfer strategies for non-faulty areas, thus upgrading the distribution network self-healing technology from "bottom-level physical isolation" to "high-level cognitive closed loop".

[0101] This application provides a collaborative fault diagnosis system for power distribution networks, such as... Figure 7 As shown, it includes an edge-side feeder terminal (FTU) 710 and a cloud master station 720.

[0102] The edge-side feeder terminal is used to perform preliminary fault location, generate an initial location probability vector and report it to the cloud master station, and receive the dynamic weight matrix fed back by the cloud master station to perform secondary fault location and physical isolation.

[0103] Specifically, the edge-side feeder terminal unit (FTU) includes the following functional modules: The data acquisition module 711 is used to acquire real-time measurement data of each node in the distribution network; The initial positioning module 712 is used to generate an initial positioning probability vector based on real-time measurement data. The initial positioning probability vector is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network. The first communication module 713 is used to report the initial positioning probability vector, the original measurement sequence and the action protection message as target data to the cloud master station. And the authorized isolation command used to receive feedback from the cloud master station. The authorized isolation command is generated by the cloud master station after it evaluates the credibility of the initial positioning probability vector and determines that the evaluation is successful. The isolation execution module 714 is used to perform fault isolation based on authorized isolation instructions; The first communication module 713 is also used to generate a fault isolation confirmation message based on the fault isolation result and to send the fault isolation message back to the cloud master station.

[0104] The cloud-based main station receives the initial positioning probability vector, performs a reliability assessment by combining multimodal constraints, selectively generates a dynamic weight matrix based on the assessment results and sends it to the edge-side feeder terminal, and performs global topology reconstruction and power restoration decisions based on the isolation results fed back by the edge-side feeder terminal.

[0105] Specifically, the cloud-based main site includes the following functional modules: The second communication module 721 is used to receive target data reported by the edge-side feeder terminal. The target data includes at least an initial location probability vector, which is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network. The credibility assessment module 722 is used to obtain multimodal constraint information, which includes at least one of the following: real-time power grid topology status, current meteorological conditions, and historical equipment operation and maintenance records. And used to obtain a confidence score based on the initial positioning probability vector and multimodal constraint information; The determination module 723 is used to determine the faulty section that has been confirmed to have a fault based on the confidence score, and send an authorized isolation command to the edge feeder terminal.

[0106] The functions of each module in each device in the embodiments of this application can be found in the corresponding descriptions in the above methods, and will not be repeated here.

[0107] Figure 8 A structural block diagram of an electronic device according to an embodiment of this application is shown. Figure 8 As shown, the electronic device includes a memory 910 and a processor 920. The memory 910 stores instructions that can be executed on the processor 920. When the processor 920 executes the instructions, it implements the power distribution network fault collaborative diagnosis method in the above embodiments. The number of memories 910 and processors 920 can be one or more. This electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0108] The electronic device may also include a communication interface 930 for communicating with external devices and exchanging data. The devices are interconnected using different buses and can be mounted on a common motherboard or otherwise as needed. The processor 920 can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). The bus can be divided into address buses, data buses, control buses, etc. For ease of illustration, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0109] Optionally, in a specific implementation, if the memory 910, processor 920, and communication interface 930 are integrated on a single chip, the memory 910, processor 920, and communication interface 930 can communicate with each other through an internal interface.

[0110] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting the Advanced Reduced Instruction Set Computing (RISC) machine (ARM) architecture.

[0111] This application provides a computer-readable storage medium (such as the memory 910 described above) that stores computer instructions, which, when executed by a processor, implement the method provided in this application.

[0112] Optionally, the memory 910 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the use of the electronic equipment in the distribution network fault collaborative diagnosis method, etc. Furthermore, the memory 910 may include high-speed random access memory and may also include non-transient memory, such as at least one disk storage device, flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 910 may optionally include memory remotely located relative to the processor 920, and these remote memories can be connected to the electronic equipment of the distribution network fault collaborative diagnosis method via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0113] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0114] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0115] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more (two or more) executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0116] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0117] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0118] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for collaborative fault diagnosis in a power distribution network, characterized in that, Applications to edge-side feeder terminals include: Acquire real-time measurement data of each node in the power distribution network; An initial location probability vector is generated based on the real-time measurement data. The initial location probability vector is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network. The initial positioning probability vector, the original measurement sequence, and the action protection message are reported as target data to the cloud main station. The system receives an authorization isolation command from the cloud master station, which is generated by the cloud master station after it performs a credibility assessment on the initial positioning probability vector and determines that the assessment is successful. Based on the authorized isolation command, perform fault isolation; A fault isolation confirmation message is generated based on the fault isolation result, and the fault isolation message is sent back to the cloud master station.

2. The method according to claim 1, characterized in that, The step of generating an initial positioning probability vector based on the real-time measurement data includes: Based on the real-time measurement data, an initial state feature vector is generated; A topological adjacency matrix is ​​constructed based on the physical topological connection relationship of the distribution network. The topological adjacency matrix is ​​then algebraically multiplied with the initial state feature vector to calculate the node connectivity feature vector. Based on the overcurrent alarm status of each node and its corresponding connectivity feature value, the distorted state that violates the physical topology logic is adaptively flipped to generate a noiseless state feature vector. The initial weight matrix, which is pre-set locally, is called and forward multiplied with the noiseless state feature vector to obtain the initial localization probability vector.

3. The method according to claim 1 or 2, characterized in that, Also includes: Receive the dynamic weight matrix sent by the cloud master station. The dynamic weight matrix is ​​the feedback parameter generated by the cloud master station for the interference segment where the initial positioning probability vector is greater than the first threshold and is confirmed to be without fault. The dynamic weight matrix and the cached real-time state feature vector are subjected to a second forward dot product calculation to generate a secondary localization probability vector. The actual fault section is determined based on the secondary positioning probability vector, a relay protection command is generated, and fault isolation is performed based on the relay protection command.

4. The method according to claim 3, characterized in that, Determining the actual fault segment based on the secondary positioning probability vector includes: The maximum probability value in the secondary positioning probability vector is determined by the maximum value index function, and the segment index corresponding to the maximum probability value is locked as the actual fault segment.

5. A method for collaborative diagnosis of faults in a distribution network, characterized in that, Applied to cloud-based main sites, including: Receive target data reported by edge-side feeder terminals, wherein the target data includes at least an initial location probability vector, which is used to characterize the probability distribution of fault occurrence in each fault section of the distribution network; Obtain multimodal constraint information, which includes at least one of the following: real-time power grid topology status, current meteorological conditions, and historical equipment operation and maintenance records; A confidence score is obtained based on the initial positioning probability vector and the multimodal constraint information; Based on the confidence score, the faulty section that has been confirmed to have failed is identified, and an authorized isolation command is sent to the edge-side feeder terminal.

6. The method according to claim 5, characterized in that, Also includes: The initial positioning probability vector is determined to be greater than the first threshold and is identified as an interference segment where no fault has occurred; Based on the physical and logical conflict between the multimodal constraint information and the initial positioning probability vector, cognitive bias features are extracted, and a spatial penalty matrix and a state compensation vector are constructed. Based on the spatial penalty matrix and the state compensation vector, a dynamic correction is applied to the preset initial weight matrix to generate a dynamic weight matrix; The dynamic weight matrix is ​​sent to the edge-side feeder terminal.

7. The method according to claim 5, characterized in that, Also includes: Receive the fault isolation confirmation message fed back by the edge-side feeder terminal; Update the global physical connectivity status of the distribution network based on the isolated fault sections, and reconstruct the isolated topology; Based on the reconstructed topology matrix, and combined with the load demand of each non-fault outage section, the available capacity of distributed power sources, and the thermal stability margin of the lines, the optimal load transfer path is calculated. Generate collaborative isolation and adaptive power transfer scheduling strategies and push them to the distribution network SCADA system or dispatcher workstation.

8. A collaborative fault diagnosis system for power distribution networks, characterized in that, Including edge-side feeder terminals and cloud master stations; The edge-side feeder terminal is used to perform preliminary fault location, generate an initial location probability vector and report it to the cloud master station, and receive the dynamic weight matrix fed back by the cloud master station and perform secondary fault location and physical isolation. The cloud-based master station is used to receive the initial positioning probability vector, perform a reliability assessment in conjunction with multimodal constraints, selectively generate a dynamic weight matrix based on the assessment results and send it to the edge-side feeder terminal, and perform global topology reconstruction and power restoration decisions based on the isolation results fed back by the edge-side feeder terminal.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 7.

Images