Fault identification method, system and device for distribution network non-power-cut operation and medium

Abstract

The invention is suitable for the technical field of power distribution networks, and provides a fault identification method for power distribution network non-outage operation, and the method comprises the steps: obtaining the monitoring data of a power distribution network at a plurality of visual angles in real time; fusing the monitoring data based on a Transform large model to obtain a fused feature representation, and carrying out fault identification on the distribution network equipment based on the fused feature representation to obtain potential fault data, the fault data including a potential fault point, a potential fault position and a potential fault type; based on the potential fault data, performing quantitative analysis on the distribution of distribution network equipment in the distribution network and the operation state of the distribution network equipment to obtain a fault risk result; dynamically adjusting a distribution network scheduling strategy based on the fault risk result, and generating a job scheduling plan; based on the operation scheduling plan, the communication technology is adopted to enable the power distribution equipment in the distribution network to work cooperatively. The fault identification accuracy and efficiency can be improved, the distribution network scheduling strategy is optimized, the cooperative work of the distribution network equipment is promoted, and the operation and maintenance level of the distribution network is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of power distribution networks, and in particular, relates to a fault identification method, system, device and medium for non-stop power supply operation of power distribution networks. Background Art

[0002] Distribution network non-stop operation is a method of maintenance of the distribution network without interrupting the power supply. This operation method is of great significance for improving power supply reliability, optimizing the business environment and meeting the high requirements of society for power supply. With the acceleration of urbanization and the increase in electricity demand, the impact of distribution network power outages on people's daily life and social production is becoming increasingly significant, and the reliability and continuity of power supply are required to be higher and higher. Therefore, in order to achieve non-stop operation of the distribution network, it is necessary to identify faults in the distribution network in daily life so as to carry out timely maintenance of the distribution network to ensure the stability and reliability of power supply.

[0003] In the prior art, by integrating advanced sensor technology, data acquisition systems and artificial intelligence algorithms, real-time monitoring and intelligent analysis of distribution network operating parameters can be achieved, thereby realizing rapid identification and precise positioning of faults. However, the prior art distribution network fault identification methods still have some shortcomings and limitations. For example, single-perspective images or video data are usually used for fault identification, and multi-angle feature fusion cannot be performed. In addition, the risk prediction of distribution network faults usually relies on expert experience, which is greatly affected by the subjective influence of experts and is prone to deviations, resulting in low accuracy of risk assessment results and low maintenance efficiency of the distribution network.

[0004] Therefore, the present application proposes a method that can quickly predict, identify, accurately locate and efficiently handle distribution network faults. Summary of the invention

[0005] The embodiments of the present application provide a fault identification method, system, device and medium for non-stop operation of distribution networks, which can improve the accuracy and efficiency of fault identification, optimize distribution network scheduling strategies, promote the coordinated work of distribution network equipment, and thus improve the operation and maintenance level of the distribution network.

[0006] In a first aspect, an embodiment of the present application provides a fault identification method for non-stop operation of a distribution network, comprising:

[0007] Arrange multiple monitoring perspectives in the distribution network to obtain monitoring data from multiple perspectives of the distribution network in real time;

[0008] Based on the Transformer large model, the monitoring data is fused to obtain a fused feature representation, and based on the fused feature representation, fault identification is performed on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location, and a potential fault type;

[0009] Based on the potential fault data, quantitatively analyze the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices to obtain fault risk results;

[0010] Based on the fault risk results, dynamically adjust the distribution network scheduling strategy and generate an operation scheduling plan;

[0011] Based on the operation scheduling plan, communication technology is used to enable various distribution equipment in the distribution network to work in coordination.

[0012] Furthermore, the real-time acquisition of monitoring data of the distribution network equipment from multiple perspectives includes:

[0013] According to the structural characteristics and distribution of each distribution network equipment, the spatial position distribution of the key parts of each distribution network equipment is determined through three-dimensional modeling and simulation analysis.

[0014] Calculating the visibility of each of the key parts at different viewing angles to obtain a visibility matrix of the key parts;

[0015] Based on the visibility matrix, construct a viewing angle occlusion map of the current distribution network, and calculate the visibility score of each viewing angle for the key parts;

[0016] Using a dynamic programming algorithm, searching for an optimal perspective combination in the perspective occlusion map, the optimal perspective combination being a perspective combination with a highest total visibility score;

[0017] Based on the optimal viewing angle combination, a monitoring device is arranged, and the monitoring device is used to monitor the distribution network device.

[0018] Furthermore, the monitoring data is fused based on the Transformer large model to obtain a fused feature representation, and based on the fused feature representation, fault identification is performed on the distribution network equipment to obtain potential fault data, including:

[0019] Based on the optimal perspective combination, monitoring images of the distribution network equipment are collected from multiple perspectives to obtain a multi-perspective image data set;

[0020] Using a pre-trained Transformer large model to extract features from the multi-view image dataset to obtain an image feature vector for each view;

[0021] The feature splicing method is used to fuse the image feature vectors of each perspective to obtain the fused feature representation;

[0022] The support vector machine is used to input the fusion feature representation into a pre-trained fault recognition model to obtain potential fault data.

[0023] Furthermore, based on the potential fault data, the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices are quantitatively analyzed to obtain fault risk results, including:

[0024] For each distribution network device in the distribution network, based on the potential fault data, identify the potential fault mode of the distribution network device, perform risk calculation on all potential fault modes in the distribution network device, and obtain a potential fault score of the distribution network device;

[0025] Obtain the operating parameters of each distribution network device, pre-process the operating parameters using an edge computing method, and obtain distribution network operation data;

[0026] According to the time scale, the distribution network operation data is divided according to the time granularity to obtain a granular data set, where the time granularity includes hours, days or weeks;

[0027] Performing feature engineering processing on each of the granular data sets to obtain device attribute characteristics of the distribution network, wherein the device attribute characteristics include device parameters, topology information, and historical operation data;

[0028] A distribution network quantitative model is constructed using a graph neural network algorithm, where the nodes of the graph neural network represent the distribution network devices in the distribution network, and the edges represent the connection relationship between the distribution network devices. The node features include the device parameters and historical operation data of each distribution network device, and the edge features include topological structure information.

[0029] Based on the potential fault score, a loss function is constructed, and the distribution network quantitative model is trained through the loss function, so that the distribution network quantitative model can predict the overall operation trend of the distribution network and generate fault risk results.

[0030] Furthermore, the loss function is used to train the distribution network quantization model so that the distribution network quantization model can predict the overall operation trend of the distribution network, including:

[0031] Acquire the operation scenario data of the distribution network in real time and construct a scenario feature vector, wherein the operation scenario data includes weather forecast data and load forecast data, wherein the weather forecast data includes temperature, humidity and wind speed, and the load forecast data is the predicted value of power consumption within a preset time;

[0032] Obtaining benchmark feature vectors of a plurality of operating condition scenarios preset in the distribution network, wherein each operating condition scenario is correspondingly provided with a similarity threshold and a benchmark model parameter, and the benchmark model parameter is a model parameter determined based on the loss function during the training process of the distribution network quantization model;

[0033] Calculating the similarity between the scene feature vector and the reference feature vector to obtain a reference similarity;

[0034] If there is an operating condition scenario whose similarity threshold is lower than the benchmark similarity, the benchmark model parameters of the operating condition scenario are used to perform quantitative analysis on the current distribution network through the distribution network quantification model;

[0035] If the similarity thresholds of the plurality of operating condition scenarios are all higher than the benchmark similarity, a genetic algorithm is used to generate a new similarity threshold and new benchmark model parameters for the current distribution network, and the new benchmark model parameters are used to perform quantitative analysis on the distribution network through the distribution network quantification model.

[0036] Furthermore, dynamically adjusting the distribution network scheduling strategy based on the fault risk result to generate a job scheduling plan includes:

[0037] Acquire real-time operation data of the distribution network, wherein the real-time operation data includes load data, voltage data and current data;

[0038] Using a power flow calculation algorithm to calculate and analyze the real-time operation data to obtain a power flow calculation result;

[0039] Based on the power flow calculation result and the fault risk result, a decision tree algorithm is used to generate an operation instruction scheme, wherein the operation instruction scheme includes an operation sequence, operation steps and operation parameters;

[0040] Based on simulation calculation and risk assessment, the operation instruction scheme is adjusted to form a distribution network scheduling strategy;

[0041] The distribution network scheduling strategy is converted into an operation scheduling plan, and the operation scheduling plan is used for real-time scheduling and fault handling of the distribution network by on-site operators.

[0042] Furthermore, based on the operation scheduling plan, communication technology is used to enable the distribution equipment in the distribution network to work in coordination, including:

[0043] According to the communication protocol and data characteristics of each distribution network device, the encoding method is dynamically selected to encode the data of different distribution network devices into a unified data format standard;

[0044] Establishing a network topology structure for the plurality of network distribution devices, dynamically sensing changes in the network topology structure by regularly sending detection messages, and updating the network connection relationship in a timely manner;

[0045] For job scheduling, the shortest path algorithm is used to obtain the optimal path for information transmission.

[0046] In a second aspect, an embodiment of the present application provides a fault identification system for non-stop operation of a distribution network, including:

[0047] The first processing module is used to arrange multiple monitoring perspectives in the distribution network and to obtain monitoring data from multiple perspectives of the distribution network in real time;

[0048] The second processing module is used to fuse the monitoring data based on the Transformer large model to obtain a fusion feature representation, and based on the fusion feature representation, perform fault identification on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location and a potential fault type;

[0049] The third processing module is used to quantitatively analyze the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices based on the potential fault data to obtain a fault risk result;

[0050] The fourth processing module is used to dynamically adjust the distribution network scheduling strategy and generate an operation scheduling plan based on the fault risk result;

[0051] The fifth processing module is used to enable the distribution equipment in the distribution network to work in coordination based on the job scheduling plan and using communication technology.

[0052] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned fault identification method for non-stop power supply operation of a distribution network when executing the computer program.

[0053] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, including a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned fault identification method for non-stop power supply operation of a distribution network is implemented.

[0054] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0055] The present application discloses a fault identification method for non-stop operation of a distribution network. By arranging multiple perspectives in the distribution network, it is ensured that the collected multi-perspective data can fully reflect the operating status of the distribution network equipment and provide sufficient information for subsequent fault identification. In addition, by adopting the Transformer large model to fuse the monitoring data, a more accurate and rich representation of equipment characteristics can be obtained, thereby improving the accuracy of fault identification. In addition, by quantitatively analyzing the distribution and operating status of each distribution network equipment in the distribution network, the overall operating trend of the distribution network is predicted, and the distribution network scheduling strategy can be dynamically adjusted to generate a more reasonable operation scheduling plan. At the same time, through communication technology, the distribution equipment in the distribution network can work together during the non-stop operation process, realize remote monitoring and intelligent scheduling of the distribution network equipment, and improve the operating efficiency and management level of the distribution network. Therefore, the present application can improve the accuracy and efficiency of fault identification, optimize the distribution network scheduling strategy, promote the collaborative work of distribution network equipment, and thus improve the operation and maintenance level of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0057] Figure 1 It is a flow chart of a method for fault identification for non-stop operation of a distribution network provided by an embodiment of the present invention;

[0058] Figure 2 It is a structural schematic diagram of a fault identification system for non-stop operation of a distribution network provided by an embodiment of the present invention;

[0059] Figure 3 It is a structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0061] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0062] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0063] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0064] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0065] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0066] See also Figure 1 As shown, the present invention is a fault identification method for non-stop operation of distribution network, comprising the following steps:

[0067] S100, real-time acquisition of monitoring data of distribution network equipment from multiple perspectives;

[0068] In some embodiments, the above step S100 includes:

[0069] According to the structural characteristics and distribution of each distribution network equipment, the spatial position distribution of the key parts of each distribution network equipment is determined through three-dimensional modeling and simulation analysis.

[0070] Calculating the visibility of each of the key parts at different viewing angles to obtain a visibility matrix of the key parts;

[0071] Based on the visibility matrix, construct a viewing angle occlusion map of the current distribution network, and calculate the visibility score of each viewing angle for the key parts;

[0072] Using a dynamic programming algorithm, searching for an optimal perspective combination in the perspective occlusion map, the optimal perspective combination being a perspective combination with a highest total visibility score;

[0073] Based on the optimal viewing angle combination, a monitoring device is arranged, and the monitoring device is used to monitor the distribution network device.

[0074] In this embodiment, for multiple distribution network equipment in the distribution network, which usually includes overhead lines, poles, cables, distribution transformers, switchgear, and reactive compensation capacitors, etc., according to the structural characteristics of different distribution network equipment, multiple monitoring perspectives are arranged in the distribution network to cover the key parts and potential fault areas of each distribution network equipment to the greatest extent, so as to ensure that the collected multi-perspective data can comprehensively reflect the operating status of the distribution network equipment and provide sufficient information for subsequent fault identification.

[0075] In this embodiment, the current distribution network is three-dimensionally modeled, and the corresponding simulation parameters such as material properties, operating rules and external loads are set according to the simulation target. The model is run using simulation software to simulate the operating status of each distribution network device in the actual working environment. During the simulation process, various performance indicators of the distribution network equipment, such as stress and temperature, can be monitored in real time. The simulation results are analyzed in detail, and areas such as stress concentration, severe wear, and excessive temperature in the distribution network equipment are taken as potential problem areas, and the functional requirements of the equipment need to be considered. For example, for overhead lines, the key parts are conductors, insulators, lightning arresters, hardware, etc.; for pole towers, the key parts include tower bodies, tower bases, crossarms, guy wires, etc.; the key parts of cables include conductors, insulation layers, shielding layers, sheaths, etc.; the key parts of distribution transformers include iron cores, windings, oil tanks, cooling devices, etc.; the key parts of switchgear include contacts, arc extinguishing devices, operating mechanisms, etc.; the key parts of reactive compensation capacitors include capacitor units, series reactors (if any), discharge devices, etc.

[0076] In this embodiment, the method for determining the key parts is as follows: the simulation results and functional requirements are combined for comprehensive evaluation. Specifically, according to the functional requirements, importance weights are assigned to each component of the distribution network equipment. The size of the weight reflects the importance of the component in the overall function of the equipment. Combined with the simulation results, the failure probability and failure consequences of each component are analyzed, and the risk level of each component is divided according to the severity of the failure probability and the failure consequences. The importance weights and risk levels of each component are quantified, and the comprehensive scores of each component are calculated. According to the comprehensive scores, the components are sorted, and multiple components with higher comprehensive scores are selected as the key parts of the corresponding distribution network equipment.

[0077] In this embodiment, for the three-dimensional model of each distribution network equipment, the projection area of ​​the key parts of each distribution network equipment at different viewing angles is calculated, and its actual surface area is obtained. The ratio of the projection area to the actual surface area is calculated to obtain the visibility of the key parts corresponding to each distribution network equipment at different viewing angles, and a visibility matrix is ​​constructed based on this.

[0078] Specifically, the visibility of each key part under different viewing angles is calculated. For example, in one embodiment, there are N key parts and M viewing angles, thereby constructing an N×M visibility matrix V, where V nm Indicates the visibility of the nth key part at the mth viewing angle, visibility V nm is a binary value, where 0 means completely invisible, 1 means completely visible, and a continuous value between 0 and 1 indicates the degree of visibility. Based on the above visibility matrix V, a view occlusion map is constructed, which describes the occlusion of key parts under different view angles for visualization analysis. For each view angle m, its visibility score for all key parts is calculated, where the calculation formula of visibility score is

[0079]

[0080] Among them, w n is the weight of the nth key part.

[0081] The entropy weight method is used to determine the weight of each key part. The entropy weight method is an objective weighting method based on information entropy. It completely relies on the characteristics of the data itself to determine the weight, avoiding the interference of subjective factors. By using the data of each key part in terms of functional requirements, safety risks, maintenance costs, etc., the weight of each part is determined by calculating the information entropy. Specifically, an evaluation matrix P containing N samples (i.e., key parts) and J evaluation indicators (such as functional requirements, safety risks, maintenance costs, etc.) is constructed. Among them, P jnIndicates the number of indicators of the nth key part under the jth evaluation index, calculates the proportion of the nth sample under the jth evaluation index, and the above proportion reflects the relative importance of the key part under the index. The calculation formula of the proportion is: proportion = (the index value of the sample / the sum of all samples under the index) × 100%; then according to the definition and properties of entropy, calculate the information entropy and redundancy of each index. Specifically, the information entropy reflects the discrete degree of the index value. The smaller the information entropy, the greater the discrete degree of the index, that is, the greater the influence of the index on the comprehensive evaluation (i.e., weight), and the redundancy (coefficient of variation) is the complement of the information entropy, which reflects the variation degree of the index value. The greater the redundancy, the greater the variation degree of the index, that is, the greater the influence of the index on the comprehensive evaluation.

[0082] Specifically, the calculation formula of information entropy is:

[0083] e_j=-k*Σ(p_ij*ln(p_ij))

[0084] Among them, e_j represents the information entropy of the j-th evaluation index, k is a constant, usually 1 / ln(N), N is the number of samples, that is, the number of key parts, and p_nj represents the proportion of the n-th sample under the j-th evaluation index.

[0085] The calculation formula of redundancy (coefficient of difference) is:

[0086] d_j=1-e_j

[0087] Among them, d_j represents the redundancy (coefficient of variation) of the j-th rating indicator.

[0088] The calculation formula for the weight of each evaluation index is:

[0089] w_j=d_j / Σd_j

[0090] Among them, w_j represents the weight of the j-th evaluation indicator, and Σd_j represents the sum of the redundancy (coefficient of difference) of all evaluation indicators.

[0091] Finally, the calculation formula for the weight of each key part is:

[0092]

[0093] In this embodiment, the perspective combination selection problem is transformed into an optimal path search problem based on the occlusion map. Under the premise of satisfying the constraint on the number of perspectives, the perspective combination with the largest total visibility score is selected. Specifically, a dynamic programming algorithm is used to search for the optimal perspective combination in the occlusion map to obtain candidate perspective combinations that meet the conditions. From the candidate perspective combinations, the combination with the highest visibility score is selected as the final optimal perspective combination to maximize the visibility of key parts.

[0094] In this embodiment, the view occlusion graph is a directed graph G=(M, E), where M is a view set, E is a set of edges, and each edge e=(m i , m j ) represents the value from the viewing angle m i You can see the viewing angle m j , that is, m j Not m i Occlusion, each view m i There is a visibility score Score m , the number of selected perspectives cannot exceed K. Based on this, a dynamic programming algorithm is used to solve the optimal perspective combination.

[0095] In this embodiment, according to the optimal viewing angle arrangement scheme and the on-site environmental conditions, the constraints and optimization goals of the light source arrangement are obtained, and the lighting simulation technology is used to simulate the lighting effects for different light source arrangement schemes to obtain the image quality evaluation results under each scheme; according to the image quality evaluation results, it is judged whether the image data quality requirements for fault identification are met. If not, the light source parameters are adjusted and the lighting simulation is performed again until the requirements are met; according to the constraints of the on-site installation conditions, such as the space size and the installation position, the physical installation scheme of the viewing angle is designed, including the bracket design and the installation angle, etc.; the physical installation scheme is modeled by using three-dimensional modeling technology, and the mechanical performance simulation is performed on it through finite element analysis to judge the feasibility and stability of the installation scheme; if the finite element analysis results show that the installation scheme has mechanical performance problems, the physical installation scheme needs to be optimized and adjusted, and the modeling and simulation analysis are performed again until the requirements are met; according to the results of the lighting simulation and finite element analysis, the final light source arrangement scheme and physical installation scheme are determined, and detailed design documents and construction drawings are generated to guide on-site installation and debugging.

[0096] S200, fusing the monitoring data based on the Transformer large model to obtain a fusion feature representation, and based on the fusion feature representation, performing fault identification on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location, and a potential fault type;

[0097] In this embodiment, the Transformer large model is used to fuse the monitoring data to obtain a more accurate and rich representation of equipment features, thereby improving the accuracy of fault identification.

[0098] In some embodiments, the step S200 includes:

[0099] Based on the optimal perspective combination, monitoring images of the distribution network equipment are collected from multiple perspectives to obtain a multi-perspective image data set;

[0100] Using a pre-trained Transformer large model to extract features from the multi-view image dataset to obtain an image feature vector for each view;

[0101] The feature splicing method is used to fuse the image feature vectors of each perspective to obtain the fused feature representation;

[0102] The support vector machine is used to input the fusion feature representation into a pre-trained fault recognition model to obtain potential fault data.

[0103] In this embodiment, images of distribution network equipment are collected from multiple perspectives to obtain a multi-perspective image data set, and the Transformer large model is used to extract features from the multi-perspective image data set to obtain an image feature vector for each perspective; the image feature vectors for each perspective are fused by feature splicing to obtain a fused feature representation. Specifically, for the pre-trained Transformer large model, the Vision Transformer model is used mainly through the powerful feature representation capability of the Transformer large model to cut the images in the multi-perspective image data set into a sequence of blocks of equal size, linearly embed the block sequence of each image and add position information, and then perform feature crossover through the standard Transformer encoder structure to capture the key information in the image and generate the corresponding image feature vector.

[0104] For the fault recognition model, a historical multi-view image data set and its corresponding fault data when the fault occurs are obtained to train the above fault recognition model, wherein the fault data includes the fault point, fault location and fault type. When training, the above multi-view image data set needs to be preprocessed, such as using data enhancement technology to expand the multi-view image data set to obtain an expanded data set, wherein the data expansion method includes operations such as rotation, scaling and translation; then, feature extraction and fusion are performed on the image data in the expanded data set to obtain a fused feature representation, and the fault recognition model is trained by the fused feature representation and the corresponding fault data. Specifically, a support vector machine is used as a fault recognition model, and the fused feature representation is used as input and the fault data is used as output. The support vector machine model, i.e., the fault recognition model, is trained to output the potential fault data in the current distribution network according to the input multi-view image data set, which is used for the staff to warn and repair the distribution network in advance. Specifically, when the potential fault data is obtained through the fault recognition model, the specific coordinates or area of ​​the potential fault position of the distribution network are determined in combination with the model diagram constructed by three-dimensional modeling in the above step S100.

[0105] S300, based on the potential fault data, quantitatively analyzing the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices to obtain a fault risk result;

[0106] In this embodiment, by quantitatively analyzing the distribution and operating status of each distribution network device, the overall operating trend of the distribution network is predicted, and the distribution network scheduling strategy can be dynamically adjusted to generate a more reasonable operation scheduling plan.

[0107] In some embodiments, the step S300 includes:

[0108] For each distribution network device in the distribution network, based on the potential fault data, identify the potential fault mode of the distribution network device, perform risk calculation on all potential fault modes in the distribution network device, and obtain a potential fault score of the distribution network device;

[0109] Obtain the operating parameters of each distribution network device, pre-process the operating parameters using an edge computing method, and obtain distribution network operation data;

[0110] According to the time scale, the distribution network operation data is divided according to the time granularity to obtain a granular data set, where the time granularity includes hours, days or weeks;

[0111] Performing feature engineering processing on each of the granular data sets to obtain device attribute characteristics of the distribution network, wherein the device attribute characteristics include device parameters, topology information, and historical operation data;

[0112] A distribution network quantitative model is constructed using a graph neural network algorithm, where the nodes of the graph neural network represent the distribution network devices in the distribution network, and the edges represent the connection relationship between the distribution network devices. The node features include the device parameters and historical operation data of each distribution network device, and the edge features include topological structure information.

[0113] Based on the potential fault score, a loss function is constructed, and the distribution network quantitative model is trained through the loss function, so that the distribution network quantitative model can predict the overall operation trend of the distribution network and generate fault risk results.

[0114] In this embodiment, for the fault mode, the fault data of historical faults and the operating data of each distribution network equipment when the fault occurs are collected from the monitoring system of the distribution network, including but not limited to parameters such as current, voltage, power, temperature, humidity, etc., and the noise, duplicate values ​​and missing values ​​in the data are removed to ensure the accuracy and completeness of the data. The data is standardized and normalized for subsequent analysis, and cluster analysis is performed on the pre-processed data to obtain multiple clusters. According to the clustering results, combined with the operating characteristics of the distribution network equipment and historical fault records, the potential fault mode represented by each cluster is explained.

[0115] In this embodiment, for each distribution network device in the distribution network, the corresponding potential fault mode is obtained based on the potential fault data and the operation data of the current distribution network device, so as to facilitate the subsequent fault analysis and prediction of the distribution network. In addition, for each distribution network device, a severity score is assigned to each potential fault mode according to the severity evaluation standard, and the frequency score of each potential fault mode is calculated according to the frequency evaluation standard. According to the detection evaluation standard, a detection score is assigned to each potential fault mode, and the risk score of the corresponding fault mode is calculated: h = S × O × D, where S represents the severity score, O represents the frequency score, and D represents the detection score. The risk calculation is performed for each potential fault mode of the distribution network device, and the risk score of each potential fault mode is added to obtain the potential fault score of the distribution network device.

[0116] In this embodiment, sensors and edge computing nodes are deployed at key locations of distribution network equipment. Sensors are used to monitor the operating parameters of the equipment in real time, such as current, voltage, power, temperature, humidity, etc. A data acquisition system is established to transmit the data collected by the sensors to the edge computing nodes by wired or wireless means. Sufficient computing power and storage resources are configured for the edge computing nodes. The data collected by the sensors is cleaned, deduplicated, normalized, and other preprocessing is performed to generate distribution network operation data to meet data processing requirements. By utilizing edge computing technology, real-time and efficient preprocessing of the operating parameters of distribution network equipment is achieved, providing strong support for subsequent data analysis and decision support.

[0117] In this embodiment, the distribution network operation data is divided into multiple granular data sets according to the time scale (such as hours, days, weeks, etc.), and feature engineering is performed on each granular data set to extract features such as equipment parameters, topological structure, and historical operation data, thereby reducing the input dimension of the distribution network quantization model and reducing the complexity of the distribution network quantization model, thereby improving the training speed and prediction performance of the model.

[0118] In this embodiment, a graph neural network algorithm is used to construct a distribution network quantitative model, which can capture the complex relationships in the distribution network. In the graph neural network structure, nodes represent various distribution network devices in the distribution network, edges represent the connection relationship between various distribution network devices, node features include device parameters and historical operation data, and edge features include topological structure information.

[0119] In this embodiment, a loss function related to the potential fault score is constructed for the training of the above-mentioned distribution network quantitative model, so that the prediction of the overall operation trend of the distribution network by the distribution network quantitative model finally trained can be strongly correlated with the potential fault data, thereby improving the accuracy of the prediction. It can be understood that if the instantaneous function value of the loss function is less than the preset threshold, the training of the distribution network quantitative model is completed. At this time, the model parameters of the distribution network quantitative model are the benchmark model parameters, which are used to perform quantitative analysis on the current distribution network. If the loss function value is greater than or equal to the preset threshold, the training continues.

[0120] Specifically, the distribution network operation data to be predicted is obtained, the distribution network operation data is divided into time granularity and feature engineering is performed to obtain the device attribute characteristics, and the device attribute characteristics are input into the trained distribution network quantization model to obtain the global feature representation and local feature representation of the distribution network. According to the global feature representation of the distribution network, the overall operation trend of the distribution network is judged. For example, if the load dimension value represented by the global feature continues to rise, it is judged that the load of the distribution network continues to increase. In addition, according to the local feature representation, the health status of the distribution network equipment is judged. For example, there is a transformer whose local feature representation is significantly different from its feature representation in its historical normal state, such as the voltage mean value drops by 10V and the current mean value increases by 5A, and it is judged that the transformer may be faulty. Based on the above judgment results, a distribution network operation status report is generated, including the overall operation trend of the distribution network and the health status information of each distribution network device, which is used to guide the distribution network operation and maintenance and decision-making.

[0121] In some embodiments, the loss function includes a basic loss function and a potential fault score loss function, and the loss function is calculated as follows:

[0122] L total =αL basic +βL fault

[0123] Among them, L total represents the loss function, L basic represents the basic loss function, L fault represents the potential fault score loss, α and β are weight coefficients used to adjust the contribution of the basic loss function and the potential fault score loss in the total loss function;

[0124] For the basic loss function, including the trend loss function and the robust loss function, the calculation formula is:

[0125] L basic =L(y)+L(a)

[0126] Among them, L(y) represents the trend loss function, and L(a) represents the robustness loss function;

[0127]

[0128] Among them, Δy t =y t -y t-1 , represents the difference between the actual load data at time t and t-1, represents the difference between the estimated load data at time t and t-1, Represents the prediction error, and δ is a threshold parameter used to control the switching of the robustness loss function between the mean square error and the mean absolute error.

[0129] As you can understand, the robustness loss function is Huber loss, which combines the advantages of mean square error (MSE) and mean absolute error (MAE). For small errors, it uses MSE (sensitive to errors), and for large errors, it uses MAE (robust to outliers).

[0130] For the potential fault score loss function, the calculation formula is:

[0131]

[0132] Devices is a collection of network distribution devices. d represents the weight of the distribution network device d, and H represents the potential fault score of the distribution network device d.

[0133] In some embodiments, the loss function is used to train the distribution network quantization model so that the distribution network quantization model can predict the overall operation trend of the distribution network, including:

[0134] Acquire the operation scenario data of the distribution network in real time and construct a scenario feature vector, wherein the operation scenario data includes weather forecast data and load forecast data, wherein the weather forecast data includes temperature, humidity and wind speed, and the load forecast data is the predicted value of power consumption within a preset time;

[0135] Obtaining benchmark feature vectors of a plurality of operating condition scenarios preset in the distribution network, wherein each operating condition scenario is correspondingly provided with a similarity threshold and a benchmark model parameter, and the benchmark model parameter is a model parameter determined based on the loss function during the training process of the distribution network quantization model;

[0136] Calculating the similarity between the scene feature vector and the reference feature vector to obtain a reference similarity;

[0137] If there is an operating condition scenario whose similarity threshold is lower than the benchmark similarity, the benchmark model parameters of the operating condition scenario are used to perform quantitative analysis on the current distribution network through the distribution network quantification model;

[0138] If the similarity thresholds of the plurality of operating condition scenarios are all higher than the benchmark similarity, a genetic algorithm is used to generate a new similarity threshold and new benchmark model parameters for the current distribution network, and the new benchmark model parameters are used to perform quantitative analysis on the distribution network through the distribution network quantification model.

[0139] In this embodiment, the operating status of the distribution network is also affected by multiple factors such as weather, load, and faults. Therefore, the distribution network quantitative model needs to consider the weather and load conditions in the corresponding scenario. Specifically, the weather forecast data within the preset time is obtained from the meteorological department, including information such as temperature, humidity, and wind speed. At the same time, the load forecast data within the preset time is obtained from the power dispatching department, and the obtained weather forecast data and load forecast data are feature vectorized to generate a scenario feature vector.

[0140] In this embodiment, the K-means clustering algorithm is used to perform cluster analysis on multiple scene feature vectors. By calculating the silhouette coefficient and the SSE value, the optimal number of clusters is selected. In a preferred embodiment, the number of clusters is 5, that is, the operating condition scene is divided into 5 scenes. Each operating condition scene is provided with a similarity threshold and a benchmark model parameter. The similarity threshold is used to compare the real-time scene feature vector of the distribution network with the benchmark feature vector corresponding to the operating condition scene to determine whether it is necessary to update the model parameters of the distribution network quantitative model according to the current operating scene data to improve the accuracy of the model. Specifically, for the weather forecast data and load forecast data obtained in real time, a scene feature vector is constructed, and the cosine similarity is used to calculate the benchmark similarity between it and the benchmark feature vector in the preset operating condition scene. If the similarity is greater than the similarity threshold in the corresponding operating condition scene, it is considered that the current operating scene belongs to the operating condition scene, and the current distribution network can be quantitatively analyzed with the corresponding model parameters in the operating condition scene. If the benchmark similarity is less than the similarity threshold in each preset operating condition scenario, the genetic algorithm is used to optimize the solution of new model parameters and set a new similarity threshold, which is then used as a new operating condition scenario to enrich the operating scenarios of the distribution network and provide a decision-making basis for the distribution network scheduling operation.

[0141] S400, dynamically adjusting the distribution network scheduling strategy based on the fault risk result and generating an operation scheduling plan;

[0142] In some embodiments, the step S400 includes:

[0143] Acquire real-time operation data of the distribution network, wherein the real-time operation data includes load data, voltage data and current data;

[0144] Using a power flow calculation algorithm to calculate and analyze the real-time operation data to obtain a power flow calculation result;

[0145] Based on the power flow calculation result and the fault risk result, a decision tree algorithm is used to generate an operation instruction scheme, wherein the operation instruction scheme includes an operation sequence, operation steps and operation parameters;

[0146] Based on simulation calculation and risk assessment, the operation instruction scheme is adjusted to form a distribution network scheduling strategy;

[0147] The distribution network scheduling strategy is converted into an operation scheduling plan, and the operation scheduling plan is used for real-time scheduling and fault handling of the distribution network by on-site operators.

[0148] In this embodiment, the acquired real-time operation data is used as the input of the power flow calculation, and the Newton-Raphson method is used for iterative calculation to obtain the operation status of the current distribution network. Specifically, based on step S300, the graph neural network structure and the electrical characteristics of each distribution network device in the distribution network are constructed to establish a mathematical model of the power grid. A suitable initial voltage value is selected for each node in the graph neural network structure, wherein the selection of the initial value determines the convergence speed and stability of the iteration. The real-time operation data is substituted into the mathematical model of the power flow calculation, and the elements of the Jacobian matrix are calculated according to the correction equation. The correction vector is solved using the Jacobian matrix and the constant term vector. According to the correction vector, the voltage value of the node is updated. It is determined whether the iteration converges. If converged, the calculation result is output; if not converged, the new node voltage value is used as the initial value, and the above iterative process is repeated. Finally, the power flow distribution outputs the voltage, phase angle of each node in the power grid and the power distribution of each branch. According to the calculation results, the operation status of the power grid is analyzed, including load distribution, voltage stability and power balance.

[0149] In this embodiment, key features such as node voltage deviation, branch power overload degree, fault point, fault location, fault type, etc. are extracted from the power flow calculation results and the fault risk analysis results. The decision tree model is trained by the extracted feature data. Specifically, the decision tree model is divided by recursively selecting the optimal features to form a tree structure. Each internal node represents a test of a feature or attribute, each branch represents a value of a feature, and each leaf node represents a job instruction scheme. The power flow calculation results and the fault risk analysis are input into the trained decision tree model. According to the extracted feature data, the corresponding path is traversed and selected in the decision tree until the leaf node is reached, and the job instruction scheme is extracted from the leaf node, including the job sequence, job steps and job parameters.

[0150] In this embodiment, through simulation calculation and risk assessment, potential risks and optimization space in the distribution network can be identified more accurately, thereby improving scheduling efficiency. The optimized operation instruction scheme can reduce fault recovery time and load transfer time, improve power supply reliability and user satisfaction, and through refined scheduling and optimized operation sequence, it can reduce equipment loss and operation and maintenance costs, and improve the economy of the distribution network.

[0151] S500: Based on the operation scheduling plan, communication technology is used to enable the distribution equipment in the distribution network to work in coordination.

[0152] In this embodiment, communication technology is used to enable the distribution equipment in the distribution network to work together during the non-stop operation process, realize remote monitoring and intelligent scheduling of distribution network equipment, and improve the operation efficiency and management level of the distribution network.

[0153] In some embodiments, the step S500 includes:

[0154] According to the communication protocol and data characteristics of each distribution network device, the encoding method is dynamically selected to encode the data of different distribution network devices into a unified data format standard;

[0155] Establishing a network topology structure for the plurality of network distribution devices, dynamically sensing changes in the network topology structure by regularly sending detection messages, and updating the network connection relationship in a timely manner;

[0156] For job scheduling, the shortest path algorithm is used to obtain the optimal path for information transmission.

[0157] In this embodiment, the data characteristics and communication protocol types of the transmission data of each distribution network device in the distribution network are obtained, and the data encoding and decoding methods corresponding to each device are determined according to the pre-established device protocol mapping table. The data of each device is encoded using a determined encoding method, and the encoded data is encapsulated into a unified data format, and metadata fields such as device identification, data type, and timestamp are added. The encapsulated data is sent to the target device through the corresponding communication protocol. After the target device receives the data, it decodes the data using the corresponding decoding method according to the information in the metadata field to obtain the original data.

[0158] In some embodiments, a data format standard library is established to define the XML or JSON structure of each type of data, as well as the name, type, value range, etc. of each field, as a unified standard for information exchange between various network distribution devices and communication protocols. If a new device type or communication protocol appears, the corresponding encoder and decoder programs are generated according to the definition of the data format standard library to achieve rapid integration of new devices or protocols.

[0159] In this embodiment, a data structure of network topology information is generated according to a unified data format, including attributes such as node identification and connection relationship. At the same time, the format and content of the detection message are preset and sent regularly in the network to dynamically perceive changes in the network topology. When the detection message is received, the message content is parsed and the locally maintained network topology information is updated to ensure the accuracy of the connection relationship. If a new node is detected to join, the new node information is sent to other nodes through a broadcast notification mechanism to enable them to update their routing strategies. When a node exits, other nodes are also notified through broadcast to adjust the route, thereby maintaining the connectivity and stability of the network.

[0160] In some embodiments, during the node communication process, an identity authentication mechanism is used to verify the legitimacy of the node identity through a digital certificate or a pre-shared key. The communication data is encrypted using a symmetric encryption algorithm (such as AES) or an asymmetric encryption algorithm (such as RSA) to ensure the security of data transmission and prevent illegal nodes from eavesdropping or tampering with data.

[0161] In this embodiment, according to the fault location range displayed in the job scheduling plan, multiple possible paths of the fault node are searched in the network topology diagram, and the Dijkstra algorithm is used to calculate the cost of each path to obtain the optimal path for transmitting the fault information. The fault information is packaged and transmitted to the network management center through the optimal path. At the same time, to ensure reliability, the same copy of the fault information is transmitted in parallel on other alternative paths. After receiving the fault information, the network management center analyzes the fault type and location. If the fault information of multiple paths is consistent, the corresponding fault handling process is started according to the fault type to ensure the synchronous transmission of the fault information. If the fault information returned by multiple paths is inconsistent, it is judged that the node may have a Byzantine error, the error node is isolated, and subsequent processing is performed according to the consistent fault information, thereby improving the parallelism and scalability of each distribution equipment in the distribution network.

[0162] See also Figure 2 As shown, the present invention also provides a fault identification system for non-stop operation of distribution network, the system comprising:

[0163] The first processing module 201 is used to arrange multiple monitoring perspectives in the distribution network and to obtain monitoring data from multiple perspectives of the distribution network in real time;

[0164] The second processing module 202 is used to fuse the monitoring data based on the Transformer large model to obtain a fused feature representation, and based on the fused feature representation, perform fault identification on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location, and a potential fault type;

[0165] The third processing module 203 is used to quantitatively analyze the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices based on the potential fault data to obtain a fault risk result;

[0166] The fourth processing module 204 is used to dynamically adjust the distribution network scheduling strategy based on the fault risk result and generate a job scheduling plan;

[0167] The fifth processing module 205 is used to enable the distribution equipment in the distribution network to work in coordination based on the operation scheduling plan and using communication technology.

[0168] It is understandable that if Figure 1 The contents of the fault identification method embodiment for non-stop power supply operation of the distribution network shown in the figure are applicable to the fault identification system embodiment for non-stop power supply operation of the distribution network. The functions specifically implemented by the fault identification system embodiment for non-stop power supply operation of the distribution network are the same as those in the embodiment shown in the figure. Figure 1 The fault identification method embodiment for non-stop power supply operation of the distribution network shown in the figure is the same as that of the embodiment shown in the figure, and the beneficial effects achieved are the same as those of the embodiment shown in the figure. Figure 1 The beneficial effects achieved by the embodiment of the fault identification method for non-stop operation of a distribution network shown are also the same.

[0169] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0170] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0171] See also Figure 3As shown, an embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, a fault identification method for non-stop power supply operation of a distribution network as described in any of the above methods is implemented.

[0172] The computer device 3 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will appreciate that Figure 3 It is only an example of computer device 3 and does not constitute a limitation on computer device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, etc.

[0173] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0174] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 3. Further, the memory 302 may also include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0175] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the fault identification method for non-stop operation of a distribution network as described in any one of the above methods is implemented.

[0176] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / computer device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, USB flash drive, mobile hard disk, disk or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0177] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A fault identification method for non-stop power supply operation of a distribution network, characterized in that: include: Arrange multiple monitoring perspectives in the distribution network to obtain monitoring data from multiple perspectives of the distribution network in real time; Based on the Transformer large model, the monitoring data is fused to obtain a fused feature representation, and based on the fused feature representation, fault identification is performed on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location, and a potential fault type; Based on the potential fault data, quantitatively analyze the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices to obtain fault risk results; Based on the fault risk results, dynamically adjust the distribution network scheduling strategy and generate an operation scheduling plan; Based on the operation scheduling plan, communication technology is used to enable various distribution equipment in the distribution network to work in coordination.

2. The method according to claim 1, characterized in that Arranging multiple monitoring perspectives in the distribution network to obtain monitoring data from multiple perspectives of the distribution network in real time includes: According to the structural characteristics of each distribution network device and the distribution of each distribution network device, the spatial position distribution of the key parts of each distribution network device is determined through three-dimensional modeling and simulation analysis; Calculating the visibility of each of the key parts at different viewing angles to obtain a visibility matrix of the key parts; Based on the visibility matrix, construct a viewing angle occlusion map of the current distribution network, and calculate the visibility score of each viewing angle for the key parts; Using a dynamic programming algorithm, searching for an optimal perspective combination in the perspective occlusion map, the optimal perspective combination being a perspective combination with a highest total visibility score; Based on the optimal viewing angle combination, a monitoring device is arranged, and the monitoring device is used to monitor the distribution network device.

3. The method according to claim 2, characterized in that The step of fusing the monitoring data based on the Transformer large model to obtain a fused feature representation, and performing fault identification on the distribution network equipment based on the fused feature representation to obtain potential fault data includes: Based on the optimal perspective combination, monitoring images of the distribution network equipment are collected from multiple perspectives to obtain a multi-perspective image data set; Using a pre-trained Transformer large model to extract features from the multi-view image dataset to obtain an image feature vector for each view; The feature splicing method is used to fuse the image feature vectors of each perspective to obtain the fused feature representation; The support vector machine is used to input the fusion feature representation into a pre-trained fault recognition model to obtain potential fault data.

4. The method according to claim 1, characterized in that The method of quantitatively analyzing the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices based on the potential fault data to obtain a fault risk result includes: For each distribution network device in the distribution network, based on the potential fault data, identify the potential fault mode of the distribution network device, perform risk calculation on all potential fault modes in the distribution network device, and obtain a potential fault score of the distribution network device; Obtain the operating parameters of each distribution network device, pre-process the operating parameters using an edge computing method, and obtain distribution network operation data; According to the time scale, the distribution network operation data is divided according to the time granularity to obtain a granular data set, where the time granularity includes hours, days or weeks; Performing feature engineering processing on each of the granular data sets to obtain device attribute characteristics of the distribution network, wherein the device attribute characteristics include device parameters, topology information, and historical operation data; A distribution network quantitative model is constructed using a graph neural network algorithm, where the nodes of the graph neural network represent the distribution network devices in the distribution network, and the edges represent the connection relationship between the distribution network devices. The node features include the device parameters and historical operation data of each distribution network device, and the edge features include topological structure information. Based on the potential fault score, a loss function is constructed, and the distribution network quantitative model is trained through the loss function, so that the distribution network quantitative model can predict the overall operation trend of the distribution network and generate fault risk results.

5. The method according to claim 4, characterized in that The loss function is used to train the distribution network quantization model so that the distribution network quantization model can predict the overall operation trend of the distribution network, including: Acquire the operation scenario data of the distribution network in real time and construct a scenario feature vector, wherein the operation scenario data includes weather forecast data and load forecast data, wherein the weather forecast data includes temperature, humidity and wind speed, and the load forecast data is the predicted value of power consumption within a preset time; Obtaining benchmark feature vectors of a plurality of operating condition scenarios preset in the distribution network, wherein each operating condition scenario is correspondingly provided with a similarity threshold and a benchmark model parameter, and the benchmark model parameter is a model parameter determined based on the loss function during the training process of the distribution network quantization model; Calculating the similarity between the scene feature vector and the reference feature vector to obtain a reference similarity; If there is an operating condition scenario whose similarity threshold is lower than the benchmark similarity, the benchmark model parameters of the operating condition scenario are used to perform quantitative analysis on the current distribution network through the distribution network quantification model; If the similarity thresholds of the plurality of operating condition scenarios are all higher than the benchmark similarity, a genetic algorithm is used to generate a new similarity threshold and new benchmark model parameters for the current distribution network, and the new benchmark model parameters are used to perform quantitative analysis on the distribution network through the distribution network quantification model.

6. The method according to claim 1, characterized in that The dynamically adjusting the distribution network scheduling strategy based on the fault risk result and generating the operation scheduling plan includes: Acquire real-time operation data of the distribution network, wherein the real-time operation data includes load data, voltage data and current data; Using a power flow calculation algorithm to calculate and analyze the real-time operation data to obtain a power flow calculation result; Based on the power flow calculation result and the fault risk result, a decision tree algorithm is used to generate an operation instruction scheme, wherein the operation instruction scheme includes an operation sequence, operation steps and operation parameters; Based on simulation calculation and risk assessment, the operation instruction scheme is adjusted to form a distribution network scheduling strategy; The distribution network scheduling strategy is converted into an operation scheduling plan, and the operation scheduling plan is used for real-time scheduling and fault handling of the distribution network by on-site operators.

7. The method according to claim 1, characterized in that The method of using communication technology based on the operation scheduling plan to enable the distribution equipment in the distribution network to work in coordination includes: According to the communication protocol and data characteristics of each distribution network device, the encoding method is dynamically selected to encode the data of different distribution network devices into a unified data format standard; Establishing a network topology structure for the plurality of network distribution devices, dynamically sensing changes in the network topology structure by regularly sending detection messages, and updating the network connection relationship in a timely manner; For job scheduling, the shortest path algorithm is used to obtain the optimal path for information transmission.

8. A fault identification system for non-stop power supply operation of distribution network, characterized in that: include: The first processing module is used to obtain monitoring data of distribution network equipment from multiple perspectives in real time; The second processing module is used to fuse the monitoring data based on the Transformer large model to obtain a fusion feature representation, and based on the fusion feature representation, perform fault identification on the distribution network equipment to obtain potential fault data, wherein the fault data includes a potential fault point, a potential fault location and a potential fault type; The third processing module is used to quantitatively analyze the distribution of distribution network devices in the distribution network and the operating status of the distribution network devices based on the potential fault data to obtain a fault risk result; The fourth processing module is used to dynamically adjust the distribution network scheduling strategy and generate an operation scheduling plan based on the fault risk result; The fifth processing module is used to enable the distribution equipment in the distribution network to work in coordination based on the job scheduling plan and using communication technology.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images