Distribution network fault positioning method and system based on intelligent decision

Abstract

The invention provides an intelligent decision-based distribution network fault positioning method and system, and the method comprises the steps: obtaining a target state data set collected by a power distribution network global intelligent perception network, carrying out the time-space semantic fusion processing, generating a multi-dimensional feature map, carrying out the context perception discrimination of the multi-dimensional feature map through a dynamic decision inference engine, and carrying out the fault positioning of a distribution network. Generating a fault risk judgment result, performing causal link tracking on the target state data set based on the fault risk judgment result, generating a fault section positioning result, generating a fault positioning instruction according to the fault section positioning result, sending the fault positioning instruction to the distribution network intelligent operation and maintenance platform, and triggering a precise maintenance process. Therefore, the global state data of the power distribution network are comprehensively utilized, the accurate evaluation of the fault risk of the power distribution network, the rapid positioning of the fault section and the efficient triggering of the accurate maintenance process are realized, and the operation reliability and safety of the power distribution network are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network operation and maintenance, and in particular, to a distribution network fault location method and system based on intelligent decision-making. Background Art

[0002] In the operation and management of the distribution network, fault location is a key link to ensure the safe and stable operation of the power grid. Traditional distribution network fault location methods mainly rely on means such as manual inspection, experience judgment, and simple fault indicators. These methods have problems such as low positioning accuracy, slow response speed, and inability to comprehensively perceive the operation state of the power grid. With the continuous expansion of the scale and complexity of the distribution network, as well as the large-scale access of new energy and distributed power sources, the operation environment of the distribution network has become more complex and changeable, and traditional fault location methods are difficult to meet the management requirements of modern distribution networks. Although existing intelligent technology-based fault location methods have improved the accuracy and efficiency of positioning to a certain extent, most of them only focus on single-type data or features, lack comprehensive analysis and spatio-temporal semantic fusion processing of the entire distribution network state data, and are difficult to accurately reflect the dynamic operation trend and fault risk evolution process of the distribution network, resulting in inaccurate and incomplete fault location results. Summary of the Invention

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present invention, embodiments of the present invention provide a distribution network fault location method based on intelligent decision-making, and the method includes: Obtain a target state data set collected by the global intelligent perception network of the distribution network, where the target state data set includes equipment operation condition data, network topology structure data, historical fault knowledge base, and environmental interference data; Perform spatio-temporal semantic fusion processing on the target state data set to generate a multi-dimensional feature map reflecting the dynamic operation trend of the distribution network, where the multi-dimensional feature map includes a state evolution trajectory in the time dimension and an equipment collaboration relationship in the space dimension; Perform context-aware discrimination processing on the multi-dimensional feature map through a pre-constructed dynamic decision-making inference engine to generate a fault risk discrimination result of the distribution network, where the fault risk discrimination result includes a risk level identifier, a risk trigger period, and a risk impact range; Perform causal link tracking processing on the target state data set based on the fault risk discrimination result to generate a fault section location result of the distribution network, where the fault section location result includes a risk propagation path, a key impact node, and an impact degree quantization value; Generate a fault location instruction including a path identifier, a node identifier, and an impact degree according to the fault section location result, and send the fault location instruction to the distribution network intelligent operation and maintenance platform to trigger a maintenance process.

[0004] In another aspect, an embodiment of the present invention further provides a distribution network fault location system based on intelligent decision-making, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions, or codes, and the processor is used to execute the programs, instructions, or codes in the machine-readable storage medium to implement the above method.

[0005] Based on the above aspects, the embodiment of the present invention obtains a set of target state data collected by the global intelligent perception network of the distribution network, which covers various data such as equipment operation conditions, network topology structure, historical fault knowledge base, and environmental interference. The target state data set is subjected to spatio-temporal semantic fusion processing to generate a multi-dimensional feature map reflecting the dynamic operation trend of the distribution network, which can accurately describe the operation characteristics of the distribution network in different time and space dimensions and the equipment cooperation relationship. Through the pre-constructed dynamic decision-making inference engine, the multi-dimensional feature map is subjected to context-aware discrimination processing to generate a fault risk discrimination result including risk level identification, risk trigger time period, and risk impact range, realizing the accurate assessment and early warning of the distribution network fault risk. Based on the fault risk discrimination result, the causal link tracking processing is performed on the target state data set to generate a fault section location result including the risk propagation path, key impact nodes, and impact degree quantization value, further improving the accuracy and pertinence of the fault location. Finally, according to the fault section location result, a fault location instruction is generated and sent to the distribution network intelligent operation and maintenance platform to trigger the precise maintenance process, realizing the rapid location and efficient processing of the distribution network fault, and significantly improving the operation reliability and safety of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a schematic flow chart of the execution of the distribution network fault location method based on intelligent decision-making provided by the embodiment of the present invention.

[0007] Figure 2 It is a schematic diagram of the exemplary hardware and software components of the distribution network fault location system based on intelligent decision-making provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0008] The present invention will be specifically described below in conjunction with the accompanying drawings of the specification. Figure 1 It is a schematic flow chart of the distribution network fault location method based on intelligent decision-making provided by an embodiment of the present invention. The distribution network fault location method based on intelligent decision-making will be introduced in detail below.

[0009] Step S110: Obtain a set of target state data collected by the global intelligent perception network of the distribution network. The set of target state data includes equipment operation condition data, network topology structure data, historical fault knowledge base, and environmental interference data.

[0010] In this embodiment, the global intelligent perception network of the distribution network is a monitoring system widely distributed at various key positions in the distribution network. Its main function is to collect various types of data related to the operation of the distribution network to construct a target state data set. This target state data set covers equipment operation condition data, network topology structure data, historical fault knowledge base, and environmental interference data.

[0011] The acquisition of equipment operation condition data depends on sensors installed on various key equipment in the distribution network. For example, voltage sensors, current sensors, and temperature sensors are installed on transformers. The voltage sensor can monitor the voltage values of the input and output of the transformer in real time, the current sensor is used to measure the magnitude of the current passing through the transformer, and the temperature sensor can detect the temperature change during the operation of the transformer. The above sensors sample the operation parameters of the equipment at a set sampling frequency to obtain a series of time-series sampling values, such as time-series sampling values of voltage, current, and temperature. These sampling values constitute a part of the equipment operation condition data and reflect the operation state of the equipment at different times.

[0012] The acquisition of network topology structure data relies on a network topology monitoring system. This system monitors and records the physical and logical connections between various equipment in the distribution network to obtain the network topology structure information of the distribution network. Specifically, it records the connection relationships between each equipment node (such as transformers, circuit breakers, switches, etc.), as well as the connection methods and paths. For example, it can record which circuit breakers a certain transformer is connected to and how these connections are achieved through which cables or lines. Network topology structure data is crucial for analyzing the power transmission path in the distribution network and the collaborative working relationship between equipment.

[0013] The historical fault knowledge base is a database formed by detailed recording and sorting of all past fault events in the distribution network, which contains information such as the specific time of each fault occurrence, the location of the fault (i.e., the equipment node involved), the type of the fault (such as short-circuit fault, open-circuit fault, etc.), and the cause of the fault.

[0014] The acquisition of environmental interference data utilizes a variety of environmental monitoring devices, such as weather stations and electromagnetic monitoring devices. The weather station can monitor the meteorological conditions in the area where the distribution network is located in real time, including environmental factors such as temperature, humidity, wind speed, and air pressure. The electromagnetic monitoring device is used to detect the electromagnetic environment around the distribution network, such as the intensity and frequency of electromagnetic interference. The above environmental factors may affect the operation of the equipment in the distribution network. For example, a high-temperature environment may cause the equipment to overheat, increasing the risk of equipment failure; strong electromagnetic interference may interfere with the normal communication and control signals of the equipment.

[0015] Step S120: Perform spatio-temporal semantic fusion processing on the target state data set to generate a multi-dimensional feature map reflecting the dynamic operation state of the distribution network. The multi-dimensional feature map includes the state evolution trajectory in the time dimension and the device collaboration relationship in the space dimension.

[0016] In this embodiment, the purpose of performing spatio-temporal semantic fusion processing on the target state data set is to integrate data of different types and dimensions, mine the internal relationships between the data, and thus generate a multi-dimensional feature map that can comprehensively reflect the dynamic operation state of the distribution network. The multi-dimensional feature map includes the state evolution trajectory in the time dimension and the device collaboration relationship in the space dimension.

[0017] Step S121: Perform time window alignment processing on the device operation condition data in the target state data set to obtain a condition data sequence with a time synchronization relationship. The condition data sequence includes time series sampling values of operation parameters in multiple dimensions.

[0018] In this embodiment, since the sensor sampling frequencies of different devices may be different, there are time differences in the collected device operation condition data. In order to effectively analyze and compare these data, it is necessary to perform time window alignment processing on the device operation condition data.

[0019] Specifically, it is first necessary to determine a unified time window size and time step. The time window size determines the data range processed each time, and the time step determines the interval between adjacent time windows. For example, set the time window size to a specific time period, and the time step to a part of this time period. Then, based on this unified time window, perform alignment operations on the time series sampling values of the operation parameters of each device.

[0020] For the time series sampling values of the operation parameters of each device, arrange them in chronological order, and then divide them according to the time window and time step. If there are missing values in the sampling values of a certain device within a certain time window, interpolation methods can be used for supplementation. For example, linear interpolation methods can be used to estimate the missing values based on the sampling values of the device at adjacent time points. Through the above processing, a condition data sequence with a time synchronization relationship is obtained. The condition data sequence includes time series sampling values of operation parameters in multiple dimensions, such as time series sampling values of voltage, current, power, etc. The sampling values of each dimension are arranged on a unified time scale.

[0021] Step S122: Perform graph model conversion processing on the network topology structure data in the target state data set to generate a distribution network topology graph including device nodes and connection edges. The weight of the connection edge is determined by the power transmission capacity between devices.

[0022] In this embodiment, in order to more intuitively represent the network topology structure of the distribution network and facilitate subsequent analysis and processing, it is necessary to convert the network topology structure data into a graph model.

[0023] First, each device in the distribution network is abstracted as a node in the graph model. Devices such as transformers, circuit breakers, and switches can all be regarded as nodes in the graph. Then, according to the connection relationships between devices recorded in the network topology structure data, connection edges are added to the graph. For example, if there is a physical connection between two devices, a connection edge is added between the corresponding nodes.

[0024] The weight of the connection edge is determined according to the power transmission capacity between devices. The power transmission capacity reflects the maximum ability of electrical energy that can be transmitted between two devices. Specifically, for each connection edge, the greater the power transmission capacity between the corresponding devices, the higher the weight of the connection edge. For example, if the power transmission capacity between two devices is large, it indicates that their connection is crucial for the electrical energy transmission of the distribution network. Therefore, the weight of the corresponding connection edge in the graph model will be set relatively high; conversely, if the power transmission capacity between two devices is small, the weight of the connection edge will be low. Through the above method, a distribution network topology graph containing device nodes and connection edges is generated, which intuitively reflects the connection relationships between devices in the distribution network and the importance of electrical energy transmission.

[0025] Step S123: Perform spatial region division processing on the environmental interference data in the target state data set to generate an environmental parameter distribution matrix for the area where the device is located. The environmental parameter distribution matrix includes spatial sampling values of environmental parameters in multiple dimensions.

[0026] In this embodiment, since the environmental interference data is collected at different spatial positions where the distribution network is located, in order to better integrate it with the device operating condition data and network topology structure data, it is necessary to perform spatial region division processing on the environmental interference data.

[0027] First, according to the geographical scope of the distribution network and the distribution of devices, the area where the distribution network is located is divided into multiple sub-regions. The division principle can be determined according to factors such as geographical coordinates and the distribution density of devices. For example, the area where the distribution network is located can be divided into several small rectangular sub-regions according to a certain longitude and latitude grid.

[0028] Then, for each sub-region, the sampling values of environmental parameters in this region are collected. Environmental parameters include multiple dimensions such as temperature, humidity, wind speed, and electromagnetic interference. For each sub-region, the sampling values of these environmental parameters are sorted and statistically analyzed to obtain the environmental parameter characteristics of this sub-region. For example, the average value of temperature, the maximum value of humidity, and the standard deviation of wind speed in this sub-region can be calculated.

[0029] Finally, arrange the environmental parameter characteristics of all sub - regions in a set order to form an environmental parameter distribution matrix, where each row represents a sub - region and each column represents an environmental parameter dimension. For example, the first column of the matrix may represent temperature, the second column may represent humidity, and so on. In the above way, an environmental parameter distribution matrix of the area where the device is located is generated, and this environmental parameter distribution matrix reflects the environmental characteristics of different regions of the distribution network.

[0030] Step S124: Input the working condition data sequence, the distribution network topology diagram, and the environmental parameter distribution matrix into the semantic fusion module. Through semantic matching processing of the historical fault knowledge base, extract the association rules between the device operation state and the historical fault modes.

[0031] In this embodiment, the main function of the semantic fusion module is to fuse the processed working condition data sequence, the distribution network topology diagram, and the environmental parameter distribution matrix, and perform semantic matching processing using the historical fault knowledge base, so as to extract the association rules between the device operation state and the historical fault modes.

[0032] Step S1241: Perform feature extraction processing on the working condition data sequence to generate a working condition feature vector containing abnormal working condition events.

[0033] In this embodiment, performing feature extraction processing on the working condition data sequence is to extract key features that can reflect the device operation state from a large number of time - series sampling values of operation parameters.

[0034] Specifically, first analyze the time - series sampling values of the operation parameters in each dimension of the working condition data sequence. For example, for the time - series sampling values of voltage, calculate its statistical features such as mean, standard deviation, maximum value, minimum value, etc. These statistical features can reflect the change trend and fluctuation of voltage over a period of time.

[0035] Then, identify abnormal working condition events through set thresholds. For example, for the time - series sampling values of voltage, if the voltage value at a certain time point exceeds the pre - set normal range (i.e., the voltage threshold), it is considered that an abnormal working condition event occurs at this time point. Organize and encode the relevant information of these abnormal working condition events (such as the time of occurrence of the abnormality, the severity of the abnormality, etc.) to form a feature representation of the abnormal working condition event.

[0036] Finally, combine the statistical features of all dimensions and the feature representation of the abnormal working condition events together to form a working condition feature vector. This working condition feature vector contains information in multiple dimensions and can comprehensively reflect the operation state of the device.

[0037] Step S1242: Perform neighborhood feature aggregation processing on the distribution network topology map to generate a topology feature vector including node degree, edge weight, and clustering coefficient.

[0038] In this embodiment, performing neighborhood feature aggregation processing on the distribution network topology map is to extract the feature information of nodes and edges in the map and generate a topology feature vector that can reflect the topology structure characteristics of the distribution network.

[0039] First, calculate the node degree of each node in the distribution network topology map. The node degree refers to the number of edges connected to the node, which reflects the connection tightness of the node in the distribution network topology map. For example, a node with a higher node degree indicates that the node has more connections with other nodes and may be in a relatively important position in the distribution network.

[0040] Then, for each connection edge in the map, record its edge weight. The edge weight is determined according to the power transmission capacity between devices and reflects the importance of the edge in the distribution network.

[0041] Next, calculate the clustering coefficient of each node. The clustering coefficient is an index to measure the connection tightness between neighboring nodes around the node. Specifically, for a node, calculate the ratio of the actual number of connection edges between its neighboring nodes to the theoretically possible number of connection edges to obtain the clustering coefficient of the node. The higher the clustering coefficient, the closer the connections between the neighboring nodes around the node.

[0042] Finally, arrange the node degree, edge weight, and clustering coefficient of each node in a set order to form a topology feature vector. This topology feature vector contains the important feature information of nodes and edges in the distribution network topology map.

[0043] Step S1243: Perform statistical analysis processing on the environmental parameter distribution matrix to generate an environmental feature vector including humidity mean, temperature variance, and electromagnetic interference peak value.

[0044] In this embodiment, performing statistical analysis processing on the environmental parameter distribution matrix is to extract the key information that can reflect the environmental characteristics from the environmental parameter data and generate an environmental feature vector.

[0045] First, for the humidity data in the environmental parameter distribution matrix, calculate its mean value. The humidity mean value reflects the overall humidity level of the area where the distribution network is located. For example, if the humidity mean value is high, it indicates that the air in this area is relatively humid, which may affect the insulation performance of the equipment.

[0046] Then, for the temperature data, calculate its variance. The temperature variance reflects the fluctuation of temperature among different sub-regions. A larger variance indicates a greater difference in temperature among different regions, which may lead to significant differences in the operating environments of the device in different regions.

[0047] Next, for the electromagnetic interference data, find its peak value. The peak value of electromagnetic interference reflects the maximum interference intensity of the electromagnetic environment around the distribution network. If the peak value of electromagnetic interference is high, it may interfere with the communication and control signals of the device, affecting the normal operation of the device.

[0048] Finally, combine the humidity mean, temperature variance, and peak value of electromagnetic interference to form an environmental feature vector, which contains the key feature information of the environmental parameters.

[0049] Step S1244: Input the operating condition feature vector, topological feature vector, and environmental feature vector into the semantic matching unit of the historical fault knowledge base, and calculate the cosine similarity between each feature vector and the historical fault mode feature vector.

[0050] In this embodiment, the main function of the semantic matching unit of the historical fault knowledge base is to compare the operating condition feature vector, topological feature vector, and environmental feature vector with the historical fault mode feature vector, and calculate the cosine similarity between them.

[0051] First, for each historical fault mode in the historical fault knowledge base, there is a corresponding historical fault mode feature vector. This historical fault mode feature vector is extracted based on information such as the device operating state, network topology structure, and environmental conditions when the fault mode occurred, and has the same dimension as the operating condition feature vector, topological feature vector, and environmental feature vector.

[0052] Then, for the operating condition feature vector, topological feature vector, and environmental feature vector, calculate their cosine similarities with each historical fault mode feature vector respectively. Cosine similarity is an index to measure the similarity between two vectors, which determines their similarity by calculating the cosine value of the angle between the two vectors. The closer the value of cosine similarity is to 1, the more similar the two vectors are; the closer the value of cosine similarity is to 0, the less similar the two vectors are.

[0053] Specifically, when calculating, for two vectors A and B, the formula for cosine similarity is: Cosine similarity = (A·B) / (||A|| * ||B||), where A·B represents the dot product of vector A and vector B, and ||A|| and ||B|| represent the norms of vector A and vector B respectively. The calculated cosine similarity value can measure the similarity degree between the current device operating state, network topology structure, and environmental conditions and the historical fault mode.

[0054] Step S1245: Screen out historical fault modes with a similarity exceeding a preset threshold according to the cosine similarity value, and extract the corresponding device state evolution rules and fault triggering conditions as association rules.

[0055] In this embodiment, according to the calculated cosine similarity value, historical fault modes with a similarity exceeding the preset threshold to the current working condition feature vector, topology feature vector, and environmental feature vector are screened out. The preset threshold is a similarity standard set according to the actual situation and experience. Only when the cosine similarity value exceeds this preset threshold is it considered that the current situation has a high similarity to this historical fault mode.

[0056] For the screened historical fault modes, extract their corresponding device state evolution rules and fault triggering conditions. The device state evolution rules describe how the operating state of the device gradually changes before the fault occurs. For example, how operating parameters such as the voltage, current, and temperature of the device change over time. The fault triggering conditions refer to the specific reasons and conditions that cause the fault, such as specific environmental conditions, abnormal operations of the device, etc.

[0057] These device state evolution rules and fault triggering conditions constitute the association rules. The association rules can help predict possible faults in the current distribution network and take corresponding preventive measures. For example, if the current device operating state and environmental conditions are similar to a certain historical fault mode, and the fault triggering conditions of this historical fault mode are also partially met in the current situation, then it can be predicted that a similar fault may occur, and thus equipment inspection and maintenance can be carried out in advance.

[0058] Step S125: Based on the association rules, perform feature cross - fusion processing on the working condition data sequence, distribution network topology map, and environmental parameter distribution matrix to generate a multi - dimensional feature map containing time evolution trajectories and spatial coordination relationships.

[0059] In this embodiment, based on the extracted association rules, feature cross - fusion processing is performed on the working condition data sequence, distribution network topology map, and environmental parameter distribution matrix. The purpose is to generate a multi - dimensional feature map that can comprehensively reflect the dynamic operation state of the distribution network. This multi - dimensional feature map contains the state evolution trajectory in the time dimension and the device coordination relationship in the spatial dimension.

[0060] Step S1251: Perform a sliding window analysis process on the time dimension of the working condition data sequence, count the occurrence frequency of abnormal working condition events within the sliding window, and generate the time evolution trajectory of the device state.

[0061] In this embodiment, performing a sliding window analysis process on the time dimension of the working condition data sequence is to observe the changes in the device state over time and generate the time evolution trajectory of the device state.

[0062] First, define the size and step of a sliding window. The size of the sliding window represents the time range covered by the window, and the step represents the time interval for each movement of the window. For example, the size of the sliding window can be set to a specific time period, and the step can be set to a part of that time period.

[0063] Then, slide the sliding window over the operating condition data sequence, moving one step each time. For the operating condition data within each sliding window, count the occurrence frequency of abnormal operating condition events. The occurrence frequency of abnormal operating condition events reflects the instability degree of the device during that time period. For example, if the occurrence frequency of abnormal operating condition events is high within a certain sliding window, it indicates that the operating state of the device is relatively unstable during that time period, and there may be potential fault risks.

[0064] Finally, arrange the occurrence frequencies of abnormal operating condition events corresponding to each sliding window in chronological order to form the time evolution trajectory of the device state. This trajectory can visually display the changing trend of the device state over time.

[0065] Step S1252: Perform path search processing on the spatial dimension of the distribution network topology map. Starting from the significant load equipment node, traverse all reachable nodes and record the edge weights on the path to generate the spatial cooperation relationship between devices.

[0066] In this embodiment, performing path search processing on the spatial dimension of the distribution network topology map is to analyze the cooperative working relationship between devices in space and generate the spatial cooperation relationship between devices.

[0067] First, determine the significant load equipment node. The significant load equipment node refers to the equipment node that undertakes a large load in the distribution network, such as a large transformer, etc. The above nodes have an important position in the distribution network and may have a greater impact on the operating states of other devices.

[0068] Then, starting from the significant load equipment node, use a path search algorithm (such as depth - first search algorithm or breadth - first search algorithm) to traverse the distribution network topology map. During the traversal process, record the edge weights on the path from the starting point to each reachable node. The edge weight reflects the power transmission capacity and connection importance between devices.

[0069] Through the traversal of all reachable nodes and the recording of edge weights, the spatial cooperation relationship between devices is obtained. This relationship describes the connection and interaction situation between devices in the distribution network in space, such as which devices have a direct power transmission relationship and which device connections are more important, etc.

[0070] Step S1253: Perform spatio-temporal joint analysis and processing on the environmental parameter distribution matrix, count the co-occurrence frequency of environmental parameters and equipment failures in the target area, and generate the coupling characteristics of environmental interference and equipment status.

[0071] In this embodiment, performing spatio-temporal joint analysis and processing on the environmental parameter distribution matrix is to deeply explore the internal relationship between environmental factors and equipment status, and generate the coupling characteristics of environmental interference and equipment status.

[0072] First, clarify the target area. The target area can be divided according to the actual situation of the distribution network, for example, divided according to geographical areas, equipment distribution areas, etc. After determining the target area, analyze the environmental parameter distribution matrix within this area.

[0073] For each environmental parameter dimension in the environmental parameter distribution matrix (such as temperature, humidity, wind speed, electromagnetic interference, etc.), combined with historical failure records, count the co-occurrence frequency of this environmental parameter and equipment failures. The calculation process of the co-occurrence frequency is as follows: at each time point and spatial position, determine whether the environmental parameter is within a certain set range, and at the same time determine whether an equipment failure has occurred. If at the same time and spatial position, the environmental parameter is within the set range and an equipment failure has occurred, it is considered that this environmental parameter and the equipment failure co-occur. Count the number of co-occurrence times at all time points and spatial positions, and divide by the total number of time points and spatial positions to obtain the co-occurrence frequency of this environmental parameter and equipment failures.

[0074] For example, for the temperature parameter, set a temperature range (such as above a certain temperature threshold), count the number of equipment failures within this temperature range, and then divide by the total number of statistical times to obtain the co-occurrence frequency of temperature and equipment failures. Through the above statistical analysis of each environmental parameter dimension, a series of co-occurrence frequency values are obtained.

[0075] These co-occurrence frequency values reflect the degree of correlation between environmental parameters and equipment failures. The higher the co-occurrence frequency, the greater the impact of this environmental parameter on equipment failures. Arrange these co-occurrence frequency values according to the environmental parameter dimension to form the coupling characteristics of environmental interference and equipment status. This coupling characteristic can reflect the interaction relationship between environmental factors and equipment status.

[0076] Step S1254: Input the time evolution trajectory, spatial coordination relationship, and coupling characteristics into the feature map construction unit, and generate a multi-dimensional feature map with unified time resolution and spatial resolution through feature dimension alignment processing.

[0077] In this embodiment, the role of the feature map construction unit is to integrate the time evolution trajectory, spatial coordination relationship, and coupling features to generate a multi-dimensional feature map. Before integration, it is necessary to perform feature dimension alignment processing to ensure that these features have a unified time resolution and spatial resolution.

[0078] Regarding the time evolution trajectory, it records the changes in the device state over time and has a certain time resolution, that is, the time interval represented by each data point. Regarding the spatial coordination relationship, it describes the connection and interaction between devices in space and has a certain spatial resolution, that is, the spatial range represented by each node or region. The coupling feature is the correlation feature between environmental parameters and device failures and also needs to be aligned with the time evolution trajectory and spatial coordination relationship in terms of time and space.

[0079] The specific process of feature dimension alignment processing is as follows: First, determine the unified time resolution and spatial resolution. The time resolution can be selected according to actual needs and the characteristics of the data. For example, a fixed time interval can be selected as the unified time resolution. The spatial resolution can be determined according to the geographical scope of the distribution network and the device distribution. For example, the distribution network can be divided into several regions of the same size as the unified spatial resolution.

[0080] Then, for the time evolution trajectory, resample or interpolate it according to the unified time resolution so that the time interval of each data point is consistent with the unified time resolution. For the spatial coordination relationship, divide or merge it according to the unified spatial resolution so that the spatial range of each node or region is consistent with the unified spatial resolution. For the coupling feature, also adjust it according to the unified time resolution and spatial resolution to ensure its alignment with the time evolution trajectory and spatial coordination relationship in terms of time and space.

[0081] Finally, input the time evolution trajectory, spatial coordination relationship, and coupling features that have undergone feature dimension alignment processing into the feature map construction unit. This unit splices and integrates these features to generate a multi-dimensional feature map with a unified time resolution and spatial resolution. This multi-dimensional feature map contains the state evolution trajectory in the time dimension, the device coordination relationship in the spatial dimension, and the coupling feature between environmental interference and device state, and can comprehensively reflect the dynamic operation situation of the distribution network.

[0082] Step S1255: Perform noise filtering processing on the multi-dimensional feature map, eliminate redundant features irrelevant to historical fault patterns, and retain the core features strongly related to fault risks as the final multi-dimensional feature map.

[0083] In this embodiment, since the multi-dimensional feature map may contain some redundant features that are irrelevant to historical fault modes, these features will increase the complexity of subsequent analysis and may affect the accuracy of fault risk discrimination. Therefore, it is necessary to perform noise filtering on the multi-dimensional feature map.

[0084] The specific process of noise filtering is as follows: First, according to the historical fault knowledge base and the previously extracted association rules, the core features that are strongly correlated with the fault risk are determined. The above-mentioned core features are the features that often appear in historical fault modes and have an important impact on fault risk discrimination. For example, certain specific device operating parameters, environmental parameters, or connection relationships between devices, etc.

[0085] Then, each feature in the multi-dimensional feature map is evaluated to determine whether it is related to the core features. The evaluation can be performed by calculating the correlation coefficient between the features. If the correlation coefficient of a certain feature with the core features is low, it indicates that the feature has a weak correlation with the fault risk and may be a redundant feature.

[0086] For the features determined to be redundant features, they are removed from the multi-dimensional feature map. Through the above processing, the core features that are strongly correlated with the fault risk are retained, and the final multi-dimensional feature map is obtained. This final multi-dimensional feature map is more concise and effective and can better reflect the fault risk situation of the distribution network.

[0087] Step S130: Perform context-aware discrimination processing on the multi-dimensional feature map through a pre-constructed dynamic decision inference engine to generate a fault risk discrimination result of the distribution network. The fault risk discrimination result includes a risk level identifier, a risk trigger period, and a risk impact range.

[0088] In this embodiment, the pre-constructed dynamic decision inference engine is used to deeply analyze and process the multi-dimensional feature map to generate a fault risk discrimination result of the distribution network. This engine can extract useful context information from the multi-dimensional feature map and perform fault risk discrimination based on this information.

[0089] Step S131: Input the multi-dimensional feature map into the context-aware layer of the dynamic decision inference engine, and extract the state mutation points in the time evolution trajectory and the abnormal connection edges in the spatial cooperation relationship as context features.

[0090] In this embodiment, the main function of the context-aware layer of the dynamic decision inference engine is to extract context features from the multi-dimensional feature map. The above context features can reflect the abnormal situation of the distribution network operation state.

[0091] For the time evolution trajectory, a state mutation point refers to a point where the device state suddenly changes over time. For example, if the operating parameters of a device such as voltage, current, and power suddenly change significantly at a certain time point, it may indicate potential fault risks in the device. State mutation points can be identified by performing differential calculations on the time evolution trajectory or by setting thresholds. For example, calculate the difference in operating parameters between adjacent time points. If the difference exceeds a certain threshold, then that time point is considered a state mutation point.

[0092] For the spatial coordination relationship, an abnormal connection edge refers to an edge whose weight or state has an abnormal change compared to the normal situation. For example, if the power transmission capacity of certain connection edges suddenly decreases, or the state of the connection edge changes from normal to abnormal, these may all indicate problems in the collaborative working relationship between devices. Abnormal connection edges can be identified by comparing the current spatial coordination relationship with the historical normal spatial coordination relationship, or by setting a threshold for edge weight changes.

[0093] Take the identified state mutation points and abnormal connection edges as context features, which can reflect the abnormal conditions of the distribution network in the time and space dimensions.

[0094] Step S132: Input the context features into the risk assessment layer of the dynamic decision inference engine. Through fuzzy logic inference processing, calculate the risk probability value of the device node, where the risk probability value is jointly determined by the severity of the state mutation point and the influence range of the abnormal connection edge.

[0095] In this embodiment, the risk assessment layer of the dynamic decision inference engine uses fuzzy logic inference to process the context features and calculate the risk probability value of the device node. Fuzzy logic inference is a reasoning method that can handle uncertain and fuzzy information and is suitable for problems with a certain degree of uncertainty such as fault risk assessment.

[0096] Step S1321: Perform severity assessment processing on the state mutation points in the context features. According to the numerical value of the abnormal operating condition event, divide the state mutation points into multiple state mutation levels.

[0097] In this embodiment, performing severity assessment processing on the state mutation points is to more accurately measure the impact of the state mutation on device operation. For each state mutation point, it is evaluated according to the numerical value of the abnormal operating condition event. The numerical value of the abnormal operating condition event can be the change range of device operating parameters, such as the change range of voltage, the change range of current, etc.

[0098] Set a series of thresholds. According to the comparison results between the numerical values of abnormal operating condition events and these thresholds, the state mutation points are divided into multiple state mutation levels. For example, set three thresholds. When the numerical value of the abnormal operating condition event is less than the first threshold, the state mutation point is divided into a low-level state mutation; when the numerical value of the abnormal operating condition event is between the first threshold and the second threshold, the state mutation point is divided into a medium-level state mutation; when the numerical value of the abnormal operating condition event is greater than the second threshold, the state mutation point is divided into a high-level state mutation. Different state mutation levels reflect different severities of state mutations. A high-level state mutation indicates that the state mutation has a greater impact on the operation of the equipment and may lead to a higher risk of equipment failure.

[0099] Step S1322: Perform an impact range evaluation process on the abnormal connection edges in the context features. According to the degree of reduction of the edge weights, the abnormal connection edges are divided into multiple abnormal connection edge levels, and the abnormal connection edge levels include local impact level, regional impact level, and global impact level.

[0100] In this embodiment, performing an impact range evaluation process on the abnormal connection edges is to determine the impact degree and range of the abnormal connection edges on the distribution network. For each abnormal connection edge, an evaluation is performed according to the degree of reduction of the edge weights. The degree of reduction of the edge weights reflects the decrease in the power transmission capacity between devices.

[0101] Set different edge weight reduction thresholds. According to the comparison results between the degree of reduction of the edge weights of the abnormal connection edges and these thresholds, the abnormal connection edges are divided into multiple abnormal connection edge levels. For example, when the degree of reduction of the edge weights is less than a certain lower threshold, the abnormal connection edge is divided into the local impact level, indicating that this abnormal connection edge only affects the operation of local devices; when the degree of reduction of the edge weights is between the lower threshold and the higher threshold, the abnormal connection edge is divided into the regional impact level, indicating that this abnormal connection edge will affect the operation of devices within a region; when the degree of reduction of the edge weights is greater than the higher threshold, the abnormal connection edge is divided into the global impact level, indicating that this abnormal connection edge will have a greater impact on the operation of the entire distribution network.

[0102] Step S1323: Construct a fuzzy inference rule base to define the influence relationship of the combination of the state mutation level and the abnormal connection edge level on the risk probability value.

[0103] In this embodiment, constructing a fuzzy inference rule base is to clarify the influence relationship of different combinations of the state mutation level and the abnormal connection edge level on the risk probability value of the equipment nodes. The fuzzy inference rule base is constructed based on expert knowledge and historical fault data.

[0104] Define a series of fuzzy rules according to different combinations of the state mutation level and the abnormal connection edge level. For example, if the state mutation level is high and the abnormal connection edge level is the global impact level, the risk probability value of the device node is high; if the state mutation level is low and the abnormal connection edge level is the local impact level, the risk probability value of the device node is low. Determine these fuzzy rules by analyzing and summarizing a large amount of historical fault data and combining the experience and judgment of experts.

[0105] Step S1324: Input the state mutation level and the abnormal connection edge level into the fuzzy inference rule base, and calculate the risk probability value of the device node through fuzzyfication, rule matching, and defuzzification processing.

[0106] In this embodiment, after inputting the state mutation level and the abnormal connection edge level into the fuzzy inference rule base, it is necessary to perform fuzzyfication, rule matching, and defuzzification processing to calculate the risk probability value of the device node.

[0107] First, perform fuzzyfication processing. Since the state mutation level and the abnormal connection edge level are discrete level values, they need to be converted into membership degree values in the fuzzy set. For example, for the state mutation level of high, its membership degree value in the "high-risk state mutation" fuzzy set can be converted to 1, and the membership degree values in other fuzzy sets are 0.

[0108] Then, perform rule matching processing. According to the rules in the fuzzy inference rule base, match the fuzzyfied state mutation level and abnormal connection edge level to find applicable rules. For each applicable rule, calculate its rule firing strength, which is calculated based on the membership degree values in the rule premise conditions.

[0109] Finally, perform defuzzification processing. Combine the firing strengths and rule conclusions of all applicable rules, and calculate the risk probability value of the device node through a defuzzification method (such as the centroid method). This risk probability value reflects the likelihood of the device node failing, which is jointly determined by the severity of the state mutation point and the influence range of the abnormal connection edge.

[0110] Step S133: Input the risk probability value into the level division layer of the dynamic decision-making inference engine, and divide the device node into multiple risk levels according to the preset risk level thresholds to generate risk level identifiers.

[0111] In this embodiment, the level division layer of the dynamic decision-making inference engine divides the risk probability value of the device node according to the preset risk level thresholds to determine the risk level of the device node.

[0112] The preset risk level thresholds are a series of values set based on historical failure data and actual experience. For example, three risk level thresholds are set, namely the low - risk threshold, the medium - risk threshold, and the high - risk threshold. When the risk probability value of a device node is less than the low - risk threshold, the device node is classified as a low - risk level; when the risk probability value of the device node is between the low - risk threshold and the medium - risk threshold, the device node is classified as a medium - risk level; when the risk probability value of the device node is greater than the medium - risk threshold, the device node is classified as a high - risk level.

[0113] The risk level of each device node is represented by an identifier, such as a letter or a number, to form a risk - level identifier. This risk - level identifier can intuitively reflect the failure risk degree of the device node.

[0114] Step S134: Perform timestamp analysis processing on the state mutation points in the time evolution trajectory to determine the start time and end time of the risk event, and generate a risk trigger period.

[0115] In this embodiment, by performing timestamp analysis processing on the state mutation points in the time evolution trajectory, the start time and end time of the risk event are determined, thereby generating a risk trigger period.

[0116] For each state mutation point, its corresponding timestamp records the time when the state mutation occurs. Analyze the timestamps of adjacent state mutation points to find the start state mutation point and the end state mutation point of the risk event. The start state mutation point is the time point when the risk event starts to show abnormalities, and the end state mutation point is the time point when the abnormal state of the risk event ends.

[0117] Take the timestamp of the start state mutation point as the start time of the risk event, and the timestamp of the end state mutation point as the end time of the risk event. These two time points constitute the risk trigger period. The risk trigger period can clarify the time range during which the risk event occurs.

[0118] Step S135: Perform regional coverage analysis processing on the abnormal connection edges in the spatial collaboration relationship to determine the range of device nodes affected by the risk event, and generate a risk impact range.

[0119] In this embodiment, performing regional coverage analysis processing on the abnormal connection edges in the spatial collaboration relationship is to determine the range of device nodes affected by the risk event and generate a risk impact range.

[0120] First, determine the area covered by the abnormal connection edge according to its position and connection relationship. For each abnormal connection edge, find the device nodes connected to it. Then, taking these device nodes as the center, determine the range of affected device nodes according to the set rules (such as distance threshold or connection relationship).

[0121] For example, a distance threshold can be set. For the device nodes connected to the abnormal connection edges, all other device nodes within the distance threshold from this node are found, and these device nodes constitute the range of device nodes affected by the risk event.

[0122] Sort out and identify these ranges of affected device nodes to generate a risk impact range. This risk impact range can intuitively show the degree and scope of the impact of the risk event in space.

[0123] Step S136: Combine the risk level identifier, the risk trigger period, and the risk impact range to form a failure risk discrimination result including the risk level identifier, the risk trigger period, and the risk impact range.

[0124] In this embodiment, the generated risk level identifier, the risk trigger period, and the risk impact range are combined to form a complete failure risk discrimination result. This result can comprehensively reflect the failure risk situation of the device nodes in the distribution network, including the level of the failure risk, the time range of occurrence, and the spatial range of the impact.

[0125] Step S140: Perform causal link tracing processing on the target state data set based on the failure risk discrimination result to generate a fault section location result of the distribution network. The fault section location result includes a risk propagation path, key impact nodes, and a quantification value of the impact degree.

[0126] In this embodiment, the purpose of performing causal link tracing processing on the target state data set based on the failure risk discrimination result is to find out the cause and propagation path of the fault, determine the key impact nodes and the quantification value of the impact degree, and generate a fault section location result.

[0127] Step S141: If the risk level corresponding to the risk level identifier in the failure risk discrimination result is higher than the set risk level, extract the subset of state data in the target state data set that overlaps with the risk trigger period.

[0128] In this embodiment, first, it is judged whether the risk level corresponding to the risk level identifier in the failure risk discrimination result is higher than the set risk level. The set risk level is a threshold set according to actual requirements and the safety standards of the distribution network. When the risk level is higher than this threshold, it indicates that the device has a relatively high failure risk and further analysis and processing are required.

[0129] If the risk level is higher than the set risk level, a subset of status data that overlaps with the risk trigger period is extracted from the target status data set. The target status data set includes equipment operating condition data, network topology data, historical fault knowledge base, environmental interference data, etc. According to the risk trigger period, the equipment operating condition data, network topology data, and environmental interference data recorded during this period are filtered out to form a subset of status data. This subset of status data can reflect the actual operating status of the distribution network during the risk event.

[0130] Step S142: Perform causal relationship mining processing on the subset of status data to construct a fault causal graph including equipment nodes and causal relationship edges, and the weight of the causal relationship edge is determined by the mutual information value of the operating parameters between the equipment nodes.

[0131] In this embodiment, performing causal relationship mining processing on the subset of status data is to find the causal relationship between equipment nodes and construct a fault causal graph.

[0132] Step S1421: Perform Granger causality test processing on the equipment operating parameters in the subset of status data to determine the causal relationship direction between equipment nodes, and the causal relationship direction indicates whether the state change of one node will cause the state change of another node.

[0133] In this embodiment, the Granger causality test is a key step in determining the causal relationship direction between equipment nodes. In the subset of status data, each equipment node has multiple dimensions of operating parameters, such as voltage, current, power, etc., and the above parameters form time series data as they change over time.

[0134] First, for the time series of the operating parameters of any two equipment nodes, denoted as parameter sequence A and parameter sequence B. The core idea of the Granger causality test is based on the predictability of the time series. If the past values of parameter sequence A can help predict the future values of parameter sequence B, while the past values of parameter sequence B cannot help predict the future values of parameter sequence A, then it can be considered that parameter sequence A is the Granger cause of parameter sequence B, that is, it reflects the possible causal relationship direction between equipment nodes from the parameter level.

[0135] Specifically, two regression models are constructed. The first regression model is to predict the current value of parameter sequence B only using the past values of parameter sequence B, denoted as model M1. The second regression model is to predict the current value of parameter sequence B using both the past values of parameter sequence A and parameter sequence B, denoted as model M2.

[0136] For model M1, methods such as the least squares method are used to estimate the parameters of the model, so that the fitting error of the model to the parameter sequence B is minimized. For model M2, the least squares method is also used to estimate its parameters. Then, calculate the residual sum of squares of the two models, denoted as RSS1 and RSS2 respectively.

[0137] Next, use the F-test to determine whether model M2 is significantly better than model M1. The calculation process of the F-test statistic is to divide (RSS1 - RSS2) by the combined value of RSS2 and the degrees of freedom. Here, the degrees of freedom are related to the number of model parameters and the number of samples. Compare the calculated F-test statistic with the critical value of the F-distribution at a given significance level.

[0138] If the F-test statistic is greater than the critical value, it means that model M2 is significantly better than model M1, that is, the past values of parameter sequence A have a significant contribution to predicting the current value of parameter sequence B. At this time, parameter sequence A can be considered the Granger cause of parameter sequence B, indicating that at the device node level, the state change of the device node corresponding to parameter sequence A may cause the state change of the device node corresponding to parameter sequence B, thus determining a causal relationship direction between these two device nodes.

[0139] Perform such Granger causality tests pairwise on the operating parameters of all device nodes in the state data subset to obtain the complete causal relationship direction information between device nodes.

[0140] Step S1422: Calculate the mutual information value of the operating parameters of the device node pairs with causal relationships, and the mutual information value represents the correlation strength of the state changes between device nodes.

[0141] After determining the causal relationship direction between device nodes, it is necessary to calculate the mutual information value of the operating parameters of the device node pairs with causal relationships to measure the correlation strength of the state changes between device nodes.

[0142] For the time series of the operating parameters of two device nodes with causal relationships, denoted as parameter sequence X and parameter sequence Y. Mutual information is an index to measure the correlation between two random variables. Based on the concept of information theory, it reflects the degree to which the information of one random variable can reduce the uncertainty of another random variable.

[0143] First, discretize parameter sequence X and parameter sequence Y. Since continuous operating parameters are not conducive to directly calculating mutual information, the value range of the parameter sequence is divided into several intervals, and each interval corresponds to a discrete state. For example, for voltage parameters, its value range can be divided into low voltage interval, normal voltage interval, high voltage interval, etc.

[0144] Then, the joint probability distribution and marginal probability distribution of the parameter sequence X and the parameter sequence Y in each discrete state are statistically calculated. The joint probability distribution represents the probability that the parameter sequences X and Y are simultaneously in two certain discrete states, and the marginal probability distribution represents the probability that the parameter sequences X and Y are in a certain discrete state respectively.

[0145] Based on the joint probability distribution and marginal probability distribution, the mutual information value is calculated. The calculation process of the mutual information value is as follows: for each state combination in the joint probability distribution, multiply the joint probability of this combination by the ratio of the joint probability of this combination to the product of the marginal probabilities, and then sum over all state combinations. The obtained result is the mutual information value of the parameter sequences X and Y.

[0146] The larger the mutual information value is, the stronger the correlation between the operating parameters of the two device nodes is, that is, the closer the association between the state changes of one device node and the state changes of the other device node is.

[0147] Step S1423: Generate a directed causal relationship edge between device nodes according to the causal relationship direction and the mutual information value. The direction of the edge is consistent with the causal relationship direction, and the weight of the edge is the mutual information value.

[0148] Generate a directed causal relationship edge between device nodes according to the causal relationship direction and the mutual information value obtained previously.

[0149] For each pair of device nodes with a causal relationship, determine the direction of the directed causal relationship edge according to the causal relationship direction. For example, if it is determined through Granger causality test that the state change of device node A will cause the state change of device node B, then the direction of the directed causal relationship edge is from device node A to device node B.

[0150] At the same time, use the mutual information value of the operating parameters of these two device nodes as the weight of the directed causal relationship edge. The larger the mutual information value is, the stronger the causal association between these two device nodes is, which is reflected as a larger weight of the edge in the directed causal relationship edge.

[0151] In the above way, a directed causal relationship edge is generated for each pair of device nodes with a causal relationship, and these edges reflect the causal relationship between device nodes and the strength of the relationship.

[0152] Step S1424: Perform graph structure assembly processing on all device nodes and directed causal relationship edges to generate a fault causal graph reflecting the causal relationship between device nodes.

[0153] After generating all the directed causal relationship edges, perform graph structure assembly processing on all device nodes and directed causal relationship edges to generate a fault causal graph reflecting the causal relationship between device nodes.

[0154] First, each device node is regarded as a vertex in the graph, and the directed causal relationship edges are regarded as the directed edges in the graph. Then, according to the actual connection and causal relationship of the device nodes, the vertices and directed edges are connected and laid out.

[0155] During the assembly process, it is necessary to ensure that the structure of the graph accurately reflects the causal relationship between device nodes. For example, if there are multiple device nodes having a causal relationship with a central device node, then these connection relationships should be correctly reflected in the graph. At the same time, it is necessary to ensure that the direction and weight of the directed edge are consistent with the direction and mutual information value of the causal relationship calculated previously.

[0156] Through the above graph structure assembly process, a complete fault causal graph is obtained, which intuitively shows the causal relationship and the strength of the relationship between device nodes in the distribution network.

[0157] Step S1425: Perform sparsification processing on the fault causal graph, and eliminate the weak causal relationship edges with mutual information values lower than the preset threshold.

[0158] In order to make the fault causal graph more concise and effective, it is necessary to perform sparsification processing on it and eliminate the weak causal relationship edges with mutual information values lower than the preset threshold.

[0159] The preset threshold is a standard set according to the actual situation and experience, and is used to distinguish strong causal relationships from weak causal relationships. For each directed causal relationship edge in the fault causal graph, check the weight of its edge (i.e., the mutual information value).

[0160] If the mutual information value of a certain directed causal relationship edge is lower than the preset threshold, it means that the causal relationship between the device nodes represented by this edge is weak and has little impact on the fault analysis. Such weak causal relationship edges are eliminated from the fault causal graph.

[0161] Through the sparsification processing, the strong causal relationship edges with higher mutual information values are retained, making the fault causal graph clearer and highlighting the main causal relationships between device nodes.

[0162] Step S143: Starting from the significant risk device nodes in the fault causal graph, traverse all reachable nodes through the depth-first search algorithm to generate a candidate set of risk propagation paths.

[0163] After obtaining the sparsified fault causal graph, starting from the significant risk device nodes, use the depth-first search algorithm to traverse all reachable nodes to generate a candidate set of risk propagation paths.

[0164] The significant risk device nodes are determined according to the previous fault risk discrimination results. These nodes have a higher risk level and may be the source of the fault or the nodes greatly affected by the fault.

[0165] The depth - first search algorithm is an algorithm used to traverse a graph. Starting from the significant risk device node, first visit this node, then select an unvisited directed edge connected to it, and visit the next node along this edge. Continue to repeat this process for the next node until it is impossible to move forward or all reachable nodes have been visited.

[0166] During the traversal process, record the paths from the starting point to each reachable node. Each recorded path is a candidate risk propagation path.

[0167] By performing such depth - first search traversals on all significant risk device nodes, a set of candidate risk propagation paths is obtained. This set of candidate risk propagation paths contains the possible risk propagation paths starting from the significant risk device nodes.

[0168] Step S144: Perform a weight screening process on the set of candidate risk propagation paths, and retain the risk propagation paths whose mutual information values exceed the preset threshold as valid risk propagation paths.

[0169] After obtaining the set of candidate risk propagation paths, it is necessary to perform a weight screening process on it to determine the valid risk propagation paths.

[0170] For each path in the set of candidate risk propagation paths, calculate the mutual information values of all directed causal relationship edges on the path. Methods such as summation and averaging can be used to synthesize the mutual information values on the path.

[0171] Then, compare the calculated synthesized mutual information value with the preset threshold. The preset threshold is a criterion for distinguishing valid paths from invalid paths. If the synthesized mutual information value of a certain path exceeds the preset threshold, it indicates that the risk propagation relationship represented by this path is relatively strong and may be a real risk propagation path, and it is retained as a valid risk propagation path.

[0172] Through the above - mentioned weight screening process, valid risk propagation paths are screened out from the set of candidate risk propagation paths, and these valid risk propagation paths are more likely to be actual risk propagation paths.

[0173] Step S145: Identify the node with the highest degree of abnormality of the operating parameters on the valid risk propagation path as the key influencing node, and the degree of abnormality is determined by the amplitude of the deviation of current, voltage, and temperature from the reference value.

[0174] After obtaining the valid risk propagation paths, it is necessary to identify the node with the highest degree of abnormality of the operating parameters on these paths as the key influencing node.

[0175] For each node on the effective risk propagation path, monitor its operating parameters such as current, voltage, temperature, etc. Each operating parameter has a reference value, which can be the rated parameter of the device or the average parameter during historical normal operation.

[0176] Calculate the deviation amplitude of the operating parameters such as current, voltage, temperature, etc. of each node from the reference value. For example, for the current parameter, subtract the reference current value from the current current value and then take the absolute value to obtain the deviation amplitude of the current. The same processing is performed on the voltage and temperature parameters.

[0177] Then, comprehensively evaluate the deviation amplitudes of these operating parameters. The weighted summation method can be used to assign different weights to different operating parameters because different operating parameters may have different degrees of influence on equipment failures. For example, the current parameter may have a greater impact on the overload failure of the equipment and is assigned a higher weight; the temperature parameter may have a greater impact on the heat dissipation and insulation performance of the equipment and is also assigned a set weight.

[0178] Through comprehensive evaluation, obtain the abnormal degree value of the operating parameters of each node. On the effective risk propagation path, find the node with the highest abnormal degree value and use it as the key influencing node. This key influencing node may be the key source of the failure or the node with the greatest impact on the failure propagation.

[0179] Step S146: Calculate the quantification value of the influence degree of the key influencing node on the risk propagation path, and form a fault section location result including the risk propagation path, the key influencing node, and the influence degree quantification value. The influence degree quantification value is jointly determined by the position of the key influencing node in the risk propagation path and the abnormal degree of the operating parameters.

[0180] After determining the key influencing node, it is necessary to calculate the quantification value of the influence degree of the key influencing node on the risk propagation path to form a complete fault section location result.

[0181] The influence degree quantification value is jointly determined by the position of the key influencing node in the risk propagation path and the abnormal degree of the operating parameters. For the position of the key influencing node in the risk propagation path, the node closer to the starting point of the risk propagation path may have a greater impact on the entire path because it may be the source of the failure or the node that is first affected by the failure. The position influence factor can be determined according to the distance of the key influencing node from the starting point of the risk propagation path. For example, the closer to the starting point, the greater the position influence factor.

[0182] For the abnormal degree of the operating parameters of the key influencing node, an abnormal degree value has been calculated previously. The position influence factor and the abnormal degree value of the operating parameters are comprehensively calculated, such as using multiplication or weighted summation, to obtain the quantification value of the influence degree of the key influencing node on the risk propagation path.

[0183] Combine the risk propagation path, key impact nodes, and the calculated quantified impact value to form a fault section location result containing the above information. This fault section location result can accurately indicate the section where the fault may occur, the key impact nodes, and the degree of influence of the key impact nodes on the fault propagation.

[0184] Step S150: Generate a fault location instruction containing a path identifier, a node identifier, and an impact degree according to the fault section location result, and send the fault location instruction to the distribution network intelligent operation and maintenance platform to trigger the precise maintenance process.

[0185] After obtaining the fault section location result, generate a fault location instruction containing a path identifier, a node identifier, and an impact degree according to the result, and send the fault location instruction to the distribution network intelligent operation and maintenance platform to trigger the precise maintenance process.

[0186] Step S151: Extract the risk propagation paths in the fault section location result, and assign a unique path identifier to each risk propagation path according to the preset path coding rules of the distribution network. The path identifier includes the starting node number and the ending node number.

[0187] First, extract the risk propagation paths from the fault section location result. Then, assign a unique path identifier to each risk propagation path according to the preset path coding rules of the distribution network.

[0188] The preset path coding rules of the distribution network are formulated to accurately identify and distinguish different risk propagation paths. The path identifier includes the starting node number and the ending node number, and these two numbers can uniquely determine the starting and ending positions of a risk propagation path.

[0189] For example, the starting node number can adopt the combination of the geographical location coding and the equipment type coding of the equipment node, and the same is true for the ending node number. In the above way, a unique path identifier is assigned to each risk propagation path, which is convenient for subsequent management and query.

[0190] Step S152: Extract the key impact nodes in the fault section location result, and assign a unique node identifier to each key impact node according to the preset node coding rules of the distribution network. The node identifier includes the node geographical location coding and the equipment type coding.

[0191] Next, extract the key impact nodes from the fault section location result. Assign a unique node identifier to each key impact node according to the preset node coding rules of the distribution network.

[0192] The node coding rule is formulated according to the actual situation of the distribution network and is used to accurately identify each device node. The node identifier includes the geographical location code of the node and the device type code. The geographical location code can adopt longitude and latitude information or regional division code, and the device type code is distinguished according to the function and type of the device, such as transformers, circuit breakers, etc.

[0193] By assigning a unique node identifier to each key impact node, the key impact nodes can be accurately located, facilitating maintenance personnel to quickly find the fault nodes.

[0194] Step S153: Extract the quantification value of the impact degree in the fault section location result, and convert the quantification value into an impact degree level identifier according to the preset impact degree grading standard.

[0195] Extract the quantification value of the impact degree from the fault section location result, and then convert the quantification value into an impact degree level identifier according to the preset impact degree grading standard.

[0196] The preset impact degree grading standard is set according to actual experience and the needs of fault handling, and is used to divide the continuous quantification value of the impact degree into different levels. For example, the quantification value of the impact degree can be divided into several intervals, and each interval corresponds to an impact degree level, such as mild impact, moderate impact, severe impact, etc.

[0197] Compare the extracted quantification value of the impact degree with the grading standard to determine the level interval it belongs to, so as to obtain the corresponding impact degree level identifier. This impact degree level identifier can intuitively reflect the impact degree of the key impact node on the fault propagation.

[0198] Step S154: Perform slicing processing on the risk propagation path in the time dimension to generate a segment identifier of the risk propagation path during the risk trigger period, and the segment identifier includes the start time and end time of the time period.

[0199] In order to more accurately describe the risk propagation path, perform slicing processing on the risk propagation path in the time dimension to generate a segment identifier of the risk propagation path during the risk trigger period.

[0200] The risk trigger period is the time range in which the risk event determined in the previous fault risk discrimination result occurs. According to the risk trigger period, the risk propagation path is divided according to time.

[0201] For example, the risk trigger period can be divided into several time periods according to the set time interval, and each time period corresponds to a segment of the risk propagation path. For each segment, record its start time and end time to form a segment identifier.

[0202] The segmentation identifier can display in detail the situation of the risk propagation path in different time periods, which helps maintenance personnel analyze the development process and trend of faults.

[0203] Step S155: Perform coordinate mapping processing on the key impact nodes in the spatial dimension to generate coordinate identifiers of the key impact nodes in the distribution network geographic information system, where the coordinate identifiers include longitude and latitude information.

[0204] Perform coordinate mapping processing on the key impact nodes in the spatial dimension to generate coordinate identifiers of the key impact nodes in the distribution network geographic information system.

[0205] The distribution network geographic information system records the geographical location information of each equipment node in the distribution network. By querying the geographic information system, the longitude and latitude information of the key impact nodes is obtained and combined into coordinate identifiers.

[0206] The coordinate identifier can accurately locate the position of the key impact node in the geographical space, facilitating maintenance personnel to quickly find the fault node during actual maintenance.

[0207] Step S156: Perform association and binding processing on the path identifier, node identifier, impact degree level identifier, segmentation identifier, and coordinate identifier to generate an identifier set containing multi-dimensional positioning information.

[0208] Perform association and binding processing on the previously generated path identifier, node identifier, impact degree level identifier, segmentation identifier, and coordinate identifier to generate an identifier set containing multi-dimensional positioning information.

[0209] The association and binding processing is to integrate these different types of identifier information so that a corresponding relationship is established between them. For example, associate the path identifier of a certain risk propagation path with the node identifier, impact degree level identifier, segmentation identifier of the key impact nodes on this path, and the coordinate identifier of the key impact nodes.

[0210] Through the association and binding processing, a complete identifier set is obtained. This identifier set contains the multi-dimensional information required for fault location and can comprehensively and accurately describe the location, impact degree, and development process of the fault.

[0211] Step S157: Input the identifier set into the instruction generation module, and perform data encapsulation processing according to the communication protocol format of the distribution network intelligent operation and maintenance platform to generate a fault location instruction containing a path identifier field, a node identifier field, and an impact degree field.

[0212] Input the generated identifier set into the instruction generation module, and perform data encapsulation processing according to the communication protocol format of the distribution network intelligent operation and maintenance platform to generate a fault location instruction.

[0213] The distribution network intelligent operation and maintenance platform has a compatible communication protocol format for receiving and processing externally sent instructions. The instruction generation module sorts and encapsulates the information in the identification set according to this communication protocol format.

[0214] Put information such as path identification, node identification, and impact degree level identification into corresponding fields, such as the path identification field, node identification field, and impact degree field. Through data encapsulation processing, the identification set is converted into a fault location instruction that conforms to the communication protocol format of the distribution network intelligent operation and maintenance platform.

[0215] Step S158: Send the fault location instruction to the distribution network intelligent operation and maintenance platform to trigger the precise maintenance process.

[0216] After generating the identification set containing multi-dimensional location information, it is necessary to send the fault location instruction encapsulated from this identification set to the distribution network intelligent operation and maintenance platform to trigger the precise maintenance process.

[0217] First of all, it is necessary to clarify the communication interface and protocol requirements of the distribution network intelligent operation and maintenance platform. Different distribution network intelligent operation and maintenance platforms may have different types of communication interfaces, such as serial communication, network communication, etc., and have their own specific communication protocols. The above protocols stipulate the data transmission format, encoding method, verification rules, etc. Therefore, before sending the fault location instruction, it is necessary to perform corresponding format conversion and encoding processing on the fault location instruction to make it conform to the communication protocol requirements of the distribution network intelligent operation and maintenance platform.

[0218] For the data transmission format, it is necessary to arrange information such as path identification, node identification, impact degree level identification, segment identification, and coordinate identification in the order and field length specified by the protocol. For example, the protocol may stipulate that the path identification is placed in the first few bits of the instruction, the node identification follows immediately, and the length of each identification is fixed. If the path identification needs to be represented by a set number of encodings, then the previously generated path identification needs to be filled or truncated to make its length meet the protocol requirements.

[0219] In terms of the encoding method, it may be necessary to convert the identification information into a specific encoding format, such as ASCII code, binary encoding, etc. For example, if the protocol uses ASCII code for data transmission, then it is necessary to convert the characters in information such as path identification and node identification into the corresponding ASCII code values. At the same time, in order to ensure the accuracy of data transmission, it is also necessary to add a check code according to the verification rules specified by the protocol. Common verification methods include parity check, cyclic redundancy check (CRC), etc. Taking CRC check as an example, it is necessary to calculate the data part in the fault location instruction to obtain a check code and add it to the end of the instruction.

[0220] After completing format conversion and encoding processing, the fault location instruction is sent to the distribution network intelligent operation and maintenance platform through the corresponding communication interface. If it is serial communication, parameters such as the baud rate, data bits, and stop bits of the serial port need to be configured, and then the serial communication program is used to send the instruction. During the sending process, it is necessary to ensure the stability of the communication line to avoid data loss or transmission errors. If it is network communication, the instruction needs to be sent to the specified IP address and port number of the distribution network intelligent operation and maintenance platform through the network connection.

[0221] When the distribution network intelligent operation and maintenance platform receives the fault location instruction, it can parse the fault location instruction. The parsing process is to decode the received encoded data according to the rules of the communication protocol to restore information such as path identification, node identification, impact degree level identification, section identification, and coordinate identification. Then, the platform will trigger the precise maintenance process based on these multi-dimensional location information.

[0222] The platform will determine the possible propagation path of the fault according to the path identification, which helps maintenance personnel understand the impact range and propagation direction of the fault. For example, through the path identification, it can be known which device node the fault starts to spread from, which intermediate nodes it passes through, and which terminal nodes are ultimately affected.

[0223] The node identification enables maintenance personnel to accurately locate the position of the key impact node. Combined with the coordinate identification, maintenance personnel can visually see the specific geographical location of the key impact node in the distribution network geographic information system and quickly reach the fault site. At the same time, the impact degree level identification provides maintenance personnel with information on the impact degree of the key impact node on the fault propagation, enabling them to reasonably arrange maintenance resources and priorities according to the impact degree.

[0224] The section identification records the section situation of the risk propagation path during the risk triggering period, which helps maintenance personnel analyze the development process of the fault. For example, by viewing the start time and end time of the time period in the section identification, maintenance personnel can understand the propagation speed and change trend of the fault in different time periods, so as to formulate more effective maintenance strategies.

[0225] After receiving the fault location instruction, the distribution network intelligent operation and maintenance platform can automatically generate a maintenance task order according to this information. The task order contains detailed information about the fault and maintenance requirements. Maintenance personnel can obtain the maintenance task order through mobile terminal devices or the operation interface of the platform and carry out precise maintenance according to the requirements of the task order, thereby improving the maintenance efficiency and reducing the impact of the fault on the operation of the distribution network.

[0226] In the above embodiments, it involves the processing of the target state data set collected by the global intelligent perception network of the distribution network, which may contain some privacy-sensitive data, such as the special operating parameters of some devices, environmental data of specific areas, etc. To protect this privacy-sensitive data, anonymization processing technology is adopted in the data collection stage. For the collected device operating condition data, environmental interference data, etc., the information that can directly or indirectly identify an individual or a specific device is removed. For example, for the operating parameters of a device, the specific device number or the name of the affiliated unit is not recorded, but is replaced by a randomly generated identifier. In this way, in the subsequent data processing and analysis process, even if the data is accidentally obtained, it is impossible to trace back to the specific device or individual through these data.

[0227] During the data transmission process, the data is encrypted using a symmetric encryption algorithm or an asymmetric encryption algorithm. In the symmetric encryption algorithm, the same key is used to encrypt and decrypt the data. For example, the AES encryption algorithm is adopted. At the data sending end, a key is used to encrypt the data into ciphertext, and at the receiving end, the same key is used to decrypt the ciphertext into plaintext. The asymmetric encryption algorithm uses a pair of keys, namely the public key and the private key. The data sending end uses the public key of the receiving end to encrypt the data, and the receiving end uses its own private key to decrypt it. This can ensure the security of the data during the transmission process and prevent the data from being stolen or tampered with.

[0228] In terms of data storage, the privacy-sensitive data is stored in a dedicated database, and strict access control is performed on the database. Only authorized personnel can access this data, and the access rights will be finely divided according to the responsibilities and needs of the personnel. For example, maintenance personnel can only access the device operating parameters related to the maintenance task and cannot access other sensitive environmental data. At the same time, the database is regularly backed up to prevent data loss or damage.

[0229] In the data processing and analysis stage, differential privacy technology protects the privacy of the data by adding noise to the data. When performing data analysis, a certain amount of noise is added to the input data so that the analysis result does not reveal the specific information of a single data point. For example, when calculating the statistical value of the device operating parameters, a random noise is added to each data point, and then the statistical calculation is performed. In this way, even if an attacker obtains the analysis result, it is impossible to infer the specific content of the original data.

[0230] The construction and training process of the dynamic decision inference engine is as follows: The dynamic decision inference engine mainly includes three core modules: the context awareness layer, the risk assessment layer, and the level classification layer.

[0231] The context awareness layer is the input layer of the dynamic decision-making inference engine, and its main function is to extract context features from the multi-dimensional feature map. This layer can be composed of multiple sub-modules, such as the state mutation point extraction sub-module and the abnormal connection edge extraction sub-module. The state mutation point extraction sub-module is responsible for identifying state mutation points from the time evolution trajectory, which can be achieved by performing operations such as differential calculation and threshold judgment on time series data. The abnormal connection edge extraction sub-module finds abnormal connection edges from the spatial cooperation relationship and determines the abnormal connection edges by comparing the changes in edge weights. The output results of these two sub-modules are passed as context features to the risk assessment layer.

[0232] The risk assessment layer is the core calculation layer of the dynamic decision-making inference engine. It receives the context features output by the context awareness layer and calculates the risk probability value of the device node through fuzzy logic inference processing. This layer includes a state mutation point severity assessment sub-module, an abnormal connection edge influence range assessment sub-module, a fuzzy inference rule base, and a fuzzy inference calculation sub-module. The state mutation point severity assessment sub-module classifies the state mutation points according to the numerical size of the abnormal working condition event, and the abnormal connection edge influence range assessment sub-module classifies the abnormal connection edges according to the degree of reduction of the edge weight. The fuzzy inference rule base stores the influence relationship between the combination of the state mutation level and the abnormal connection edge level on the risk probability value. The fuzzy inference calculation sub-module calculates the risk probability value of the device node through fuzzyfication, rule matching, and defuzzyfication processing based on the input state mutation level and abnormal connection edge level.

[0233] The level classification layer is the output layer of the dynamic decision-making inference engine. It classifies the risk probability value output by the risk assessment layer according to the preset risk level threshold to determine the risk level of the device node. This layer can be composed of a simple threshold comparison sub-module, which compares the risk probability value with different thresholds to determine the risk level identifier of the device node.

[0234] The above modules are connected in sequence according to the hierarchical structure. The output of the context awareness layer is used as the input of the risk assessment layer, and the output of the risk assessment layer is used as the input of the level classification layer, forming a complete inference process.

[0235] Before training the dynamic decision-making inference engine, a large amount of training data needs to be prepared. The training data includes multi-dimensional feature map samples and corresponding fault risk discrimination result labels. The multi-dimensional feature map samples can be obtained by performing spatio-temporal semantic fusion processing on the historical target state data set, and the fault risk discrimination result labels can be marked according to historical fault records and expert experience.

[0236] Step 1: Initialize the parameters of the dynamic decision-making inference engine. For the sub-modules of the context awareness layer, initialize the threshold parameters of the state mutation point extraction sub-module and the abnormal connection edge extraction sub-module. For example, set the differential threshold for state mutation point extraction, the edge weight change threshold for abnormal connection edge extraction, etc. For the fuzzy inference rule base of the risk assessment layer, initialize the premise conditions and conclusion parameters of the rules. For example, set the division thresholds for the state mutation level and the abnormal connection edge level, as well as the calculation parameters for the rule activation intensity. For the level division layer, initialize the risk level threshold.

[0237] Step 2: Input the multi-dimensional feature map samples in the training data into the context awareness layer to extract context features. According to the initialized threshold parameters, the state mutation point extraction sub-module and the abnormal connection edge extraction sub-module process the time evolution trajectory and spatial coordination relationship in the multi-dimensional feature map respectively to obtain context features.

[0238] Step 3: Input the context features into the risk assessment layer to calculate the risk probability value. The state mutation point severity assessment sub-module and the abnormal connection edge influence range assessment sub-module evaluate the context features respectively to obtain the state mutation level and the abnormal connection edge level. Then, the fuzzy inference calculation sub-module performs fuzzification, rule matching, and defuzzification processing on the state mutation level and the abnormal connection edge level according to the rules in the fuzzy inference rule base to calculate the risk probability value of the device node.

[0239] Step 4: Input the risk probability value into the level division layer to perform risk level division. The threshold comparison sub-module compares the risk probability value with the preset risk level threshold to determine the risk level identifier of the device node.

[0240] Step 5: Calculate the training error. Compare the risk level identifier output by the level division layer with the fault risk discrimination result label in the training data to calculate the error. The error can be calculated using methods such as the cross-entropy loss function and the mean square error.

[0241] Step 6: Update the parameters according to the error. Use an optimization algorithm (such as the stochastic gradient descent algorithm, Adagrad algorithm, etc.) to update the parameters of the dynamic decision-making inference engine. For example, for the threshold parameters of the context awareness layer, the fuzzy inference rule base parameters of the risk assessment layer, and the risk level threshold of the level division layer, adjust them according to the magnitude and direction of the error to gradually reduce the error.

[0242] Repeat the process from Step 2 to Step 6 until the training error reaches the preset threshold or the number of training times reaches the preset upper limit.

[0243] During the training process, some training parameters also need to be set, such as the learning rate, the training batch size, the number of training epochs, etc. The learning rate controls the step size of parameter updates. An overly large learning rate may lead to unstable parameter updates, while an overly small learning rate will result in a too slow training speed. The training batch size determines the number of samples used in each training. An appropriate batch size can improve the training efficiency. The number of training epochs represents the number of times the entire training data is trained. A sufficient number of training epochs can ensure that the model fully learns the features and patterns in the data.

[0244] With this design, the dynamic decision-making inference engine can learn the internal correlation between the multi-dimensional feature map and the fault risk discrimination result, so as to accurately discriminate the fault risk of the distribution network in practical applications.

[0245] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of a distribution network fault location system 100 based on intelligent decision-making that can implement the idea of the present application provided by some embodiments of the present application. For example, the processor 120 can be used on the distribution network fault location system 100 based on intelligent decision-making and is used to execute the functions in the present application.

[0246] The distribution network fault location system 100 based on intelligent decision-making can be a general-purpose server or a special-purpose server, both of which can be used to implement the distribution network fault location method based on intelligent decision-making of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0247] For example, the distribution network fault location system 100 based on intelligent decision-making can include a network port 110 connected to the network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the distribution network fault location system 100 based on intelligent decision-making can also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The distribution network fault location system 100 based on intelligent decision-making also includes an I / O interface 150 between the computer and other input / output devices.

[0248] For ease of explanation, only one processor is described in the distribution network fault location system 100 based on intelligent decision-making. However, it should be noted that the distribution network fault location system 100 based on intelligent decision-making in this application may also include multiple processors. Therefore, the steps performed by one processor described in this application can also be jointly performed or separately performed by multiple processors. For example, if the processor of the distribution network fault location system 100 based on intelligent decision-making performs step A and step B, it should be understood that step A and step B can also be jointly performed by two different processors or separately performed in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0249] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned distribution network fault location method based on intelligent decision-making is implemented.

[0250] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing or description thereof.

Claims

1. A distribution network fault location method based on intelligent decision-making, characterized in that, The method includes: Obtaining a set of target status data collected by the global intelligent perception network of the distribution network, where the set of target status data includes equipment operation condition data, network topology structure data, historical fault knowledge base, and environmental interference data; Performing spatio-temporal semantic fusion processing on the set of target status data to generate a multi-dimensional feature map reflecting the dynamic operation status of the distribution network, where the multi-dimensional feature map includes a state evolution trajectory in the time dimension and a device collaboration relationship in the space dimension; Performing context-aware discrimination processing on the multi-dimensional feature map through a pre-constructed dynamic decision-making inference engine to generate a fault risk discrimination result of the distribution network, where the fault risk discrimination result includes a risk level identifier, a risk trigger period, and a risk impact range; Performing causal link tracking processing on the set of target status data based on the fault risk discrimination result to generate a fault section location result of the distribution network, where the fault section location result includes a risk propagation path, key influence nodes, and an influence degree quantization value; Generating a fault location instruction including a path identifier, a node identifier, and an influence degree according to the fault section location result, and sending the fault location instruction to the distribution network intelligent operation and maintenance platform to trigger a maintenance process.

2. The method for locating distribution network faults based on intelligent decision-making according to claim 1, characterized in that, The performing spatio-temporal semantic fusion processing on the set of target status data to generate a multi-dimensional feature map reflecting the dynamic operation status of the distribution network includes: Performing time window alignment processing on the equipment operation condition data in the set of target status data to obtain a sequence of condition data with a time synchronization relationship, where the sequence of condition data includes time series sampling values of operation parameters in multiple dimensions; Performing graph model conversion processing on the network topology structure data in the set of target status data to generate a distribution network topology graph including equipment nodes and connection edges, where the weight of the connection edges is determined by the power transmission capacity between devices; Performing spatial region division processing on the environmental interference data in the set of target status data to generate an environmental parameter distribution matrix of the region where the equipment is located, where the environmental parameter distribution matrix includes spatial sampling values of environmental parameters in multiple dimensions; Inputting the sequence of condition data, the distribution network topology graph, and the environmental parameter distribution matrix into a semantic fusion module, and extracting the association rules between the equipment operation status and historical fault modes through semantic matching processing of the historical fault knowledge base; Performing feature cross-fusion processing on the sequence of condition data, the distribution network topology graph, and the environmental parameter distribution matrix based on the association rules to generate a multi-dimensional feature map including a time evolution trajectory and a spatial collaboration relationship, where the time evolution trajectory reflects the change law of the equipment status over time, and the spatial collaboration relationship reflects the mutual influence of the operation status between devices.

3. The method for locating distribution network faults based on intelligent decision-making according to claim 2, wherein The inputting the sequence of condition data, the distribution network topology graph, and the environmental parameter distribution matrix into a semantic fusion module, and extracting the association rules between the equipment operation status and historical fault modes through semantic matching processing of the historical fault knowledge base includes: Performing feature extraction processing on the sequence of condition data to generate a condition feature vector including abnormal condition events; Performing neighborhood feature aggregation processing on the distribution network topology graph to generate a topological feature vector including node degree, edge weight, and clustering coefficient; Performing statistical analysis on the environmental parameter distribution matrix to generate an environmental feature vector including humidity mean, temperature variance, and electromagnetic interference peak value; Inputting the working condition feature vector, topology feature vector and environment feature vector into the semantic matching unit of the historical fault knowledge base, and calculating the cosine similarity between each feature vector and the historical fault mode feature vector; According to the cosine similarity value, historical fault modes whose similarity exceeds a preset threshold are screened out, and the corresponding equipment state evolution rules and fault triggering conditions are extracted as association rules.

4. The method for locating distribution network faults based on intelligent decision-making according to claim 2, characterized in that, The method of performing feature cross-fusion processing on the operating condition data sequence, the distribution network topology map and the environmental parameter distribution matrix based on the association rules to generate a multi-dimensional feature map containing a time evolution trajectory and a spatial collaborative relationship includes: Perform sliding window analysis and processing on the operating condition data sequence in the time dimension, count the occurrence frequency of abnormal operating condition events in the sliding window, and generate a time evolution trajectory of the equipment state; Performing a path search process in the spatial dimension on the distribution network topology map, starting from a significant load device node, traversing all reachable nodes and recording edge weights on the path, and generating a spatial collaborative relationship between devices; Performing spatiotemporal joint analysis on the environmental parameter distribution matrix, counting the co-occurrence frequency of environmental parameters and equipment failures in the target area, and generating coupling characteristics of environmental interference and equipment status; Input the time evolution trajectory, spatial synergy relationship and coupling features into a feature map construction unit, and generate a multi-dimensional feature map with uniform time resolution and spatial resolution through feature dimension alignment processing; The multidimensional feature map is subjected to noise filtering processing to remove redundant features irrelevant to the historical failure mode, and the core features strongly correlated with the failure risk are retained as the final multidimensional feature map.

5. The method for locating distribution network faults based on intelligent decision-making according to claim 1, characterized in that, The pre-built dynamic decision-making inference engine performs context-aware discrimination processing on the multi-dimensional feature map to generate a fault risk discrimination result of the distribution network, including: Inputting the multidimensional feature map into the context perception layer of the dynamic decision-making reasoning engine, extracting the state mutation points in the time evolution trajectory and the abnormal connection edges in the spatial collaborative relationship as context features; Input the context features into the risk assessment layer of the dynamic decision-making reasoning engine, and calculate the risk probability value of the device node through fuzzy logic reasoning processing, wherein the risk probability value is jointly determined by the severity of the state mutation point and the influence range of the abnormal connection edge; Inputting the risk probability value into the level classification layer of the dynamic decision-making reasoning engine, classifying the device node into multiple risk levels according to a preset risk level threshold, and generating a risk level identifier; Performing timestamp analysis on the state mutation points in the time evolution trajectory to determine the start time and end time of the risk event and generate a risk trigger period; Perform regional coverage analysis on the abnormal connection edges in the spatial coordination relationship, determine the range of device nodes affected by the risk event, generate the risk impact range, and form a fault risk discrimination result including a risk level identifier, a risk trigger period, and a risk impact range.

6. The method for locating distribution network faults based on intelligent decision-making according to claim 5, wherein Input the context features into the risk assessment layer of the dynamic decision-making inference engine, and calculate the risk probability value of the device node through fuzzy logic reasoning processing, including: Perform severity assessment on the state mutation points in the context features, and divide the state mutation points into multiple state mutation levels according to the numerical size of the abnormal working condition event; Perform influence range assessment on the abnormal connection edges in the context features, and divide the abnormal connection edges into multiple abnormal connection edge levels according to the degree of reduction of the edge weight. The abnormal connection edge levels include local influence level, regional influence level, and global influence level; Construct a fuzzy inference rule base to define the influence relationship between the combination of the state mutation level and the abnormal connection edge level on the risk probability value; Input the state mutation level and the abnormal connection edge level into the fuzzy inference rule base, and generate the risk probability value of the device node through fuzzyfication, rule matching, and defuzzyfication processing.

7. The method for locating distribution network faults based on intelligent decision-making according to claim 1, wherein Perform causal link tracing on the target state data set based on the fault risk discrimination result to generate a fault section location result for the distribution network, including: If the risk level corresponding to the risk level identifier in the fault risk discrimination result is higher than the set risk level, extract the subset of state data overlapping with the risk trigger period in the target state data set; Perform causal relationship mining on the subset of state data, and construct a fault causal graph including device nodes and causal relationship edges. The weight of the causal relationship edge is determined by the mutual information value of the operating parameters between device nodes; Starting from the significant risk device node in the fault causal graph, traverse all reachable nodes through the depth-first search algorithm to generate a candidate set of risk propagation paths; Perform weight screening on the candidate set of risk propagation paths, and retain the risk propagation paths with mutual information values exceeding the preset threshold as effective risk propagation paths; Identify the node with the highest degree of abnormality in the operating parameters on the effective risk propagation path as the key impact node. The degree of abnormality is determined by the amplitude of the deviation of current, voltage, and temperature from the reference value; Calculate the quantization value of the influence degree of the key impact node on the risk propagation path, and form a fault section location result including the risk propagation path, the key impact node, and the influence degree quantization value. The influence degree quantization value is jointly determined by the position of the key impact node in the risk propagation path and the degree of abnormality of the operating parameters.

8. The method for locating distribution network faults based on intelligent decision-making according to claim 7, characterized in that Perform causal relationship mining on the subset of state data to construct a fault causal graph including device nodes and causal relationship edges, including: Perform Granger causality test on the device operating parameters in the subset of state data to determine the causal relationship direction between device nodes. The causal relationship direction indicates whether the state change of one node will cause the state change of another node. Calculate the mutual information value of the operating parameters of device nodes with causal relationships, where the mutual information value represents the correlation strength of the state changes between device nodes; Generate a directed causal relationship edge between device nodes according to the causal relationship direction and the mutual information value, where the direction of the edge is consistent with the causal relationship direction, and the weight of the edge is the mutual information value; Perform graph structure assembly processing on all device nodes and directed causal relationship edges to generate a fault causal graph reflecting the causal relationships between device nodes; Perform sparsification processing on the fault causal graph to remove weak causal relationship edges with mutual information values lower than a preset threshold.

9. The method for locating distribution network faults based on intelligent decision-making according to claim 1, wherein, The generating a fault location instruction including a path identifier, a node identifier, and an impact degree according to the fault section location result includes: Extract the risk propagation paths in the fault section location result, and assign a unique path identifier to each risk propagation path according to the preset path coding rules of the distribution network. The path identifier includes the starting node number and the ending node number; Extract the key impact nodes in the fault section location result, and assign a unique node identifier to each key impact node according to the preset node coding rules of the distribution network. The node identifier includes the node geographical location coding and the device type coding; Extract the impact degree quantization value in the fault section location result, and convert the quantization value into an impact degree level identifier according to the preset impact degree grading standard; Perform slicing processing on the risk propagation paths in the time dimension to generate segment identifiers of the risk propagation paths during the risk trigger period. The segment identifier includes the start time and the end time of the time period; Perform coordinate mapping processing on the key impact nodes in the space dimension to generate coordinate identifiers of the key impact nodes in the distribution network geographic information system. The coordinate identifier includes longitude and latitude information; Perform association and binding processing on the path identifier, the node identifier, the impact degree level identifier, the segment identifier, and the coordinate identifier to generate an identifier set containing multi-dimensional location information; Input the identifier set into the instruction generation module, and perform data encapsulation processing according to the communication protocol format of the distribution network intelligent operation and maintenance platform to generate a fault location instruction including a path identifier field, a node identifier field, and an impact degree field.

10. A distribution network fault location system based on intelligent decision-making, characterized in that, It includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions, or codes, and the processor is used to execute the programs, instructions, or codes in the memory to implement the distribution network fault location method according to any one of claims 1-9 above.

Images