Method and system for intelligently analyzing state of low-voltage equipment of distribution network based on edge calculation

Abstract

The invention discloses a distribution network low-voltage equipment state intelligent analysis method and system based on edge computing, and relates to the technical field of distribution network equipment state detection.The method comprises the steps that mesh network topology between edge computing nodes is constructed, and task allocation weights between the nodes are set; constructing a fault knowledge graph based on the multi-dimensional state features of the edge computing nodes; collaborative calculation is carried out based on the edge calculation nodes, calculation results of the edge calculation nodes are fused based on a weighted voting mechanism of a consistency algorithm, and a state evaluation result is obtained. According to the method, the multi-dimensional state features and the edge computing technology are combined, and intelligent analysis of the distribution network low-voltage equipment state is achieved. A mesh network topology is constructed based on electrical characteristics and environment characteristics, and efficient allocation of computing resources is realized through comprehensive evaluation of load complementation characteristics and state evaluation. The accuracy of fault propagation path identification is improved through double constraints of a feature association propagation chain and an equipment physical connection relationship.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network equipment status detection technology, and specifically to a method and system for intelligent analysis of the status of distribution network low-voltage equipment based on edge computing. Background Art

[0002] Condition monitoring of low-voltage equipment in distribution networks is crucial for ensuring the safe and stable operation of power systems. Traditional centralized monitoring architectures rely on centralized processing in data centers, which can lead to significant data transmission delays, inefficient computing resource utilization, and poor system scalability. With the development of edge computing technology, offloading computing tasks to edge nodes close to the data source has become an effective approach to addressing these issues. Currently, the application of edge computing in distribution network condition monitoring primarily focuses on single-node data processing and simple load balancing, lacking in-depth exploration of the collaborative computing capabilities between nodes.

[0003] Existing edge computing architectures have three major problems when handling low-voltage equipment status analysis tasks in distribution networks: First, the task allocation mechanism between nodes is overly simplistic, failing to fully account for differences in the computational complexity of device status characteristics and differences in node diagnostic capabilities; second, a lack of in-depth analysis of the correlations between device status characteristics prevents effective identification and prediction of fault propagation paths; and finally, an imperfect collaborative reasoning mechanism between nodes makes it difficult to accurately diagnose complex fault scenarios. These issues severely restrict the application of edge computing in distribution network status monitoring. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the centralized method used in the existing distribution network low-voltage equipment status monitoring has large data transmission delays, low computing resource utilization efficiency, poor system scalability, and the task allocation mechanism between nodes is too simple, which cannot effectively identify and predict the propagation path of faults, as well as how to achieve collaborative computing and reasoning between nodes to improve the accuracy of fault diagnosis.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing, comprising constructing a network topology between edge computing nodes, wherein the network topology is a Mesh network, and setting task allocation weights between nodes, wherein the task allocation weights of the nodes are determined based on historical task processing data, computing load, and the complexity of the computing tasks. A fault knowledge graph is constructed based on the multi-dimensional status characteristics of edge computing nodes. Collaborative computing is performed between edge computing nodes, and the calculation results of each node are integrated through the weighted voting mechanism of the consistency algorithm to obtain the final distribution network low-voltage equipment status assessment result.

[0007] As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in distribution networks based on edge computing described in the present invention, the method of constructing the network topology between edge computing nodes includes collecting historical task processing data of each edge computing node, calculating similarity based on multi-dimensional state characteristics, and constructing a node affinity matrix.

[0008] The edge computing nodes are grouped according to the node affinity matrix. If the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of each of the two nodes, and the load complementarity characteristics of the first node and the second node are greater than the load complementarity characteristics of each node with other nodes, the first node and the second node are divided into a group to form a mesh subnetwork.

[0009] If the first node and the second node only satisfy that the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of each of the two nodes, the two nodes are set to a standby collaborative relationship.

[0010] If the collaborative diagnosis accuracy and independent diagnosis accuracy of the first node and the second node do not meet the above conditions, the two nodes will remain in an independent operation state.

[0011] Weights are assigned to node computation tasks within each mesh subnetwork based on the computational complexity and state evaluation accuracy of multi-dimensional state characteristics.

[0012] As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing according to the present invention, wherein: the construction of the network topology between edge computing nodes also includes a collaborative diagnosis accuracy equal to the ratio of the number of electrical equipment status diagnosis tasks completed collaboratively by the two nodes to the total number of electrical equipment status diagnosis tasks undertaken, expressed as: .

[0013] Among them, R represents the accuracy of collaborative diagnosis, Indicates the number of electrical equipment status diagnosis tasks completed by the two nodes in collaboration. and They represent the number of the kth type of electrical equipment status diagnosis tasks completed independently by the two nodes, and M represents the total number of types of electrical equipment status diagnosis tasks.

[0014] The load complementarity characteristic is obtained by calculating the time complementarity of the load through the distribution of sampling data volume and multi-dimensional state characteristics of the distribution terminal, which is expressed as: .

[0015] Among them, C represents the load complementary characteristic, and Respectively represent the calculation load values ​​of the two nodes at time t, and They represent the amount of sampled data from the distribution terminals of the two nodes at time t, and T represents the time window.

[0016] The computational complexity of multi-dimensional state features is determined by the data size of the electrical and environmental features that need to be processed, and is expressed as: .

[0017] Among them, H represents the computational complexity of multi-dimensional state features, and Represents the data size of the pth electrical feature and the qth environmental feature respectively, p and q are the serial numbers of the electrical feature and the environmental feature respectively, p=1,2,...,P, q=1,2,...,Q, P and Q represent the number of electrical features and environmental features respectively, and represent the complexity weight coefficients of the pth electrical feature and the qth environmental feature, respectively. and is the balance factor.

[0018] The state assessment accuracy is the degree of consistency between the electrical equipment state diagnosis result output by the node and the actual operating state, which can be expressed as: .

[0019] Among them, A represents the state assessment accuracy, Indicates the status diagnosis result of the nth sample in the mth type fault, represents the actual operating status of the corresponding sample, Z represents the number of fault types, Indicates the number of samples of type m fault.

[0020] The task allocation weight is expressed as: .

[0021] in, represents the task assignment weight, Assign a weight coefficient to the task, and 0≤λ≤1.

[0022] As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in distribution networks based on edge computing described in the present invention, the method comprises: constructing a fault knowledge graph based on the multidimensional status characteristics of edge computing nodes, including receiving multidimensional status characteristics, constructing a feature association propagation chain, and characterizing the causal relationship between the multidimensional status characteristics through the feature association propagation chain.

[0023] Establish a mapping relationship between electrical equipment status and fault type, including: If the multi-dimensional state feature sequence of the electrical equipment to be analyzed shows a single feature abnormality, a fault map is constructed based on the causal relationship of the feature in the electrical features and environmental features.

[0024] If the multi-dimensional state feature sequence shows a multi-feature anomaly, the root feature is determined based on the causal relationship between the multi-dimensional state features, and a fault classification map is established step by step according to the causal relationship.

[0025] If a new change pattern appears in the multi-dimensional state feature sequence, a new fault feature template is constructed in the fault knowledge graph.

[0026] Based on the physical connection relationship and feature association propagation chain of electrical equipment, the electrical equipment association relationship in the fault knowledge graph is constructed, and the electrical equipment groups with common feature propagation paths are organized into fault propagation subgraphs.

[0027] As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in distribution networks based on edge computing described in the present invention, wherein: the multi-dimensional status characteristics based on edge computing nodes, the construction of the fault knowledge graph also includes the feature propagation path being jointly determined by the feature timing correlation and the physical connection relationship of the electrical equipment, and the physical connection relationship of the electrical equipment includes the power supply line connection relationship, the relative distance of the equipment installation location, and the upstream and downstream relationship of the power supply branch.

[0028] The weight of a characteristic propagation path is equal to a weighted combination of the characteristic timing correlation and the strength of the physical connection relationship of the electrical equipment.

[0029] The causal relationship is calculated by the causal analysis method based on transfer entropy and is expressed as: .

[0030] in, represents the transfer entropy of feature X to Y, represents the value of feature Y at time t+1, and Represent the k-order and l-order historical state vectors of Y and X respectively.

[0031] The feature temporal correlation is expressed as: .

[0032] in, represents the correlation coefficient under time lag τ, x and y are the time series of two features, N is the sequence length, and represent the mean values ​​of the series.

[0033] The strength of the physical connection relationship of electrical equipment is expressed as: .

[0034] in, Indicates the strength of the physical connection relationship between electrical devices i and j, Indicates the number of hops in the power supply path between electrical devices i and j, represents the installation location distance between electrical equipment i and j, Indicates the maximum installation position distance, Indicates the power supply branch correlation between electrical equipment i and j. If they are located in the same power supply branch, it is 1, otherwise it is 0. 、 and is the weight coefficient.

[0035] The edge weight of the fault propagation subgraph is determined based on the characteristic coupling degree and physical distance between electrical devices, and is expressed as: .

[0036] in, represents the edge weight between electrical equipment i and j, represents the characteristic coupling degree between electrical devices i and j, represents the physical distance between electrical devices i and j. α is the balancing factor.

[0037] The degree of feature coupling is calculated by mutual information entropy: .

[0038] in, represents the mutual information entropy of features X and Y, represents the joint probability distribution of X and Y, and Respectively represent the edge probability distribution of X and Y. As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing according to the present invention, wherein: the collaborative computing between edge computing nodes includes allocating the electrical equipment status diagnosis task to the edge computing nodes in the mesh subnetwork.

[0039] If the computing load of a node exceeds the processing capacity limit, the newly added electrical equipment status diagnosis task will be reallocated to other nodes with the highest task allocation weight and computing load that meets the requirements.

[0040] The computational load is expressed as: .

[0041] Where L represents the computational load, represents the historical average processing delay of the i-th electrical equipment status diagnosis task, represents the current computational load factor of the i-th electrical equipment status diagnosis task, and K represents the number of electrical equipment status diagnosis tasks undertaken by the node.

[0042] As a preferred solution of the method for intelligent analysis of the status of low-voltage equipment in the distribution network based on edge computing described in the present invention, the method comprises: fusing the calculation results of each node through a weighted voting mechanism of a consistency algorithm, including using a graph neural network to perform distributed reasoning based on the temporal correlation of the feature propagation path and the physical connection relationship of the electrical equipment, and realizing early warning and positioning prediction of distribution network low-voltage equipment failures by iteratively updating the feature representation of the node.

[0043] The update weight of the node feature is determined according to the feature coupling degree and the strength of the physical connection relationship of the electrical equipment, which can be expressed as: .

[0044] Among them, G represents the update weight of node features, I represents the degree of feature coupling, Indicates the strength of the physical connection relationship, is the feature update coefficient, and 0≤ ≤1.

[0045] The voting weight of a node in the weighted voting mechanism is positively correlated with the accuracy of state evaluation.

[0046] Another object of the present invention is to provide an intelligent analysis system for the status of low-voltage equipment in a distribution network based on edge computing. The system can achieve efficient allocation of computing resources through a mesh network topology constructed based on electrical characteristics and environmental characteristics, and through a comprehensive evaluation of load complementary characteristics and status assessment accuracy, thereby solving the problem of low resource allocation efficiency in the current centralized method used in the status monitoring of low-voltage equipment in the distribution network.

[0047] As a preferred solution of the intelligent analysis system of the state of low-voltage equipment in the distribution network based on edge computing described in the present invention, it includes: a network construction module, a graph construction module, and an evaluation module. The network construction module is used to extract multi-dimensional state features from the collected electrical equipment state quantities, and construct a mesh network topology between edge computing nodes based on the multi-dimensional state features, and set the task allocation weights between nodes. The graph construction module is used to construct a fault knowledge graph based on the multi-dimensional state features of the edge computing nodes. The evaluation module is used to collaboratively calculate the electrical equipment state diagnosis task between edge computing nodes through a mesh network based on the task allocation weight, use a graph neural network to perform distributed reasoning on the feature propagation path in the fault knowledge graph, realize fault warning and positioning, and fuse the calculation results of the edge computing nodes based on the weighted voting mechanism of the consistency algorithm to obtain a state evaluation result.

[0048] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing.

[0049] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing.

[0050] Beneficial effects of the present invention: The present invention realizes intelligent analysis of the status of low-voltage equipment in the distribution network by combining multi-dimensional state features and edge computing technology. First, the mesh network topology constructed based on electrical characteristics and environmental characteristics realizes efficient allocation of computing resources through comprehensive evaluation of load complementary characteristics and state assessment accuracy. Secondly, the method for constructing a multi-level fault knowledge graph improves the accuracy of fault propagation path identification through the dual constraints of feature association propagation chain and physical connection relationship of equipment. Finally, the distributed reasoning mechanism based on graph neural network, combined with the dynamic weight update of feature coupling degree and physical connection relationship strength, enhances the diagnostic capability in complex fault scenarios. This multi-level, multi-dimensional analysis method not only improves the utilization efficiency of edge computing resources, but also enhances the accuracy and explainability of fault diagnosis. The present invention achieves better results in resource allocation efficiency, accuracy and resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 An overall flow chart of a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing provided in the first embodiment of the present invention.

[0053] Figure 2 A module interaction diagram of a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing, provided as the third embodiment of the present invention. DETAILED DESCRIPTION

[0054] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0055] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing, comprising: S1: Build a mesh network topology between edge computing nodes and set task allocation weights between nodes. The task allocation weights of nodes are determined based on historical task processing data, computing load, and the complexity of computing tasks.

[0056] Furthermore, multi-dimensional state features are extracted based on the collected electrical equipment state quantities, and the mesh network topology between edge computing nodes is constructed based on the multi-dimensional state features, and the task allocation weights between nodes are set.

[0057] It should be noted that the multi-dimensional state characteristics include electrical characteristics and environmental characteristics. Electrical characteristics include voltage, current, active power, reactive power, and power factor, while environmental characteristics include temperature, humidity, and vibration.

[0058] It should also be noted that the historical task processing data of each edge computing node is collected, the similarity is calculated based on the multi-dimensional state characteristics, and the node affinity matrix is ​​constructed.

[0059] Among them, the node affinity matrix represents the ability of nodes to collaboratively process electrical equipment status diagnosis tasks.

[0060] It should be noted that the node affinity matrix originates from the concept of the attention matrix in graph neural networks, and the present invention applies it to the characterization of the collaborative capabilities of edge nodes. Each element of the matrix can be calculated by weighting multi-dimensional indicators such as the collaborative diagnosis accuracy and diagnostic latency of nodes on similar fault type tasks. For example, a Transformer-based self-attention mechanism can be used to calculate the association strength between nodes, or a spectral clustering method can be used to analyze the collaborative pattern of nodes, thereby quantifying the collaborative diagnostic capabilities and state similarity between nodes.

[0061] The similarity of multidimensional state features is based on distribution network state estimation theory and is calculated using feature vector distance. Algorithms such as cosine similarity and DTW (Dynamic Time Warping) can be used to calculate the similarity of multidimensional state features, or the Mahalanobis distance can be used to consider the correlation between features, thereby evaluating the node's ability to handle similar fault types.

[0062] Furthermore, the edge computing nodes are grouped according to the node affinity matrix: if the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of each of the two nodes, and the load complementarity characteristics of the first node and the second node are greater than the load complementarity characteristics of each node with other nodes, then the first node and the second node are divided into a group to form a mesh subnetwork.

[0063] If the first node and the second node only satisfy that the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of each of the two nodes, the two nodes are set to a standby collaborative relationship.

[0064] If the collaborative diagnosis accuracy and independent diagnosis accuracy of the first node and the second node do not meet the above conditions, the two nodes will remain in an independent operation state.

[0065] Weights are assigned to node computation tasks within each mesh subnetwork based on the computational complexity and state evaluation accuracy of multi-dimensional state characteristics.

[0066] It should be noted that the collaborative diagnosis accuracy is equal to the ratio of the number of electrical equipment status diagnosis tasks completed by the two nodes in collaboration to the total number of electrical equipment status diagnosis tasks undertaken, which is expressed as: .

[0067] Among them, R represents the accuracy of collaborative diagnosis, Indicates the number of electrical equipment status diagnosis tasks completed by the two nodes in collaboration. and Respectively represent the number of k-th type electrical equipment status diagnosis tasks completed independently by the two nodes, and M represents the total number of types of electrical equipment status diagnosis tasks. Based on the traditional accuracy evaluation method, it is used to evaluate the collaborative effect of edge nodes in the distribution network scenario. In the status diagnosis of low-voltage equipment in the distribution network, the types of electrical equipment status diagnosis tasks mainly include: abnormal diagnosis tasks based on electrical characteristics (such as overvoltage / undervoltage detection, harmonic content analysis, three-phase imbalance diagnosis, power factor abnormality diagnosis, etc.), status diagnosis tasks based on environmental characteristics (such as temperature over-limit warning, vibration abnormality detection, humidity impact analysis, etc.), and composite fault diagnosis tasks based on feature combination (such as temperature rise abnormality caused by overload, leakage caused by insulation degradation, etc.). The division of these task types directly corresponds to the multi-dimensional status features (including electrical features and environmental features) extracted by the present invention, ensuring the comprehensiveness and accuracy of status diagnosis.

[0068] The load complementarity characteristic is obtained by calculating the time complementarity of the load through the distribution of sampling data volume and multi-dimensional state characteristics of the distribution terminal, which is expressed as: .

[0069] Among them, C represents the load complementary characteristic, and Respectively represent the calculation load values ​​of the two nodes at time t, and Represent the amount of sampled data from the distribution terminals of the two nodes at time t, and T represents the time window. The calculation formula for the load complementarity characteristic combines the load difference and data distribution correlation. Among them, the first item evaluates the degree of load difference between node pairs, avoiding the problem of load peak and valley overlap caused by homogeneous node combinations in traditional methods. The second item solves the problem of resource scheduling lag caused by the burstiness of distribution network monitoring data through time series correlation analysis of the amount of sampled data from the distribution terminals. This design not only improves the utilization efficiency of edge computing resources, but also can adapt to the time-varying characteristics of low-voltage equipment status monitoring in the distribution network. In addition, the computational complexity of this method is low, making it suitable for real-time execution in an edge computing environment.

[0070] The computational complexity of multi-dimensional state features is determined by the data size of the electrical and environmental features that need to be processed, and is expressed as: .

[0071] Among them, H represents the computational complexity of multi-dimensional state features, and Represents the data size of the pth electrical feature and the qth environmental feature respectively, p and q are the serial numbers of the electrical feature and the environmental feature respectively, p=1,2,...,P, q=1,2,...,Q, P and Q represent the number of electrical features and environmental features respectively, and represent the complexity weight coefficients of the pth electrical feature and the qth environmental feature, respectively. and The balancing factor is used to balance the weights of electrical and environmental features in the computational complexity formula for multidimensional state features. This is necessary because electrical features (such as the fluctuation rate and current harmonic content calculated from voltage data) require more complex calculations, while environmental features (such as temperature change rate and vibration spectrum) are relatively simple to process. Furthermore, the data size and computational resource consumption of these two types of features differ significantly.

[0072] The state assessment accuracy is the degree of consistency between the electrical equipment state diagnosis result output by the node and the actual operating state, which can be expressed as: .

[0073] Among them, A represents the state assessment accuracy, Indicates the status diagnosis result of the nth sample in the mth type fault, represents the actual operating status of the corresponding sample, Z represents the number of fault types, Indicates the number of samples of type m fault.

[0074] For example, assume that an edge node in a distribution network monitors the operating status of a transformer. The node needs to diagnose two fault types: winding overheating and core failure (Z = 2). For each fault type, samples are collected at multiple time points: five samples for winding overheating (N1 = 5) and three samples for core failure (N2 = 3). The node performs fault diagnosis by analyzing the characteristics of the electrical equipment's status and outputs a diagnostic result value between 0 and 1 (0 indicates normal and 1 indicates a severe fault). This formula reflects the accuracy of the edge node's electrical equipment status diagnosis by calculating the deviation between the diagnostic result and the actual status.

[0075] The task allocation weight is expressed as: .

[0076] in, represents the task assignment weight, Assign a weight coefficient to the task, and 0≤λ≤1.

[0077] When the value is large, task allocation tends to favor nodes with low computational complexity, which is suitable for diagnostic tasks with high real-time requirements. When the value is small, task allocation tends to favor nodes with high state assessment accuracy, which is suitable for diagnostic tasks with high accuracy requirements. This formula determines the task allocation weight by combining the inverse of computational complexity and state assessment accuracy. Nodes with lower complexity and higher assessment accuracy receive higher task allocation weights.

[0078] S2: Build a fault knowledge graph based on the multi-dimensional state characteristics of edge computing nodes.

[0079] Specifically, the fault knowledge graph includes the mapping relationship between electrical equipment status and fault type, as well as the association relationship between electrical equipment.

[0080] Furthermore, it receives multi-dimensional state features and builds a feature association propagation chain.

[0081] The feature association propagation chain represents the causal relationship between multi-dimensional state features. The causal relationship is determined by transfer entropy.

[0082] The causal relationship is calculated by the causal analysis method based on transfer entropy: .

[0083] in, It represents the transfer entropy of feature X to Y. Represents the value of feature Y at time t+1. and denote the k-order and l-order historical state vectors of Y and X, respectively. p(·) denotes the probability distribution.

[0084] It should be noted that the calculation formula for the causal relationship between multi-dimensional state features is used to quantify the causal strength between features, and the larger the transfer entropy value, the more significant the causal relationship. The transfer entropy calculation formula originated from the field of information theory and was originally used to quantify the flow of information in time series data. The present invention has improved it: by introducing k-order and l-order historical feature vectors, it is able to capture the long-term dependencies between the multi-dimensional state features of distribution equipment. This improvement solves the limitations of traditional methods in dealing with the propagation of state features of distribution equipment. Traditional methods often only consider single-step information transmission and cannot identify complex fault propagation chains. For example, certain equipment failures may require analysis of feature changes on multiple time scales to accurately track their propagation paths. Through this improvement, the present invention can more accurately identify the root characteristics and propagation paths of faults, providing a more reliable basis for fault diagnosis.

[0085] It should be noted that the feature association propagation chain is constructed by quantifying the causal relationship between multi-dimensional state features. Specifically, the transfer entropy is calculated for each of the multi-dimensional state features extracted from S1. This reflects the degree of influence of feature X on Y. When the transfer entropy exceeds the set threshold, it is determined that a causal relationship exists between features X and Y, and the direction of the relationship is from X to Y. This constructs a directed network structure that represents the causal relationship between multidimensional state features, providing basic support for subsequent fault mapping.

[0086] It should also be noted that a mapping relationship between the electrical equipment status and the fault type is established.

[0087] If the multidimensional state feature sequence of the electrical equipment being analyzed exhibits a single feature anomaly, a fault map is constructed based on the causal relationship between that feature in the electrical and environmental features. If the multidimensional state feature sequence exhibits multiple feature anomalies, the root feature is determined based on the causal relationship between the multidimensional state features, and a fault classification map is established step by step based on the causal relationship. If a new change pattern appears in the multidimensional state feature sequence, a new fault feature template is constructed in the fault knowledge graph.

[0088] New change patterns include the mutation, periodic, attenuation, and fluctuation characteristics of multidimensional state feature time series curves, as well as the combined variation characteristics of multiple features. The mutation characteristic characterizes the amplitude and duration of a feature's jump. The jump point is detected by calculating the difference between feature values ​​at adjacent moments and setting a differential threshold. The mutation characteristic is identified by combining the duration of the period before and after the jump point. The periodic characteristic characterizes the recurrence pattern of a feature. Fast Fourier transform is used to analyze the spectrum of the feature sequence, and periodic patterns are identified by the dominant frequency component. The attenuation characteristic characterizes the decreasing trend of a feature. The attenuation envelope of the feature sequence is fitted using an exponential fitting method, and the attenuation rate is quantified using the fitting parameters. The fluctuation characteristic characterizes the oscillation frequency and amplitude of a feature. Wavelet transform is used for time-frequency analysis, and oscillation components of different scales are identified through energy distribution characteristics. The combined variation characteristic characterizes the phase relationship between multiple features. The instantaneous phase of the feature sequence is calculated using the Hilbert transform, and the synchronization between features is analyzed by phase difference.

[0089] It should be noted that different mapping strategies are employed for different anomaly patterns. For single-feature anomalies (e.g., a sudden voltage change at a specific moment), the present invention first identifies a set of features causally related to the anomaly based on transfer entropy analysis. The fault type is then determined by combining the variation patterns of these features, such as mutation, periodicity, attenuation, and fluctuation. For multi-feature anomalies (e.g., simultaneous voltage and current anomalies), the transfer entropy values ​​of each anomaly feature are compared to identify the root feature with the strongest causal influence. The fault type is then determined based on the variation patterns of this root feature and the resulting sequence of other feature changes. When a new variation pattern emerges, a new fault type feature template is constructed and stored in the knowledge graph based on the temporal characteristics of the relevant features (including mutation, periodicity, attenuation, fluctuation, etc.) and causal relationships. This hierarchical, multi-strategy mapping mechanism improves the accuracy and interpretability of fault diagnosis.

[0090] Furthermore, the electrical equipment association relationship in the fault knowledge graph is constructed based on the physical connection relationship and feature association propagation chain of the electrical equipment, and the electrical equipment groups with common feature propagation paths are organized into fault propagation subgraphs.

[0091] Specifically, the characteristic propagation path is determined by both the characteristic temporal correlation and the physical connection relationships between electrical devices. The physical connection relationships between electrical devices include the power supply line connections, the relative distances between device installation locations, and the upstream and downstream relationships between power supply branches. The weight of the characteristic propagation path is a weighted combination of the characteristic temporal correlation and the strength of the physical connection relationships between electrical devices. The edge weights of the fault propagation subgraph are determined based on the characteristic coupling degree and physical distance between electrical devices.

[0092] The characteristic propagation path reflects the propagation pattern of device status characteristics in the power distribution network. It is determined by two dimensions: characteristic temporal correlation and physical connection relationship. Temporal correlation reflects the propagation relationship of characteristics in the time dimension and can capture how faults propagate between different devices over time. Physical connection relationship provides spatial dimension constraints to ensure that the propagation path conforms to the actual network structure. This two-dimensional comprehensive determination method enables the system to more accurately track the propagation path of faults, avoiding the misjudgment that may result from considering only a single dimension. For example, when the characteristics of two devices are observed to have a high temporal correlation, their physical connection relationship is checked. Only when a physical and reasonable propagation path exists will it be confirmed as a valid propagation path.

[0093] The feature temporal correlation is expressed as: .

[0094] in, represents the correlation coefficient under time lag τ. x and y are the time series of two features. N is the length of the series. and represent the mean values ​​of the series.

[0095] It should be noted that the calculation formula for the timing correlation of the characteristic propagation path is used to identify the propagation delay between features and support the dynamic construction of the fault propagation path. The timing correlation calculation formula is based on the classical correlation coefficient calculation. The present invention introduces the time lag τ to enable the formula to capture the time delay effect of characteristic propagation, thereby solving the problem that traditional synchronous correlation analysis cannot effectively identify the timing of fault propagation. For example, in a power distribution network, the failure of a device may take a certain amount of time to affect related devices. This time delay effect is often ignored in traditional methods. Through this improvement, the present invention can more accurately characterize the propagation timing characteristics of the fault and improve the accuracy of fault diagnosis.

[0096] The strength of the physical connection relationship of electrical equipment is calculated according to the following formula: .

[0097] in, Indicates the strength of the physical connection relationship between electrical devices i and j, Indicates the number of hops in the power supply path between electrical devices i and j, represents the installation location distance between electrical equipment i and j, Indicates the maximum installation position distance, Indicates the power supply branch correlation between electrical equipment i and j (1 if they are in the same power supply branch, otherwise 0), 、 and is the weight coefficient.

[0098] It should be noted that the formula for calculating the strength of the physical connections between electrical devices integrates multiple physical characteristics of the power supply network. While the calculation methods for each parameter (such as hop count and distance) are known, this invention proposes a weighted combination of three factors. This invention introduces the power branch correlation, addressing the problem that traditional methods often overlook the impact of power supply topology on fault propagation. This design more accurately describes the physical connections between devices and provides more reliable physical constraints for fault propagation analysis in complex ring network structures.

[0099] The degree of feature coupling is calculated by mutual information entropy: .

[0100] in, Represents the mutual information entropy of features X and Y. represents the joint probability distribution of X and Y. and denote the marginal probability distributions of X and Y respectively.

[0101] It should be noted that the calculation formula for the degree of feature coupling measures the statistical correlation between two features and reflects the coupling strength of the features. Although the information entropy calculation formula is a classic information theory tool, the present invention has made adaptive adjustments when applying it to the feature coupling measurement of distribution equipment. Traditional mutual information calculations often assume that the data distribution is static, while the feature correlation in the distribution network is dynamically changing. The present invention enables the mutual information entropy to reflect the time-varying characteristics of the coupling relationship between features by dynamically updating the probability distribution estimate during the calculation process. This solves the problem that traditional static coupling analysis cannot adapt to the dynamic operating characteristics of the distribution network. Therefore, the present invention can better identify the correlation between transient faults and gradual faults.

[0102] The edge weight of the fault propagation subgraph is determined based on the characteristic coupling degree and physical distance between electrical devices, and is expressed as: .

[0103] in, represents the edge weight between electrical equipment i and j. Indicates the characteristic coupling degree between electrical devices i and j (calculated by mutual information entropy). represents the physical distance between electrical devices i and j. α is the balancing factor.

[0104] It should be noted that this invention combines characteristic coupling and physical distance through a balancing factor to account for a unique phenomenon in power distribution networks: the fault propagation characteristics of adjacent devices may be less affected by physical distance, while fault correlation between distant devices is more dependent on the physical path. By dynamically adjusting the balancing factor, this invention can adaptively balance the impact of these two factors in different scenarios, overcoming the limitations of traditional fixed-weight methods in complex networks.

[0105] S3: Collaborative computing is performed between edge computing nodes. The calculation results of each node are integrated through the weighted voting mechanism of the consistency algorithm to obtain the final distribution network low-voltage equipment status assessment result.

[0106] Furthermore, the electrical equipment status diagnosis task is collaboratively calculated among edge computing nodes through a mesh network based on the task allocation weight. A graph neural network is used to perform distributed reasoning on the feature propagation path in the fault knowledge graph to achieve fault warning and location. The calculation results of the edge computing nodes are fused based on the weighted voting mechanism of the consistency algorithm to obtain the status assessment result.

[0107] It should be noted that the processing capacity limit refers to the maximum computational load that an edge computing node can handle within a unit time window T. When the node's computational load L reaches this limit, a task reallocation mechanism is triggered to ensure that the node's processing efficiency of electrical equipment status diagnosis tasks is not affected by excessive load. The computational load meets the requirements when there is sufficient margin between the node's current computational load L and its processing capacity limit, ensuring that newly added electrical equipment status diagnosis tasks are processed in a timely manner and avoiding diagnostic delays caused by task accumulation.

[0108] Specifically, the load is calculated according to the following formula: .

[0109] Where L represents the computational load. represents the historical average processing delay of the i-th electrical equipment status diagnosis task. represents the current computational load factor of the i-th electrical equipment status diagnosis task. K represents the number of electrical equipment status diagnosis tasks undertaken by the node. T represents the time window.

[0110] It should also be noted that graph neural networks are used to perform distributed reasoning based on the temporal correlation of feature propagation paths and the physical connection relationships of electrical equipment, and to achieve early warning and location prediction of distribution network low-voltage equipment failures by iteratively updating the feature representation of nodes.

[0111] The update weight of the node feature is determined according to the feature coupling degree and the strength of the physical connection relationship of the electrical equipment: .

[0112] Where G represents the update weight of node features and I represents the degree of feature coupling. Indicates the strength of the physical connection relationship. is the feature update coefficient (0≤ ≤1), which is used to balance the influence of feature relevance and physical constraints in state updates.

[0113] Furthermore, during the distributed reasoning process, the graph neural network updates the feature representation of the node by simultaneously utilizing the temporal correlation of the feature propagation path and the physical connection strength of the electrical equipment. Specifically, for each node, its feature representation is iteratively updated by aggregating information from adjacent nodes, and the weight G in the update process is determined by the feature coupling degree I and the physical connection strength. This dual-constraint feature update mechanism can simultaneously consider the temporal evolution of electrical equipment status characteristics and the topological constraints of physical space, effectively improving the accuracy of fault warning and location prediction for low-voltage equipment in distribution networks.

[0114] It should be noted that the weighted voting mechanism based on the consensus algorithm integrates the calculation results of the edge computing nodes to obtain the status evaluation results.

[0115] Specifically, the voting weight of a node in the weighted voting mechanism is positively correlated with the accuracy of the state assessment. The voting weight of a node is calculated using the following formula: .

[0116] in, represents the voting weight of a node. A represents the state assessment accuracy. ω is the voting weight exponent (ω>1), which is used to strengthen the influence of high-precision nodes in the result fusion. This formula exponentially amplifies the impact of differences in state assessment accuracy on voting weight, helping to improve the accuracy of the final fusion result.

[0117] Specifically, the weighted voting mechanism of the consensus algorithm determines the voting weight of each edge computing node in the result fusion through the state evaluation accuracy A The voting weight increases exponentially with the improvement of state assessment accuracy. This nonlinear mapping relationship can highlight the contribution of high-precision nodes. During the fusion process, each node will have a corresponding impact on the state assessment results of the distribution network low-voltage equipment according to its voting weight. The final fusion result ensures the dominant role of high-precision nodes while not completely excluding the contributions of other nodes, thus achieving more reliable state assessment at the system level.

[0118] In summary, the present invention realizes the intelligent analysis of the status of low-voltage equipment in the distribution network by combining multi-dimensional state features and edge computing technology. First, the mesh network topology constructed based on electrical characteristics and environmental characteristics realizes the efficient allocation of computing resources through the comprehensive evaluation of load complementary characteristics and state assessment accuracy. Secondly, the construction method of the multi-level fault knowledge graph improves the accuracy of fault propagation path identification through the dual constraints of feature association propagation chain and physical connection relationship of equipment. Finally, the distributed reasoning mechanism based on graph neural network, combined with the dynamic weight update of feature coupling degree and physical connection relationship strength, enhances the diagnostic capability in complex fault scenarios. This multi-level and multi-dimensional analysis method not only improves the utilization efficiency of edge computing resources, but also enhances the accuracy and explainability of fault diagnosis.

[0119] Example 2, an embodiment of the present invention, provides a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0120] First, the experimental scenario was set in a city's low-voltage power distribution network, selecting multiple edge computing nodes for collaborative computing. The experimental data was collected from actual power distribution equipment, covering multi-dimensional state characteristics such as voltage, current, temperature, and load.

[0121] To build a mesh network topology, historical task processing data from each edge computing node is collected. The similarity of state characteristics between different nodes is calculated, and a node affinity matrix is ​​constructed based on this data. Based on this affinity matrix, the computing nodes are grouped to form multiple mesh subnetworks. To verify the rationality of the network topology, the collaborative diagnosis accuracy and load complementarity between different nodes are measured. If the collaborative diagnosis accuracy of two nodes is higher than their independent diagnosis accuracy and the load complementarity is high, they are grouped together; otherwise, they remain independent.

[0122] This experiment uses multidimensional state features to construct a fault knowledge graph, primarily encompassing electrical and environmental features and the relationships between them. During this construction process, the system characterizes the causal relationships between features through a feature propagation chain and, combined with the physical connections of the distribution network, forms a complete fault diagnosis knowledge base.

[0123] To improve the accuracy of status diagnosis, the experiment used a weighted voting mechanism based on a consensus algorithm to integrate the computational results of edge computing nodes. Computational tasks were first assigned to each edge computing node based on computational complexity and status assessment accuracy. If the computational load of a node exceeded a set threshold, some tasks were reassigned to other nodes with lower computational loads. After each node completed the computation, a weighted vote was performed based on the node's status assessment accuracy to ultimately determine the overall status assessment result.

[0124] The experiment ran for 30 days, collecting data on the computing efficiency, task allocation balance, and fault diagnosis accuracy of each edge computing node under different load conditions, and comparing them with traditional central server-based fault diagnosis methods. The experiment focused on the following key indicators: Calculates the load balancing degree (unit: %).

[0125] Average task response time (unit: ms).

[0126] Fault diagnosis accuracy (unit: %).

[0127] State assessment accuracy (unit: %).

[0128] Task assignment success rate (unit: %).

[0129] Table 1 Experimental data comparison table Node number Calculate load balancing (%) Average task response time (ms) Fault diagnosis accuracy (%) State assessment accuracy (%) Task assignment success rate (%) Node A 92.4 34.6 97.1 96.3 98.5 Node B 89.7 40.2 95.8 94.7 97.9 Node C 90.5 37.8 96.3 95.4 98.2 Node D 87.2 45.1 94.5 93.2 96.8 Node E 88.9 42.7 95.2 94.1 97.5 Node F 85.6 48.3 93.9 91.8 95.9 It can be seen from the experimental data that the method of intelligent analysis of the status of low-voltage equipment in the distribution network based on edge computing of the present invention shows significant advantages in computing load balancing, task allocation optimization and fault diagnosis accuracy.

[0130] This experiment used a dynamic task allocation weighting method to achieve a computational load balance of over 85% across all nodes, with a peak of 92.4%. Compared to traditional centralized computing methods, this method effectively avoids single-point overload and improves computing resource utilization.

[0131] Due to the distributed processing of tasks, edge computing nodes can execute tasks locally, reducing data transmission and processing latency. Experimental data shows that the average task response time ranges from 34.6ms to 48.3ms, which is approximately 40% lower than the traditional central server architecture and improves the real-time nature of fault diagnosis.

[0132] Through collaborative computing and a consensus voting mechanism, information between different nodes can be supplemented, improving overall fault diagnosis accuracy. Experimental data shows that the fault diagnosis accuracy of all nodes is above 93%, reaching a maximum of 97.1%, an improvement of approximately 5% compared to traditional methods.

[0133] This invention combines the causal relationships of multidimensional state features and builds a fault knowledge graph to improve the accuracy of state assessment. Experimental data shows that the state assessment accuracy is stable between 91.8% and 96.3%, a significant improvement over traditional centralized methods (typically between 85% and 90%).

[0134] During the experiment, the task allocation scheme based on computational complexity and load complementarity kept the task allocation success rate between 95.9% and 98.5%, indicating that tasks can be reasonably allocated according to the actual computing power of each node, ensuring the efficient use of computing resources.

[0135] Example 3, reference Figure 2 , is an embodiment of the present invention, which provides an intelligent analysis system for the status of low-voltage equipment in a distribution network based on edge computing, including a network construction module, a graph construction module, and an evaluation module.

[0136] Among them, the network construction module is used to build the network topology between edge computing nodes. The network topology is a Mesh network, and the task allocation weights between nodes are set. The task allocation weights of the nodes are determined based on the historical task processing data, computing load and complexity of the computing tasks. The graph construction module is used to build a fault knowledge graph based on the multi-dimensional state characteristics of the edge computing nodes. The evaluation module is used to perform collaborative calculations between edge computing nodes, and through the weighted voting mechanism of the consistency algorithm, the calculation results of each node are integrated to obtain the final distribution network low-voltage equipment status evaluation results.

[0137] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0138] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0139] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0140] It should be understood that various aspects of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gates for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gates, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc. It should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will understand that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and such modifications are intended to be encompassed by the claims of the present invention.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing, characterized in that: include: Build a mesh network topology between edge computing nodes and set task allocation weights between nodes. The task allocation weights of nodes are determined based on historical task processing data, computing load, and the complexity of computing tasks. Build a fault knowledge graph based on the multi-dimensional status characteristics of edge computing nodes; Collaborative computing is performed between edge computing nodes. Through the weighted voting mechanism of the consensus algorithm, the calculation results of each node are integrated to obtain the final distribution network low-voltage equipment status assessment results. The constructing of the network topology between the edge computing nodes includes collecting historical task processing data of each edge computing node, calculating similarity based on multi-dimensional state characteristics, and constructing a node affinity matrix; The edge computing nodes are grouped according to the node affinity matrix. If the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of the two nodes, and the load complementarity between the first node and the second node is greater than the load complementarity between each node and other nodes, the first node and the second node are grouped together to form a mesh subnetwork. If the first node and the second node only satisfy the condition that the collaborative diagnosis accuracy between the first node and the second node is greater than the independent diagnosis accuracy of the two nodes, then the two nodes are set to a standby collaborative relationship; If the collaborative diagnosis accuracy and independent diagnosis accuracy of the first node and the second node do not meet the conditions, the two nodes will remain in independent operation; Weights are assigned to node computation tasks within each mesh subnetwork based on the computational complexity and state evaluation accuracy of multi-dimensional state characteristics.

2. The method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing according to claim 1, characterized in that: The construction of the network topology between the edge computing nodes also includes a collaborative diagnosis accuracy rate equal to the ratio of the number of electrical equipment status diagnosis tasks completed by the two nodes in collaboration to the total number of electrical equipment status diagnosis tasks undertaken, expressed as: ； Among them, R represents the accuracy of collaborative diagnosis, Indicates the number of electrical equipment status diagnosis tasks completed by the two nodes in collaboration. and They represent the number of the kth type of electrical equipment status diagnosis tasks completed independently by the two nodes, and M represents the total number of types of electrical equipment status diagnosis tasks; The load complementarity characteristic is obtained by calculating the time complementarity of the load through the distribution of sampling data volume and multi-dimensional state characteristics of the distribution terminal, which is expressed as: ； Among them, C represents the load complementary characteristic, and Respectively represent the calculation load values ​​of the two nodes at time t, and They represent the amount of sampled data from the distribution terminals of the two nodes at time t, and T represents the time window; The computational complexity of multi-dimensional state features is determined by the data size of the electrical and environmental features that need to be processed, and is expressed as: ； Among them, H represents the computational complexity of multi-dimensional state features, and Represents the data size of the pth electrical feature and the qth environmental feature respectively, p and q are the serial numbers of the electrical feature and the environmental feature respectively, p=1,2,...,P, q=1,2,...,Q, P and Q represent the number of electrical features and environmental features respectively, and represent the complexity weight coefficients of the pth electrical feature and the qth environmental feature, respectively. and is the balance factor; The state assessment accuracy is the degree of consistency between the electrical equipment state diagnosis result output by the node and the actual operating state, which can be expressed as: ； Among them, A represents the state assessment accuracy, Indicates the status diagnosis result of the nth sample in the mth type fault, represents the actual operating status of the corresponding sample, Z represents the number of fault types, represents the number of samples of type m fault; The task allocation weight is expressed as: ； in, represents the task assignment weight, Assign a weight coefficient to the task, and 0≤λ≤1.

3. The method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing according to claim 2, characterized in that: The method of constructing a fault knowledge graph based on the multi-dimensional state features of the edge computing node includes receiving the multi-dimensional state features, Constructing a feature association propagation chain to represent the causal relationship between multi-dimensional state features; Establish a mapping relationship between electrical equipment status and fault type, including: If the multi-dimensional state feature sequence of the electrical equipment to be analyzed shows a single feature abnormality, a fault map is constructed based on the causal relationship of the feature in the electrical features and environmental features; If the multi-dimensional state feature sequence shows a multi-feature anomaly, the root feature is determined based on the causal relationship between the multi-dimensional state features, and a fault classification map is established step by step according to the causal relationship; If a new change pattern appears in the multi-dimensional state feature sequence, a new fault feature template is constructed in the fault knowledge graph; Based on the physical connection relationship and feature association propagation chain of electrical equipment, the electrical equipment association relationship in the fault knowledge graph is constructed, and the electrical equipment groups with common feature propagation paths are organized into fault propagation subgraphs.

4. The method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing according to claim 3, characterized in that: The multi-dimensional state characteristics of the edge computing nodes are used to construct the fault knowledge graph, which also includes the feature propagation path being determined by the feature time sequence correlation and the physical connection relationship of the electrical equipment. The physical connection relationship of the electrical equipment includes the power supply line connection relationship, the relative distance of the equipment installation location, and the upstream and downstream relationship of the power supply branch; The weight of the characteristic propagation path is equal to the weighted combination of the characteristic temporal correlation and the strength of the physical connection relationship of the electrical equipment; The causal relationship is calculated by the causal analysis method based on transfer entropy and is expressed as: ； in, represents the transfer entropy of feature X to Y, represents the value of feature Y at time t+1, and Represent the k-order and l-order historical state vectors of Y and X respectively; The feature temporal correlation is expressed as: ； in, represents the correlation coefficient under time lag τ, x and y are the time series of two features, N is the sequence length, and They represent the mean values ​​of the series respectively; The strength of the physical connection relationship of electrical equipment is expressed as: ； in, Indicates the strength of the physical connection relationship between electrical devices i and j, Indicates the number of hops in the power supply path between electrical devices i and j, represents the installation location distance between electrical equipment i and j, Indicates the maximum installation position distance, Indicates the power supply branch correlation between electrical equipment i and j. If they are located in the same power supply branch, it is 1, otherwise it is 0. 、 and is the weight coefficient; The edge weight of the fault propagation subgraph is determined based on the characteristic coupling degree and physical distance between electrical devices, and is expressed as: ； in, represents the edge weight between electrical equipment i and j, represents the characteristic coupling degree between electrical devices i and j, represents the physical distance between electrical devices i and j; α is the balance factor; The degree of feature coupling is calculated by mutual information entropy: ； in, represents the mutual information entropy of features X and Y, represents the joint probability distribution of X and Y, and denote the marginal probability distributions of X and Y respectively.

5. The method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing according to claim 4, characterized in that: The collaborative computing between edge computing nodes includes assigning electrical equipment status diagnosis tasks to edge computing nodes in the mesh subnetwork; If the computing load of a node exceeds the upper limit of its processing capacity, the newly added electrical equipment status diagnosis task will be reallocated to other nodes with the highest task allocation weight and computing load that meets the requirements; The computational load is expressed as: ； Where L represents the computational load, represents the historical average processing delay of the i-th electrical equipment status diagnosis task, represents the current computational load factor of the i-th electrical equipment status diagnosis task, and K represents the number of electrical equipment status diagnosis tasks undertaken by the node.

6. The method for intelligent analysis of low-voltage equipment status in a distribution network based on edge computing according to claim 4, characterized in that: The weighted voting mechanism of the consistency algorithm is used to fuse the calculation results of each node, including using a graph neural network to perform distributed reasoning based on the temporal correlation of feature propagation paths and the physical connection relationship of electrical equipment, and to achieve early warning and location prediction of distribution network low-voltage equipment failures by iteratively updating the feature representation of the nodes; The update weight of the node feature is determined according to the feature coupling degree and the strength of the physical connection relationship of the electrical equipment, which can be expressed as: ； Among them, G represents the update weight of node features, I represents the degree of feature coupling, Indicates the strength of the physical connection relationship, is the feature update coefficient, and 0≤ ≤1; The voting weight of a node in the weighted voting mechanism is positively correlated with the accuracy of state evaluation.

7. A system for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing, applying a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing as described in any one of claims 1 to 6, characterized in that: include: Network building module, graph building module, evaluation module; The network construction module is used to build the network topology between edge computing nodes, which is a Mesh network, and set the task allocation weights between nodes, where the task allocation weights of the nodes are determined based on historical task processing data, computing load, and the complexity of the computing tasks; The graph construction module is used to construct a fault knowledge graph based on the multi-dimensional state characteristics of the edge computing node; The evaluation module is used to perform collaborative calculations between edge computing nodes, and to integrate the calculation results of each node through the weighted voting mechanism of the consistency algorithm to obtain the final distribution network low-voltage equipment status evaluation results.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for intelligent analysis of the status of low-voltage equipment in a distribution network based on edge computing according to any one of claims 1 to 6 are implemented.

Images