Methods, devices, computer equipment, readable storage media and program products for handling power distribution network faults

Abstract

This application relates to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for handling power distribution network faults. The method includes: real-time acquisition of multi-source electrical data from the power distribution network through sensing devices at each node; extraction and fusion of the temporal and spatial features of the multi-source electrical data using a hierarchical strategy to obtain multi-dimensional fault precursor features; inputting the multi-dimensional fault precursor features into a pre-constructed adaptive early warning model to obtain fault prediction results for the power distribution network; generating early warning information based on the fault prediction results and dynamic early warning thresholds; and pushing the early warning information to multiple personnel based on preset rules. This method can uncover the spatiotemporal coupling characteristics of power distribution network operation, accurately capture fault precursors, and achieve dynamic adaptation of fault early warning.

Description

Technical Field

[0001] This application relates to the field of big data processing technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for handling power distribution network faults. Background Technology

[0002] With the accelerated construction of new power systems, distribution network fault early warning technology has emerged. This technology, characterized by early identification of potential distribution network faults and proactive avoidance of power supply faults, has become an important support for distribution network operation and maintenance.

[0003] In traditional technologies, SCADA (Supervisory Control and Data Acquisition) systems are used to set fixed thresholds for key nodes in the distribution network to generate simple alarms. Maintenance personnel then conduct manual inspections, fault location, and handling based on the alarm information. This approach primarily relies on passive response to handle distribution network faults. However, this process suffers from insufficient understanding of the spatiotemporal coupling characteristics of the distribution network, difficulty in capturing early signs of hidden faults, and a tendency to create a "protection dead zone" for fault identification. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for handling distribution network faults that can explore the spatiotemporal coupling characteristics of distribution network operation, accurately capture fault precursors, and achieve dynamic adaptation of fault warnings, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for handling power distribution network faults, including:

[0006] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0007] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0008] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0009] Based on preset rules, the warning information is pushed to multiple staff members.

[0010] In one embodiment, a hierarchical strategy is used to extract and fuse the temporal and spatial features of multi-source electrical data to obtain multi-dimensional fault precursor features, including:

[0011] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0012] In one embodiment, the process of generating the dynamic warning threshold includes:

[0013] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0014] In one embodiment, the process of constructing the adaptive early warning model includes:

[0015] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0016] In one embodiment, an initial adaptive early warning model is iteratively trained based on a fault dataset, wherein one iteration includes:

[0017] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0018] In one embodiment, multi-source electrical data of the distribution network are collected in real time by sensing devices at each node of the distribution network, including:

[0019] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0020] Secondly, this application also provides a power distribution network fault handling device, comprising:

[0021] The data acquisition module is used to collect multi-source electrical data of the power distribution network in real time through sensing devices at each node of the power distribution network;

[0022] The fusion module is used to extract and fuse the temporal and spatial features of multi-source electrical data through a hierarchical strategy to obtain multi-dimensional fault precursor features.

[0023] The generation module is used to input multi-dimensional fault precursor features into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; and to generate early warning information based on the fault prediction results and dynamic early warning thresholds.

[0024] The push module is used to send early warning information to multiple staff members based on preset rules.

[0025] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0026] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0027] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0028] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0029] Based on preset rules, the warning information is pushed to multiple staff members.

[0030] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0031] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0032] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0033] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0034] Based on preset rules, the warning information is pushed to multiple staff members.

[0035] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0036] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0037] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0038] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0039] Based on preset rules, the warning information is pushed to multiple staff members.

[0040] The aforementioned distribution network fault handling method, device, computer equipment, computer-readable storage medium, and computer program product collect multi-source electrical data of the distribution network in real time through sensing devices at each node of the distribution network; extract and fuse the temporal and spatial features of the multi-source electrical data using a hierarchical strategy to obtain multi-dimensional fault precursor features; input the multi-dimensional fault precursor features into a pre-constructed adaptive early warning model to obtain fault prediction results for the distribution network; generate early warning information based on the fault prediction results and dynamic early warning thresholds; and push the early warning information to multiple personnel based on preset rules. This application deeply mines the spatiotemporal coupling characteristics of multi-source electrical data through a hierarchical strategy, breaking through the limitations of traditional single-dimensional feature analysis. Combined with the collaborative judgment of the adaptive early warning model and dynamic early warning thresholds, it can accurately capture hidden fault precursors in the distribution network, effectively eliminate fault identification and protection dead zones, and improve the accuracy and adaptability of fault prediction. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a power distribution network fault handling method in one embodiment;

[0043] Figure 2 This is a schematic diagram of the process for generating fault precursor features in one embodiment;

[0044] Figure 3 This is a structural block diagram of a power distribution network fault handling device in one embodiment;

[0045] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0047] In one embodiment, such as Figure 1As shown, a method for handling power distribution network faults is provided. This embodiment illustrates the method applied to a terminal, but it is understood that the method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0048] Step 102: Collect multi-source electrical data of the distribution network in real time through the sensing devices at each node of the distribution network.

[0049] Optionally, sensing devices are deployed at various monitoring nodes on the distribution network's lines, transformers, switching equipment, and end-user side. The multi-source electrical data covers time-series electrical quantity data, including three-phase voltage, three-phase current, active power, reactive power, power factor, voltage harmonics, current harmonics, and load data during the operation of the distribution network. During the acquisition process, the multi-source electrical data is collected synchronously in real time to ensure the temporal consistency of data at each node, laying a data foundation for subsequent spatiotemporal feature extraction.

[0050] Step 104: Extract the time and spatial features of multi-source electrical data through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor features.

[0051] Optionally, fusion can be achieved by concatenating features from the time dimension and features from the spatial dimension, or by using a neural network.

[0052] Step 106: Input the multi-dimensional fault precursor features into the pre-built adaptive early warning model to obtain the fault prediction results of the distribution network; generate early warning information based on the fault prediction results and dynamic early warning thresholds.

[0053] Optionally, the adaptive early warning model can be a hybrid network based on convolutional neural networks and long short-term memory networks, combined with a fault determination rule base, or other hybrid models; no limitation is made here. The fault prediction results include the probability of fault occurrence, potential fault types, and risk levels for each area and device in the distribution network.

[0054] Step 108: Based on preset rules, push the warning information to multiple staff members.

[0055] The preset rules are based on a tiered push rule according to the warning level. The warning level is divided into three levels: general warning, relatively severe warning, and serious warning, according to the comparison between the fault prediction result and the dynamic warning threshold. General warning information is pushed to the on-site staff of the regional distribution network operation and maintenance team, relatively severe warning information is pushed to the regional operation and maintenance manager and core operation and maintenance staff, and serious warning information is pushed to the staff of the distribution network operation and maintenance management department, the emergency command center staff, and the corresponding responsible operation and maintenance personnel. During the push process, the complete content of the warning information and the priority prompt for handling are attached simultaneously to ensure that staff at different levels can accurately obtain the corresponding warning information and realize the efficient transmission of fault warnings and the tiered activation of handling response.

[0056] The aforementioned distribution network fault handling method, device, computer equipment, computer-readable storage medium, and computer program product collect multi-source electrical data of the distribution network in real time through sensing devices at each node of the distribution network; extract and fuse the temporal and spatial features of the multi-source electrical data using a hierarchical strategy to obtain multi-dimensional fault precursor features; input the multi-dimensional fault precursor features into a pre-constructed adaptive early warning model to obtain fault prediction results for the distribution network; generate early warning information based on the fault prediction results and dynamic early warning thresholds; and push the early warning information to multiple personnel based on preset rules. This application deeply mines the spatiotemporal coupling characteristics of multi-source electrical data through a hierarchical strategy, breaking through the limitations of traditional single-dimensional feature analysis. Combined with the collaborative judgment of the adaptive early warning model and dynamic early warning thresholds, it can accurately capture hidden fault precursors in the distribution network, effectively eliminate fault identification and protection dead zones, and improve the accuracy and adaptability of fault prediction.

[0057] In one exemplary embodiment, such as Figure 2 As shown, a hierarchical strategy is used to extract and fuse the temporal and spatial features of multi-source electrical data to obtain multi-dimensional fault precursor features, including:

[0058] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0059] In terms of time dimension, the sliding window algorithm is used to segment the time-series electrical data in the multi-source electrical data, extract the trend characteristics, abrupt change characteristics and periodic characteristics of the data in each time period, and capture the dynamic change law of the distribution network operation status. In terms of space dimension, the equipment association map is constructed by combining the distribution network topology, extracting the data coupling characteristics between different nodes of the same line, adjacent lines and related equipment, and identifying the spatial transmission law of equipment status abnormalities. The time and space features are integrated to form multi-dimensional fault precursor features. Specifically, in the time dimension feature extraction stage, a fixed-duration sliding window algorithm is used to process time-series data, with a window duration of 5-10 minutes and a step size of 2-3 minutes. Continuous time-series data such as voltage, current, and load are segmented. For each segment, trend features are extracted through linear regression fitting to identify gradual changes such as slow voltage shifts and steady load increases. Abrupt change features are extracted by calculating the threshold difference between adjacent data points to capture abnormal fluctuation signals such as sudden current increases and power factor abrupt changes. Periodic features are extracted using spectral analysis to match the inherent characteristics of the distribution network, such as daily load peak-valley cycles and weekly electricity consumption patterns, accurately capturing potential fault precursors that deviate from normal time-series patterns. Time-series correlation is performed on the features of each time period to construct a time-series feature chain, tracing the generation and development trajectory of abnormal signals. In the spatial dimension feature extraction stage, the distribution network topology in multi-source electrical data is combined, with lines and equipment as nodes and electrical connections as the basis for further analysis. Using connections as edges, a device association graph is constructed, labeling attributes such as device model, operating years, and association paths for each node. Based on the graph, data coupling characteristics of different monitoring nodes on the same line are extracted to analyze the coordinated changes in upstream and downstream voltage and current. Cross-influence characteristics of adjacent lines are extracted to identify the transmission effect of an anomaly on a neighboring line. Linkage characteristics between main and auxiliary equipment are extracted, such as the correlation between transformer temperature anomalies and corresponding switch contact temperatures, clarifying the spatial transmission path and impact range of device anomalies. In the dual-dimensional feature fusion stage, temporal and spatial features are fused using feature splicing to form a multi-dimensional feature set. The contribution of each feature to fault early warning is calculated using a random forest feature importance evaluation algorithm. Invalid features with a contribution below a set threshold are removed, such as minor fluctuations in environmental temperature and humidity that are not significantly related to the fault. Core features are retained, and finally, simplified fault precursor features are output.

[0060] For example, with a preset duration of 5-10 minutes, a sliding window algorithm is used to segment the time-series data of voltage, current, and load in 2-3 minute increments. Trend features are obtained by fitting each segment of data using linear regression. Adjacent abrupt change features are obtained by calculating the difference between adjacent data points and comparing it with a preset difference threshold. Periodic features are obtained by matching the electricity consumption cycle pattern using spectral analysis. These three types of features are integrated into time-dimensional features. Combining the physical connections of the distribution network's lines, transformers, and switching equipment, an equipment association graph is constructed with equipment / lines as nodes and electrical connections as edges. Based on the graph, the voltage and current coordination coefficients of upstream and downstream nodes on the same line are calculated to obtain data coupling features. The load transmission and fault propagation patterns of adjacent lines are analyzed to obtain cross-influence features. The state correlation change patterns between transformers and corresponding switching equipment are mined to obtain linkage features. These three types of features are integrated into spatial-dimensional features. Feature vectors of time-dimensional and spatial-dimensional features are fused using feature splicing. Invalid features with a contribution below a preset threshold are eliminated using a random forest feature importance evaluation algorithm, ultimately forming multi-dimensional fault precursor features.

[0061] In this embodiment, by refining electrical data through multi-dimensional feature extraction, the dynamic change patterns and abnormal fluctuation signals in the operation of the power distribution network can be accurately captured, the spatial transmission path and impact range of abnormal equipment status can be clearly identified, and more targeted and effective fault precursor features can be extracted. This effectively solves the problem that single-dimensional feature analysis in traditional technology is difficult to capture hidden fault precursors, and lays a solid feature foundation for subsequent accurate fault prediction.

[0062] In one exemplary embodiment, the process of generating the dynamic warning threshold includes:

[0063] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0064] Based on the operating characteristics and safety standards of different regions and types of equipment in the distribution network, an initial early warning threshold system is constructed, and the early warning levels are divided into general early warning, relatively severe early warning, and serious early warning. Combining real-time operating data and historical fault data, statistical analysis methods are used to dynamically calibrate the initial threshold, and load factor, environmental correction factor and equipment aging factor are introduced to dynamically adjust the threshold. For special scenarios such as extreme weather and major load changes, an emergency threshold calibration mechanism is triggered to automatically optimize the threshold range. Specifically, three levels of early warning thresholds are established based on the probability of fault occurrence and the scope of impact, clearly defining the boundaries of each threshold level and the corresponding handling priorities. The general early warning threshold corresponds to a fault occurrence probability of 5%-15%, with the impact limited to a single device or a local branch line, posing no risk of large-scale power outages. The more severe early warning threshold corresponds to a fault occurrence probability of 15%-30%, with the impact extending to the entire line or multiple related devices, potentially causing localized power outages. The critical early warning threshold corresponds to a fault occurrence probability exceeding 30%, affecting the regional distribution network and posing a risk of large-scale power outages and equipment damage. For different types of equipment such as transformers, circuit breakers, and lines, specific initial thresholds are set based on their operating years, rated parameters, and historical fault frequency. Value; During the dynamic parameter calibration phase, real-time operating data (i.e., multi-source electrical data) and historical fault data are used as data sources, and three core correction coefficients are introduced to achieve dynamic adjustment of the threshold; the load factor is updated in real time according to the peak and valley changes of regional electricity load, appropriately reducing the threshold sensitivity during peak load and increasing the threshold accuracy during low load; the environmental correction factor is adjusted in combination with environmental data such as temperature, icing, and precipitation, compressing the threshold range under extreme weather conditions and strengthening the early warning response; the equipment aging factor is set according to the equipment's operating years and the degree of insulation performance degradation, shifting the threshold range of aging equipment towards the low probability end, providing early warning of potential faults; the threshold boundaries at each level are updated through a combination of periodic calibration and real-time fine-tuning; the dynamic calibration formula for the early warning threshold is:

[0065]

[0066] in, The warning threshold is the one after dynamic calibration. The initial warning threshold is based on equipment characteristics and safety standards. , , The weighting coefficients are determined through training based on historical fault data. For load factor, As an environmental correction factor, This represents the equipment aging factor.

[0067] For example, based on the characteristics of urban and suburban distribution networks and the rated operating parameters and safety maintenance standards of different equipment such as transformers, switching equipment, and transmission lines, initial warning thresholds of three levels—general warning, relatively severe warning, and serious warning—are first defined, and specific initial threshold ranges are set for different regions and different types of equipment. Then, real-time collected multi-source electrical data reflects the current operating status of the distribution network. Statistical analysis is performed by combining the probability of fault occurrence and the pattern of operating condition changes in historical multi-source electrical data. At the same time, load factors that change in real time with the peak and valley of regional electricity load, environmental correction factors that are dynamically updated based on environmental data such as temperature, icing, and precipitation, and equipment aging coefficients set according to the equipment's operating years and insulation performance degradation are introduced. The three levels of initial warning thresholds are numerically corrected and the ranges are adjusted through a preset threshold calibration formula. For special scenarios such as extreme weather and major load changes, an emergency threshold calibration mechanism is triggered to further optimize the threshold ranges, ultimately obtaining dynamic warning thresholds that are adapted to the actual operating conditions of the distribution network and can be dynamically updated in real time.

[0068] In this embodiment, by setting multi-level initial warning thresholds based on the differentiated characteristics of distribution network areas and equipment types, the threshold system can be made to fit the actual operating requirements and safety standards of different equipment, avoiding the problem of insufficient adaptability of a single threshold. At the same time, based on real-time and historical multi-source electrical data, and by introducing multi-dimensional correction factors such as load, environment, and equipment aging, the initial thresholds are dynamically adjusted, allowing the warning thresholds to adapt in real time to changes in the actual operating conditions of the distribution network. This effectively solves the problem that traditional fixed thresholds cannot match dynamic changes in operating conditions and are prone to misjudgment and missed judgment. It provides a scientific and dynamic threshold basis for the accurate determination of fault warning levels, improving the accuracy and rationality of fault warnings.

[0069] In one exemplary embodiment, the process of constructing an adaptive early warning model includes:

[0070] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0071] For example, the hybrid network adopts an architecture that combines a convolutional neural network (CNN) and a long short-term memory (LSTM) network. The CNN extracts key local features from fault precursor samples, while the LSTM captures long-term dependencies in temporal feature samples. These two networks are combined to form a hybrid network with deep spatiotemporal feature mining capabilities. A fault judgment rule base is constructed based on distribution network equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling records. The output of the hybrid network is jointly judged with the rule base to form an initial adaptive early warning model. Multi-source electrical data samples containing normal operating conditions, abnormal operating conditions, and historical faults, along with corresponding fault label data, are collected. Multi-dimensional fault precursor feature samples are extracted and fused according to a spatiotemporal hierarchical strategy to construct a fully labeled fault dataset. This fault dataset is used to iteratively train and validate the initial adaptive early warning model, continuously optimizing network parameters and fault judgment logic, ultimately resulting in a trained adaptive early warning model with high accuracy and strong generalization ability.

[0072] In this embodiment, by constructing a hybrid network and combining it with a fault determination rule base to form an adaptive early warning model, it is possible to simultaneously and deeply mine the spatial features and temporal dependencies of fault precursors. By constructing a high-quality sample set based on multi-source electrical data and fault data and completing iterative training, the model can adaptively learn the fault patterns under complex operating conditions of the distribution network, effectively improving the accuracy and robustness of fault prediction and solving the problems of low early warning accuracy and weak generalization ability of traditional single models.

[0073] In an exemplary embodiment, an initial adaptive early warning model is iteratively trained based on a fault dataset, wherein one iteration includes:

[0074] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0075] For example, the fault precursor feature samples in the fault dataset are divided into a training set and a validation set in a 7:3 ratio. Batch feature samples from the training set are input into the hybrid network optimized in the previous iteration to obtain the initial fault prediction for the current iteration. The fault judgment threshold and typical fault feature matching rules for the corresponding device / region are retrieved from the fault judgment rule base to perform rule verification on the initial fault prediction result, forming the fault prediction output for the current iteration. Based on the real fault labels labeled in the fault dataset, the deviation between the fault prediction output and the real labels is calculated using a loss function to obtain the network loss for the current iteration. The network loss is backpropagated based on the gradient descent algorithm to update the parameters of the hybrid network and generate the adaptive early warning model for the current iteration. It is determined whether the current network loss has stopped decreasing for five consecutive iterations. If it has not stopped decreasing for five consecutive iterations, the next batch of training samples is selected for the next iteration. If it has, the iteration is terminated, and the current optimized model is used as the trained adaptive early warning model.

[0076] In this embodiment, by combining the intelligent prediction of the hybrid network with the rule verification of the fault judgment rule base for iterative training, the dual advantages of deep learning and expert rules can be fully utilized to improve the reliability of fault prediction. The hybrid network parameters are updated in reverse based on the network loss, which allows the model to gradually converge and accurately match the fault pattern. By setting preset conditions to determine whether to terminate training, overfitting can be avoided while ensuring the accuracy of the early warning, and finally a stable, efficient and highly adaptive fault early warning model is obtained.

[0077] In one exemplary embodiment, multi-source electrical data of the distribution network are collected in real time by sensing devices at each node of the distribution network, including:

[0078] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0079] For example, by deploying sensing devices such as voltage transformers, current transformers, and load monitoring terminals at key monitoring nodes on the distribution network lines, transformers, switching equipment, and end-user side, electrical data such as three-phase voltage, three-phase current, and real-time load of each node during actual operation are collected synchronously, forming initial multi-source electrical data containing multiple devices, multiple time periods, and multiple regions. Abnormal jump values, missing values, and noise interference data in the initial multi-source electrical data are identified, removed, and supplemented. The data format, sampling frequency, and timestamp are unified to generate standardized, unified, and directly usable multi-source electrical data in a standardized format.

[0080] In this embodiment, by deploying sensing devices on multiple types of equipment and key nodes in the power distribution network for full-domain data collection, it is possible to ensure comprehensive data coverage and reliable sources, avoiding monitoring blind spots. By screening, denoising, and standardizing the initial electrical data, the data quality can be effectively improved, providing an accurate and stable data foundation for subsequent spatiotemporal feature extraction, model training, and fault early warning.

[0081] In an exemplary embodiment, a power distribution network fault handling method includes: real-time acquisition of multi-source electrical data of the power distribution network through sensing devices at each node of the power distribution network; segmentation of the time-series electrical data in the multi-source electrical data based on a preset duration to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features; constructing an equipment association map based on the power distribution network topology data in the multi-source electrical data; extracting data coupling features of different nodes on the same line, cross-influence features of adjacent lines, and linkage features between main equipment and auxiliary equipment based on the equipment association map, forming spatial-dimensional features; fusing the extracted time-dimensional features and spatial-dimensional features to form multi-dimensional fault precursor features; inputting the multi-dimensional fault precursor features into a pre-constructed adaptive early warning model to obtain fault prediction results for the power distribution network; generating early warning information based on the fault prediction results and dynamic early warning thresholds; and pushing the early warning information to multiple personnel based on preset rules. The generation process of dynamic early warning thresholds includes: generating multi-level initial early warning thresholds based on the operating characteristics and safety standards of different areas and types of equipment in the distribution network; dynamically adjusting the multi-level initial early warning thresholds based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor, and equipment aging factor to obtain dynamic early warning thresholds. The construction process of the adaptive early warning model includes: constructing a hybrid network; building a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; constructing an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collecting multi-source electrical data samples and corresponding fault data in real time through sensing devices at each node of the distribution network; extracting and fusing time-dimensional and spatial-dimensional feature samples from the multi-source electrical data samples using a hierarchical strategy to obtain multi-dimensional fault precursor feature samples; generating a fault dataset based on the fault precursor feature samples and fault data; iteratively training the initial adaptive early warning model based on the fault dataset to obtain a trained adaptive early warning model; one iteration includes: converting faults in the fault dataset... The precursor feature samples are input into the hybrid network of the adaptive early warning model obtained from the previous iteration of training to obtain the initial fault prediction for the current iteration of training. The initial fault prediction for the current iteration of training is verified based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration of training. The loss is calculated based on the fault prediction output for the current iteration of training and the fault data in the fault dataset to obtain the network loss for the current iteration of training. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration of training to obtain the adaptive early warning model for the current iteration of training. It is determined whether the network loss for the current iteration of training meets the preset conditions. If it does not meet the conditions, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model for the current iteration is taken as the completed adaptive early warning model.

[0082] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0083] In one exemplary embodiment, such as Figure 3 As shown, a power distribution network fault handling device is provided, comprising: a data acquisition module 301, a data fusion module 302, a data generation module 303, and a data push module 304, wherein:

[0084] The data acquisition module is used to collect multi-source electrical data of the power distribution network in real time through sensing devices at each node of the power distribution network;

[0085] The fusion module is used to extract and fuse the temporal and spatial features of multi-source electrical data through a hierarchical strategy to obtain multi-dimensional fault precursor features.

[0086] The generation module is used to input multi-dimensional fault precursor features into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; and to generate early warning information based on the fault prediction results and dynamic early warning thresholds.

[0087] The push module is used to send early warning information to multiple staff members based on preset rules.

[0088] In one exemplary embodiment, the fusion module is further configured to:

[0089] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0090] In one exemplary embodiment, the generation module is further configured to:

[0091] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0092] In one exemplary embodiment, it further includes: a construction module for:

[0093] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0094] In one exemplary embodiment, the building module is further configured to:

[0095] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0096] In one exemplary embodiment, the acquisition module is used for:

[0097] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0098] Each module in the aforementioned power distribution network fault handling device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0099] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores multi-source electrical data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a power distribution network fault handling method.

[0100] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0101] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0102] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0103] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0104] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0105] Based on preset rules, the warning information is pushed to multiple staff members.

[0106] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0107] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0108] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0109] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0110] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0111] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0112] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0113] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0114] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0115] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0116] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0117] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0118] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0119] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0120] Based on preset rules, the warning information is pushed to multiple staff members.

[0121] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0122] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0123] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0124] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0125] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0126] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0127] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0128] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0129] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0130] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0131] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0132] Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network.

[0133] By extracting and fusing the temporal and spatial features of multi-source electrical data using a hierarchical strategy, multi-dimensional fault precursor features are obtained.

[0134] Multi-dimensional fault precursor features are input into a pre-built adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated.

[0135] Based on preset rules, the warning information is pushed to multiple staff members.

[0136] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0137] Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented and processed to obtain trend features reflecting data change trends, adjacent abrupt change features reflecting differences between adjacent data, and periodic features reflecting the periodicity of data, forming time-dimensional features. Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed. Based on the equipment association map, the data coupling features of different nodes on the same line, the cross-influence features of adjacent lines, and the linkage features between main equipment and auxiliary equipment are extracted to form spatial-dimensional features. The extracted time-dimensional features and spatial-dimensional features are fused to form multi-dimensional fault precursor features.

[0138] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0139] Based on the operating characteristics and safety standards of different regions and types of equipment in the power distribution network, multi-level initial warning thresholds are generated. Based on multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor and equipment aging factor, the multi-level initial warning thresholds are dynamically adjusted to obtain dynamic warning thresholds.

[0140] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0141] Construct a hybrid network; build a fault judgment rule base for the distribution network based on equipment operation and maintenance standards, fault diagnosis specifications, and historical fault handling experience; construct an initial adaptive early warning model based on the hybrid network and the fault judgment rule base; collect multi-source electrical data samples and corresponding fault data of the distribution network in real time through sensing devices on each node of the distribution network; extract time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor feature samples; generate a fault dataset based on the fault precursor feature samples and fault data; iteratively train the initial adaptive early warning model based on the fault dataset to obtain the trained adaptive early warning model.

[0142] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0143] The fault precursor feature samples from the fault dataset are input into the hybrid network of the adaptive early warning model obtained from the previous iteration to obtain the initial fault prediction for the current iteration. The initial fault prediction for the current iteration is validated based on the fault judgment rule base in the initial adaptive early warning model to obtain the fault prediction output for the current iteration. The loss is calculated based on the fault prediction output for the current iteration and the fault data in the fault dataset to obtain the network loss for the current iteration. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss for the current iteration to obtain the adaptive early warning model for the current iteration. It is determined whether the network loss for the current iteration meets the preset conditions. If not, the next iteration of training is performed. If it does meet the conditions, the adaptive early warning model of the current iteration is taken as the completed adaptive early warning model.

[0144] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0145] By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment, and end-user side, voltage, current, and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; the initial multi-source electrical data is then filtered and denoised to generate multi-source electrical data in a standardized format.

[0146] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0148] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for handling faults in a power distribution network, characterized in that, The method includes: Multi-source electrical data of the power distribution network are collected in real time by sensing devices at each node of the power distribution network. By extracting the temporal and spatial features of the multi-source electrical data through a hierarchical strategy and fusing them, multi-dimensional fault precursor features are obtained. The multi-dimensional fault precursor features are input into a pre-constructed adaptive early warning model to obtain fault prediction results for the distribution network; based on the fault prediction results and dynamic early warning thresholds, early warning information is generated. Based on preset rules, the warning information is pushed to multiple staff members; The process involves extracting and fusing the temporal and spatial features of the multi-source electrical data using a hierarchical strategy to obtain multi-dimensional fault precursor features, including: Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented to obtain segmented multi-source electrical data. Linear regression fitting was performed on the segmented multi-source electrical data to obtain trend features that reflect the data change trend; The difference between adjacent data points is calculated for the segmented multi-source electrical data and compared with a preset difference threshold to obtain the adjacent abrupt change characteristics that reflect the difference between adjacent data. By matching the segmented multi-source electrical data with the daily load data of the distribution network using the spectrum analysis method, periodic features reflecting the periodicity of the data are obtained; the trend features, the adjacent abrupt change features, and the periodic features constitute the time dimension features; Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed; based on the equipment association map, data coupling characteristics of different nodes on the same line, cross-influence characteristics of adjacent lines, and linkage characteristics between main equipment and auxiliary equipment are extracted to form spatial dimension characteristics; The extracted temporal and spatial features are fused to form multi-dimensional fault precursor features.

2. The method according to claim 1, characterized in that, The process of generating the dynamic early warning threshold includes: Based on the operating characteristics and safety standards of different areas and types of equipment in the power distribution network, multi-level initial early warning thresholds are generated. Based on the multi-source electrical data, historical multi-source electrical data, load factor, environmental correction factor, and equipment aging factor, the multi-level initial warning threshold is dynamically adjusted to obtain a dynamic warning threshold.

3. The method according to claim 1, characterized in that, The construction process of the adaptive early warning model includes: Construct a hybrid network; build a fault judgment rule base for the distribution network based on the equipment operation and maintenance standards, fault diagnosis specifications and historical fault handling experience of the distribution network. An initial adaptive early warning model is constructed based on the hybrid network and the fault determination rule base. The multi-source electrical data samples and corresponding fault data of the distribution network are collected in real time by sensing devices at each node of the distribution network. By extracting the time dimension feature samples and spatial dimension feature samples of the multi-source electrical data samples through a hierarchical strategy and fusing them, multi-dimensional fault precursor feature samples are obtained. A fault dataset is generated based on the fault precursor feature samples and the fault data. The initial adaptive early warning model is iteratively trained based on the fault dataset to obtain the trained adaptive early warning model.

4. The method according to claim 3, characterized in that, The initial adaptive early warning model is iteratively trained based on the fault dataset, wherein one iteration includes: Input the fault precursor feature samples from the fault dataset into the hybrid network of the adaptive early warning model obtained from the previous iteration of training to obtain the initial fault prediction for the current iteration of training. The initial fault prediction of the current iteration training is verified based on the fault determination rule base in the initial adaptive early warning model to obtain the fault prediction output of the current iteration training. The network loss for the current iteration of training is calculated based on the fault prediction output of the current iteration and the fault data in the fault dataset. The parameters of the hybrid network in the initial adaptive early warning model are updated based on the network loss of the current iteration training to obtain the adaptive early warning model of the current iteration training. Determine whether the network loss of the current iteration of training has reached a preset condition; If the target is not met, proceed to the next iteration of training; If this is achieved, the adaptive early warning model of the current iteration will be used as the completed adaptive early warning model.

5. The method according to claim 1, characterized in that, The real-time collection of multi-source electrical data of the distribution network through sensing devices at each node of the distribution network includes: By deploying sensing devices at various nodes in the distribution network lines, transformers, switching equipment and end users, voltage, current and load data during the operation of the distribution network are collected in real time to form initial multi-source electrical data; The initial multi-source electrical data is filtered and denoised to generate multi-source electrical data in a standardized format.

6. A power distribution network fault handling device, characterized in that, The device includes: The data acquisition module is used to collect multi-source electrical data of the power distribution network in real time through sensing devices at each node of the power distribution network; The fusion module is used to extract the temporal and spatial features of the multi-source electrical data through a hierarchical strategy and fuse them to obtain multi-dimensional fault precursor features. The generation module is used to input the multi-dimensional fault precursor features into a pre-constructed adaptive early warning model to obtain fault prediction results for the distribution network; and to generate early warning information based on the fault prediction results and dynamic early warning thresholds. The push module is used to push the warning information to multiple staff members based on preset rules; The process involves extracting and fusing the temporal and spatial features of the multi-source electrical data using a hierarchical strategy to obtain multi-dimensional fault precursor features, including: Based on a preset duration, the time-series electrical data in the multi-source electrical data is segmented to obtain segmented multi-source electrical data. Linear regression fitting was performed on the segmented multi-source electrical data to obtain trend features that reflect the data change trend; The difference between adjacent data points is calculated for the segmented multi-source electrical data and compared with a preset difference threshold to obtain the adjacent abrupt change characteristics that reflect the difference between adjacent data. By matching the segmented multi-source electrical data with the daily load data of the distribution network using the spectrum analysis method, periodic features reflecting the periodicity of the data are obtained; the trend features, the adjacent abrupt change features, and the periodic features constitute the time dimension features; Based on the distribution network topology data in the multi-source electrical data, an equipment association map is constructed; based on the equipment association map, data coupling characteristics of different nodes on the same line, cross-influence characteristics of adjacent lines, and linkage characteristics between main equipment and auxiliary equipment are extracted to form spatial dimension characteristics; The extracted temporal and spatial features are fused to form multi-dimensional fault precursor features.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images