Distribution network equipment detection method and device, terminal equipment and storage medium

Abstract

The invention discloses distribution network equipment detection, and belongs to the field of power systems, and the method comprises the steps: obtaining micrometeorological monitoring data and vulnerability index data of to-be-detected distribution network equipment in real time; processing the micrometeorological monitoring data and the vulnerability index data through a space-time interpolation algorithm to obtain a space-time standardized data set; inputting the time-space standardized data set into a random forest algorithm, and calculating to obtain a dynamic association feature between the micrometeorological monitoring data and the vulnerability index data; calculating to obtain a risk assessment result according to the dynamic association characteristics; when the risk assessment result is judged to be high risk, processing the space-time standardized data set through a long short-term memory network to generate a fault prediction result; wherein the fault prediction result comprises a fault probability and a possible fault position; and the to-be-detected distribution network equipment is detected according to the fault prediction result, so that the problem of low detection precision of the distribution network equipment in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of power systems, and more particularly to a method, apparatus, terminal equipment, and storage medium for detecting distribution network equipment. Background Technology

[0002] The safe operation of distribution network equipment is directly related to the stability of the power system and the power supply security for countless households. Especially against the backdrop of frequent extreme weather events, the impact of micrometeorological factors such as localized strong winds, lightning, or high temperatures on distribution network equipment is becoming increasingly prominent. Researching how to combine micrometeorological forecasting with equipment status analysis to predict faults in advance and achieve intelligent monitoring is not only crucial for improving power grid reliability but also for socio-economic stability. Existing technologies largely rely on single weather forecasts or equipment status monitoring, lacking dynamic correlation analysis between micrometeorology and equipment vulnerability. This makes it difficult to quickly generate targeted emergency strategies under complex and changing weather conditions, leading to delayed fault response or inefficient resource allocation. Furthermore, existing methods often overlook the spatiotemporal heterogeneity of micrometeorological monitoring data when addressing the correlation between micrometeorology and distribution network equipment faults.

[0003] Micrometeorological monitoring data, due to its localized and instantaneous nature, changes frequently and is difficult to capture accurately. Meanwhile, power distribution network equipment is widely distributed and operates in complex environments, making it difficult for traditional methods to effectively combine these dynamic changes with equipment vulnerability analysis. For example, a short-term heavy rainfall in a certain area may cause a line short circuit, but existing systems cannot quickly determine which equipment is most vulnerable based on real-time micrometeorological changes and equipment aging. This spatiotemporal heterogeneity leads to insufficient accuracy in fault prediction, thus affecting the efficiency of emergency plan generation. The core technical challenge lies in how to efficiently integrate micrometeorological monitoring data with equipment vulnerability analysis results to support the generation of dynamic emergency plans.

[0004] The rapid changes in micrometeorological monitoring data require the system to have real-time processing capabilities. However, the processing and analysis of large amounts of data need to be completed within a short period of time; otherwise, the optimal intervention opportunity may be missed. Furthermore, equipment vulnerability analysis involves multiple factors, such as material aging and operational load. Another key challenge is how to correlate and model these factors with micrometeorological monitoring data and generate targeted contingency plans in a short time. Insufficient correlation between the two can lead to a lack of specificity in the contingency plans. For example, during thunderstorms, the system may not be able to accurately determine which lines are more prone to failure due to aging, thus failing to allocate repair resources appropriately.

[0005] As can be seen from the above, existing power distribution equipment detection methods often neglect the spatiotemporal heterogeneity of micro-meteorological monitoring data when dealing with the correlation between micro-meteorological conditions and power distribution equipment faults, resulting in low detection accuracy of power distribution equipment. Summary of the Invention

[0006] This invention provides a method, apparatus, terminal equipment, and storage medium for detecting distribution network equipment, which can solve the problem of low detection accuracy of distribution network equipment in the prior art.

[0007] The method for detecting distribution network equipment provided by this invention includes: Real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested; Spatiotemporally standardized datasets are obtained by processing micrometeorological monitoring data and vulnerability index data through spatiotemporal interpolation algorithms; The spatiotemporally standardized dataset is input into the random forest algorithm to calculate the dynamic correlation characteristics between micrometeorological monitoring data and vulnerability index data; The risk assessment results are calculated based on dynamic correlation characteristics; When the risk assessment result is determined to be high risk, the spatiotemporally standardized dataset is processed by a long short-term memory network to generate a fault prediction result; the fault prediction result includes the fault probability and the possible fault location. The distribution network equipment to be tested is tested based on the fault prediction results.

[0008] Furthermore, the spatiotemporally normalized dataset is input into the random forest algorithm to calculate the dynamic correlation features between micrometeorological monitoring data and vulnerability index data, including: The spatiotemporally standardized dataset is input into the random forest algorithm so that the random forest algorithm can perform decision tree ensemble on the spatiotemporal variables and meteorological variables in the spatiotemporally standardized dataset to obtain the dynamic correlation characteristics between micro-meteorological monitoring data and vulnerability index data.

[0009] Furthermore, the risk assessment results are calculated based on dynamic correlation characteristics, including: Calculate the association strength of dynamic association features; The correlation strength is compared with a first preset threshold, and the potential failure risk level is generated based on the comparison result. Generate maintenance priorities based on the level of potential failure risk; The risk assessment results were calculated based on maintenance priorities and micrometeorological monitoring data.

[0010] Furthermore, when the risk assessment result indicates a high risk, a spatiotemporally standardized dataset is processed using a Long Short-Term Memory (LSTM) network to generate fault prediction results, including: When the risk assessment result is greater than the third preset threshold, it is judged as high risk; The spatiotemporal normalized dataset is processed using a long short-term memory network to generate normalized sequences; Calculate the failure probability based on the standardized sequence; Fault prediction results are generated based on the fault probability and location coordinate data in the spatiotemporally standardized dataset.

[0011] Furthermore, the correlation strength is compared with a first preset threshold, and a potential failure risk level is generated based on the comparison result, including: When the correlation strength is greater than the first preset threshold, it is determined to be at the first-level risk level; When the correlation strength is greater than the second preset threshold and less than or equal to the first preset threshold, it is determined to be at the level of Level 2 risk. When the correlation strength is less than or equal to the second preset threshold, it is judged as a level three risk level; Potential failure risk levels are generated based on Level 1, Level 2, and Level 3 risk levels.

[0012] Furthermore, real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment under test, including: Micro-meteorological monitoring data is acquired in real time through a micro-meteorological sensor network; the micro-meteorological monitoring data includes wind speed data, lightning intensity data, temperature data, humidity data, and rainfall data; Vulnerability index data is acquired in real time through equipment status sensors; the vulnerability index data includes: equipment aging index data, equipment load index data, equipment operating status index data, equipment mechanical performance index data, equipment insulation performance index data, and equipment environmental adaptability index data.

[0013] Furthermore, by processing micrometeorological monitoring data and vulnerability index data through spatiotemporal interpolation algorithms, a spatiotemporally standardized dataset is obtained, including: Based on micrometeorological monitoring data and vulnerability index data, a comprehensive vulnerability index set was calculated; The geographic location and timestamp of the comprehensive vulnerability index are added according to the spatiotemporal annotation method to generate the annotated comprehensive vulnerability index set; wherein, the geographic location is generated from the location of the device status sensor, and the timestamp is generated from the collection time; The annotation-based comprehensive vulnerability index set is adjusted through a calibration mechanism to generate the original spatiotemporally annotated dataset. After cleaning the original spatiotemporally labeled dataset, spatiotemporally heterogeneous data is processed using a spatiotemporally interpolation algorithm to obtain a spatiotemporally standardized dataset.

[0014] Another embodiment of the present invention provides a power distribution network equipment detection device, comprising: an acquisition module, a first processing module, a first calculation module, a second calculation module, a second processing module, and a detection module; The acquisition module is used to acquire real-time micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested; The first processing module is used to process micro-meteorological monitoring data and vulnerability index data through spatiotemporal interpolation algorithms to obtain a spatiotemporally standardized dataset; The first calculation module is used to input the spatiotemporally standardized dataset into the random forest algorithm to calculate the dynamic correlation characteristics between micro-meteorological monitoring data and vulnerability index data; The second calculation module is used to calculate the risk assessment results based on dynamic correlation characteristics; The second processing module is used to process the spatiotemporally standardized dataset through a long short-term memory network to generate fault prediction results when the risk assessment result determines that the risk is high. The fault prediction results include the fault probability and the possible fault location. The detection module is used to detect the distribution network equipment to be tested based on the fault prediction results.

[0015] Another embodiment of the present invention provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps of the distribution network equipment detection method provided by the present invention.

[0016] Another embodiment of the present invention provides a computer-readable storage medium item, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to perform the steps of the distribution network equipment detection method provided by the present invention.

[0017] The following benefits can be obtained by implementing the present invention: The technical solution of this invention processes micro-meteorological monitoring data and vulnerability index data through a spatiotemporal interpolation algorithm to obtain a spatiotemporally standardized dataset. This standardized dataset is then input into a random forest algorithm to calculate the dynamic correlation characteristics between the micro-meteorological monitoring data and the vulnerability index data. This effectively improves the correlation between external micro-meteorological factors and the internal vulnerability of equipment, solving the technical problem of insufficient correlation between the two and providing more accurate and comprehensive data support for equipment fault determination. A risk assessment result is calculated based on the dynamic correlation characteristics. When the risk assessment result indicates a high risk, the spatiotemporally standardized dataset is processed through a long short-term memory network to generate a fault prediction result. This fault prediction result includes the fault probability and possible fault location. Through the dual determination of the risk assessment result and the fault prediction result, accurate determination of equipment faults can be achieved, thereby effectively improving the detection accuracy of the distribution network equipment under test.

[0018] Meanwhile, the technical solution of this invention can prevent serious accidents such as pole collapse, wire breakage, and insulation breakdown by providing accurate early warning of extreme micro-meteorological events such as strong winds and lightning, which greatly improves the safety of the power grid itself and public safety.

[0019] Furthermore, the technical solution of this invention avoids the huge economic losses to users and society caused by power outages through preventative maintenance. It transforms general inspections into precise maintenance, focusing limited human and material resources on essential tasks, reducing ineffective and inefficient work, and significantly saving maintenance costs. Timely intervention on high-risk equipment prevents it from operating with defects and eventually failing, thereby extending the equipment's lifespan. By preventing faults, the average power outage time per household will be significantly reduced, power supply reliability will be significantly improved, and the user experience will be fundamentally enhanced. Attached Figure Description

[0020] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for detecting distribution network equipment according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a power distribution equipment detection device provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0024] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0027] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0028] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0029] See Figure 1 To address the problem of low detection accuracy of distribution network equipment in existing technologies, an embodiment of the present invention provides a method for detecting distribution network equipment, comprising: S101: Real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested.

[0030] In this embodiment, real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment under test includes: Micro-meteorological monitoring data is acquired in real time through a micro-meteorological sensor network; the micro-meteorological monitoring data includes wind speed data, lightning intensity data, temperature data, humidity data, and rainfall data; Vulnerability index data is acquired in real time through equipment status sensors; the vulnerability index data includes: equipment aging index data, equipment load index data, equipment operating status index data, equipment mechanical performance index data, equipment insulation performance index data, and equipment environmental adaptability index data.

[0031] Furthermore, the micro-meteorological sensor network comprises multiple sensor nodes distributed near power transmission lines. These nodes are equipped with anemometers, lightning detectors, and temperature sensors to monitor micro-meteorological phenomena such as strong winds, lightning, and high temperatures. For example, along a high-voltage transmission line, a sensor node can be installed at regular intervals. These nodes transmit data to a central processing system via wireless communication modules, thereby achieving continuous monitoring of local micro-meteorological conditions. Further, equipment status sensors are used to collect data on equipment vulnerability indicators. Equipment status sensors can be installed on power equipment such as conductors, insulators, and towers to monitor aging and load conditions.

[0032] Furthermore, equipment aging indicators include: conductor aging degree: detecting corrosion or fatigue on the conductor surface using vibration sensors; insulation aging degree: detecting the aging of insulation materials using insulation performance testing equipment; equipment operating years: the time the equipment has been in use since installation, used to assess its aging degree; and material aging indicators: such as the degree of corrosion of metal parts and the decrease in insulation resistance of insulation materials.

[0033] Equipment load indicators include: Current load: real-time current intensity of the equipment measured by a current sensor; Voltage fluctuation: voltage change of the equipment monitored by a voltage sensor; Load factor: the ratio of the actual load of the equipment to the rated load, used to assess whether the equipment is in an overload state; Power factor: the power factor of the equipment during operation, reflecting the load characteristics and efficiency of the equipment.

[0034] Equipment operating status indicators include: Equipment operating deviation value: the deviation between the actual operating parameters of the equipment and the rated parameters; Equipment fault history record: the number and type of past faults of the equipment, used to assess its reliability; Equipment maintenance record: the number of maintenance and maintenance content of the equipment, reflecting the maintenance status of the equipment; Equipment operating temperature: the operating temperature of the equipment is monitored by a temperature sensor, and excessively high temperatures may cause equipment failure.

[0035] The mechanical performance indicators of the equipment include: conductor tension: the actual tension of the conductor during operation; excessive tension may lead to conductor breakage; equipment structural stability: such as the tilt angle of the tower and the settlement of the foundation; mechanical stress: the mechanical stress borne by the equipment, such as the impact of wind pressure and snow pressure on the equipment.

[0036] The insulation performance indicators of the equipment include: insulation resistance: the insulation resistance of the equipment is measured by an insulation resistance tester; insulation breakdown probability: the breakdown probability is assessed based on the aging degree of the insulation material and the operating environment; leakage current: the leakage current of the insulation part of the equipment, reflecting the decline in insulation performance.

[0037] Equipment environmental adaptability indicators include: equipment wind resistance: the equipment's ability to withstand strong winds, such as wind pressure threshold; equipment corrosion resistance: the equipment's ability to resist corrosion in humid or salt spray environments; and equipment lightning strike resistance: the equipment's ability to withstand lightning strikes under lightning conditions, such as grounding resistance.

[0038] Specifically, aging indicators can be detected by vibration sensors to check for corrosion or fatigue on the conductor surface, while load indicators are measured by current sensors to check real-time current intensity. These vulnerability indicators are collected in real time by equipment condition sensors and combined with microclimate monitoring data to assess the potential failure risk of the equipment.

[0039] S102: Process micro-meteorological monitoring data and vulnerability index data through spatiotemporal interpolation algorithms to obtain a spatiotemporally standardized dataset.

[0040] In this embodiment, micrometeorological monitoring data and vulnerability index data are processed using a spatiotemporal interpolation algorithm to obtain a spatiotemporally standardized dataset, including: Based on micrometeorological monitoring data and vulnerability index data, a comprehensive vulnerability index set was calculated; The geographic location and timestamp of the comprehensive vulnerability index are added according to the spatiotemporal annotation method to generate the annotated comprehensive vulnerability index set; wherein, the geographic location is generated from the location of the device status sensor, and the timestamp is generated from the collection time; The annotation-based comprehensive vulnerability index set is adjusted through a calibration mechanism to generate the original spatiotemporally annotated dataset. After cleaning the original spatiotemporally labeled dataset, spatiotemporally heterogeneous data is processed using a spatiotemporally interpolation algorithm to obtain a spatiotemporally standardized dataset.

[0041] In one possible implementation, each data point is appended with geographic coordinates and a timestamp, for example, using a GPS module to record sensor locations and labeling the acquisition time with millisecond-level precision. This spatiotemporally labeled raw dataset supports subsequent power equipment risk assessments, ensuring data accuracy in both time and space. It should be noted that this acquisition method can be flexibly applied in different scenarios within the power transmission field. For example, in mountainous transmission lines, the sensor network can be augmented with slope sensors to monitor micro-meteorological changes influenced by terrain, while in plains areas, the density of lightning detectors is emphasized. In this way, the system achieves comprehensive monitoring of various micro-meteorological conditions and equipment status. Understandably, this data acquisition process helps improve the operational safety of power equipment, such as timely identification of conductor overload risks under strong wind conditions, thus providing a data foundation for preventative maintenance.

[0042] Specifically, noise removal and outlier removal are performed on the original spatiotemporally labeled dataset to generate the original spatiotemporally labeled dataset; spatiotemporal continuity is calculated based on micro-meteorological monitoring data; and heterogeneous data is processed at the interpolation points of the micro-meteorological monitoring data based on spatiotemporal continuity and spatiotemporal interpolation algorithm to obtain homogenized data. Based on homogenized data, a spatiotemporally standardized dataset.

[0043] Specifically, microclimate monitoring data typically originates from sensor networks, such as temperature, humidity, or wind speed measurement devices. This data may be affected by noise due to equipment malfunctions or environmental interference, such as random fluctuations or systematic errors. Data cleaning methods include statistical filtering, such as using median filters to smooth sequential data and replacing outliers by calculating local medians, thereby preserving the true meteorological trends.

[0044] It should be noted that outlier identification can be based on threshold judgment. For example, a temperature change rate exceeding a preset limit is considered an anomaly and is subsequently removed or corrected. This cleaning process ensures the reliability of the dataset, providing a high-quality foundation for subsequent processing. Furthermore, spatiotemporal interpolation algorithms are used to address the spatiotemporal heterogeneity of micrometeorological data. Spatiotemporal heterogeneity refers to the uneven distribution of data in both time series and spatial dimensions. For example, in urban meteorological monitoring, wind speed data may be missing or discontinuous at different monitoring points due to terrain differences. Spatiotemporal interpolation algorithms, such as the inverse distance weighting method, can be used to estimate missing values. The specific process involves calculating the spatiotemporal distance between known points and target points, and assigning weights based on the distance (e.g., the weight is inversely proportional to the distance), thereby generating the interpolation result. This method considers the correlation in the time dimension, such as the influence of data from adjacent hours on the current estimate, ensuring that the interpolated data is spatiotemporally continuous.

[0045] For example, after cleaning, the data is directly input into the interpolation module, and the output is then standardized to form the final dataset, which improves the accuracy of the data. For example, in weather forecasting applications, homogenizing the dataset reduces prediction errors and supports the construction of more reliable microclimate models.

[0046] For example, in an agricultural meteorological monitoring scenario, assuming the original dataset comes from field sensors, noise may originate from sudden spikes in soil moisture readings. During data cleaning, the Z-score method is first applied to detect anomalies, i.e., calculating the deviation of each data point from the mean; if the deviation exceeds three times the standard deviation, it is marked as an anomaly and removed. Subsequently, spatiotemporal interpolation is used to process missing rainfall data, for example, using Kriging interpolation. This algorithm is based on a variogram model of spatiotemporal covariance, and the process includes fitting a variogram to quantify heterogeneity, and then filling in the blanks using minimum variance estimation. This implementation demonstrates the effectiveness of the algorithm in handling localized weather variability, generating a more uniform data distribution.

[0047] By interpolating the time series data, data from different monitoring points are aligned at the same timestamp. Simultaneously, spatial data normalization is applied to adjust the numerical range to between 0 and 1, eliminating dimensional differences. This standardization ensures consistency of input for subsequent algorithms and is applicable to various transmission line scenarios in the field of power equipment risk assessment, such as high-voltage lines in mountainous areas or low-voltage networks in plains. Further, the obtained spatiotemporally standardized dataset is input into a random forest algorithm for feature extraction. The random forest algorithm is an ensemble learning method consisting of multiple decision trees, each built based on a random subset of the dataset, with feature importance determined through a voting mechanism. The specific process includes: first, randomly sampling multiple training subsets from the standardized dataset; then, in each decision tree, using the Gini index or information gain as a splitting criterion, selecting the feature that best distinguishes between micrometeorological factors and equipment vulnerability, such as the rate of change of wind speed or the elastic modulus of conductor material; finally, summing the feature importance scores of all trees and extracting the top-ranked feature vectors. In the power sector, this extraction method can highlight key variables, such as the impact of wind pressure on conductors during strong winds, thus providing refined input for subsequent modeling.

[0048] In one possible implementation, the number of trees can be adjusted from 100 to 500 to balance computational efficiency and accuracy.

[0049] Furthermore, in the Random Forest algorithm, the training subset is generated through bootstrap sampling. The specific steps are as follows: Bootstrapping: Randomly select samples from the original standardized dataset, allowing for repetition. The number of samples selected each time is the same as the original dataset, but some samples may be selected repeatedly, while others may not be selected. The resulting subset is called the training subset.

[0050] Generate multiple training subsets: Typically, multiple training subsets (e.g., 100 or more) are generated, each subset being used to train a decision tree.

[0051] In a random forest, each decision tree, when splitting a node, randomly selects a subset of features (usually the square root of the total number of features) as candidate features. Then, it selects the optimal feature from these candidate features for splitting. The specific steps are as follows: Randomly select candidate features: At each split node, randomly select a subset of features from all features as candidate features.

[0052] Selecting the optimal feature: Using the Gini index or information gain as the splitting criterion, select the optimal feature from the candidate features for splitting.

[0053] Repeat the above process until the decision tree reaches the preset depth or other stopping conditions are met.

[0054] Feature completeness of training subsets: Each training subset is randomly drawn from the original dataset, so each subset contains all features (including all vulnerability features). This means that each decision tree has access to all features during training.

[0055] Random selection of candidate features: At each split node, although the random forest randomly selects a subset of features as candidate features, this does not affect the fact that each training subset contains all features.

[0056] Feature importance calculation: During training, the importance of each feature is calculated for each decision tree. Feature importance is usually measured by calculating the reduction in impurity (such as the Gini index or information gain) of that feature across all split nodes.

[0057] Summarizing Feature Importance: The Random Forest algorithm sums the feature importance scores of all decision trees to obtain the average importance of each feature. The specific steps are as follows: For each feature, calculate its importance score across all decision trees. Summarize these scores to obtain the average importance of each feature.

[0058] The impact of randomly selected candidate features: Although candidate features are randomly selected at each split node, because a random forest contains multiple decision trees, each tree independently selects candidate features and calculates feature importance. Therefore, the aggregated feature importance score can better reflect the overall importance of each feature.

[0059] Error issue: Theoretically, certain vulnerability features may be selected less frequently in some decision trees, but this will not lead to significant error. Because random forests reduce variance through the ensemble of multiple trees, the aggregated feature importance scores can balance this randomness, thus providing a more accurate assessment of feature importance.

[0060] For example, the following data will be used to illustrate this: Wind speed (W); Lightning intensity (L); Temperature (T); Humidity (H); Rainfall (R); Equipment aging index (A); Equipment load index (Ld); Generate training subset: 100 training subsets are generated by randomly selecting samples from the original dataset. Each subset contains all features.

[0061] Training the decision tree: For each training subset, train a decision tree. At each split node, randomly select a subset of features (e.g., 3) as candidate features. Use the Gini index or information gain to select the optimal feature for splitting.

[0062] Calculate feature importance: For each decision tree, calculate the importance of each feature. Summarize the feature importance scores of all decision trees to obtain the average importance of each feature.

[0063] Assume the aggregated feature importance scores are as follows: Wind speed (W): 0.25; Lightning intensity (L): 0.20; Temperature (T): 0.15; Humidity (H): 0.10; Rainfall (R): 0.05; Equipment aging index (A): 0.15; Equipment load index (Ld): 0.10; By using this aggregation method, random forests can better reflect the overall importance of each feature, and even if some features are selected less often in some decision trees, it will not lead to significant errors.

[0064] In one specific embodiment, a comprehensive vulnerability index is calculated from micrometeorological monitoring data and vulnerability index data using a weighted summation method, specifically as follows: W(t): Wind speed data (unit: meters per second); L(t): Lightning intensity data (unit: joules / square meter) T(t): Temperature data (unit: degrees Celsius); A(t): Equipment aging index (dimensionless, e.g., a value between 0 and 1); Ld(t): Equipment load index (dimensionless, e.g., a value between 0 and 1); A comprehensive vulnerability index V(t) can be defined as follows: V(t)=α*W(t)+β*L(t)+γ*T(t)+δ*A(t)+ϵ*L d (t); Where α, β, γ, δ, and ϵ are weighting coefficients, which are adjusted according to the actual application scenario.

[0065] The choice of weighting coefficients can be determined based on the importance and relevance of the data. For example, if wind speed has a significant impact on equipment, the value of α can be increased. Weighting coefficients can be determined using the following methods: Expert experience method: Manually adjust the weighting coefficients based on the experience of domain experts. Data-driven method: Automatically adjust the weighting coefficients by training a machine learning model using historical data. Normalization method: Normalize all weighting coefficients so that their sum is 1. For example, by using min-max normalization, all variables can be scaled to the range of 0 to 1 to ensure the comparability of different meteorological parameters such as temperature and wind speed.

[0066] In one possible implementation, this step also includes spatiotemporal gridding, which maps the data to a uniform grid, such as one grid point per square kilometer, and applies averaging to eliminate residual heterogeneity.

[0067] Understandably, the process of determining such standardized datasets emphasizes homogenization, that is, making the data consistent across spatiotemporal scales through the steps described above.

[0068] For example, in the field of environmental monitoring, the processed dataset can be used to construct continuous microclimate maps, avoiding analytical biases caused by blank areas in the original data. In another embodiment, for mountain micrometeorological data, noise removal can be combined with machine learning methods, such as the isolated forest algorithm to detect anomalies. This algorithm isolates outlier points by constructing random trees, involving multiple random partitioning of data subsets until anomalies are isolated, thus efficiently processing high-dimensional meteorological data. Spatiotemporal interpolation employs a spatiotemporal kriging variant, considering altitude as an auxiliary variable. The calculation process includes estimating the spatiotemporal variability function and solving a system of linear equations to generate accurate interpolations. This approach extends the applicability of the method, maintaining data uniformity in complex terrain. Furthermore, the logic of the entire process transitions from data cleaning to interpolation and standardization, ensuring the integrity and usability of the micrometeorological data.

[0069] Furthermore, the purpose of the spatiotemporal annotation method is to add a geographic location and timestamp to each data point for subsequent analysis. The specific steps are as follows: Step 1: Data Collection Data is collected in real time through a micro-weather sensor network and equipment status sensors. The geographic coordinates (longitude and latitude) of each sensor location are recorded via a GPS module. Each data point is accompanied by a timestamp (accurate to the second).

[0070] Step 2: Spatiotemporal Labeling For each data point Add geographic coordinates And timestamp ti.

[0071] The format of the labeled data points is as follows: .

[0072] Step 3: Deviation Calibration If the labeled dataset has biases (e.g., sensor errors or data anomalies), the indicator values ​​are adjusted through a calibration mechanism. The calibration mechanism can be based on preset standard values ​​or historical data.

[0073] For example, wind speed sensor: W(t) = 10 m / s; Lightning sensor: L(t) = 50 joules / square meter; Temperature sensor: T(t) = 35 degrees Celsius; Equipment aging index: A(t) = 0.8; Equipment load index: Ld(t) = 0.7; Weighting coefficients: α = 0.3; β=0.2; γ=0.2; δ=0.2; ϵ=0.1; Comprehensive vulnerability index V(t): V(t)=0.3*10+0.2*50+0.2*35+0.2*0.8+0.1*0.7=3+10+7+0.16+0.07=20.23; Geographic coordinates: (lat=34.0522, lon=−118.2437); Timestamp: t=2025−11−0310:00:00; The labeled data points are: (34.0522,−118.2437,2025−11−0310:00:00,10,50,35,0.8,0.7).

[0074] S103: Input the spatiotemporally standardized dataset into the random forest algorithm to calculate the dynamic correlation characteristics between micrometeorological monitoring data and vulnerability index data.

[0075] In this embodiment, the spatiotemporally normalized dataset is input into the random forest algorithm to calculate the dynamic correlation features between micrometeorological monitoring data and vulnerability index data, including: The spatiotemporally standardized dataset is input into the random forest algorithm so that the random forest algorithm can perform decision tree ensemble on the spatiotemporal variables and meteorological variables in the spatiotemporally standardized dataset to obtain the dynamic correlation characteristics between micro-meteorological monitoring data and vulnerability index data.

[0076] The spatiotemporal standardized dataset originates from the monitoring system of power transmission lines, including time-series data of micrometeorological monitoring elements such as wind speed, temperature, and humidity, as well as spatial distribution data of equipment vulnerability indicators such as conductor tension and insulation aging. The spatiotemporal standardization process aims to unify the data format and scale.

[0077] It should be noted that the feature extraction process of the Random Forest algorithm places particular emphasis on the processing of high-dimensional data.

[0078] For example, in a power transmission system, the dataset may contain hundreds of variables, such as real-time wind direction, equipment installation height, and historical fault records. By using bootstrapping sampling and random feature selection in the algorithm, overfitting is avoided, and the average reduction in impurity value for each feature is calculated as a quantitative indicator of its importance. In practical applications, this method can identify the dominant role of micrometeorological factors, such as sudden gusts, in equipment vulnerability. Compared to a single decision tree, it is more robust, thus improving the reliability of feature extraction. Further processing is performed based on the extracted features, modeling the dynamic correlation between micrometeorological factors and equipment vulnerability.

[0079] Specifically, dynamic correlation modeling employs a time-series correlation analysis framework, such as using vector autoregression models or graph neural networks to capture time-varying relationships between variables. In the power equipment scenario, micrometeorological factors include local wind loads and temperature fluctuations, while equipment vulnerability manifests as conductor fatigue accumulation or insulation breakdown probability. The modeling process first constructs a correlation matrix, mapping eigenvectors to a time window; then, it calculates dynamic correlation coefficients, such as updating Pearson correlation over a sliding window to quantify how increased wind speed leads to intensified conductor vibration; next, it incorporates equipment vulnerability parameters, such as material stress thresholds, to simulate the nonlinear evolution of the correlation. For example...

[0080] In one embodiment, for high-voltage transmission lines, the model considers a wind load factor, defined as the product of the square of the wind speed and the air density, to calculate the wind pressure value the conductor can withstand: the wind pressure value equals the wind load factor multiplied by the projected area of ​​the conductor, then divided by the conductor diameter, thereby assessing the vibration-induced vulnerability increase. This modeling is applicable to different power line environments, such as coastal or inland areas, demonstrating the flexibility of the technical solution.

[0081] Preferably, in dynamic correlation modeling, Bayesian networks can be introduced to handle uncertainty. The specific process involves defining nodes as micrometeorological factors and equipment vulnerability indicators, with edges representing conditional probability dependencies; and updating the posterior probability distribution through training with historical data.

[0082] For example, when wind speed exceeds a threshold, the equipment vulnerability probability increases from 0.2 to 0.6. This method can dynamically capture correlation changes and effectively predict the impact of sudden events such as storms on power lines in power risk assessment.

[0083] Understandably, the process of determining the level of potential failure risk is based on the aforementioned modeling results.

[0084] Specifically, the risk assessment model quantifies the risk level by integrating dynamically correlated outputs and using a multilayer perceptron or threshold classifier.

[0085] For example, the correlation strength can be mapped to a risk score, with 0 to 1 representing low to high; if the correlation coefficient exceeds 0.7, it is judged as high risk, triggering an early warning. In the power sector, this model is applicable to substation equipment to assess the correlation between temperature rise and insulation vulnerability, leading to potential short-circuit faults.

[0086] In one embodiment, for transmission lines in mountainous areas, the acquired dataset includes satellite meteorological data and ground sensor readings. After standardization, features such as wind pressure gradients are extracted using a random forest model. Then, dynamic correlations are modeled to calculate the stress distribution of the conductors under wind load. Finally, the risk level is determined; if it exceeds a safety threshold, the model outputs a high-risk warning. This implementation demonstrates the application of the solution in complex terrain. Furthermore, another implementation targets plains power grids, where the dataset focuses on the correlation between humidity and corrosion vulnerability. Feature extraction emphasizes the rate of humidity change, modeling uses a sliding window to analyze dynamic relationships, risk assessment is based on probability thresholds, and preventative maintenance recommendations are provided.

[0087] For example, this risk assessment model can achieve real-time monitoring in actual deployment. Through the logical flow of the above steps, a closed chain is formed from data acquisition to risk output, improving the accuracy of fault prediction in power equipment management. In the effect description, this implementation method ensures accurate judgment of risks induced by microclimates through detailed feature extraction and dynamic modeling, covering multiple scenarios in the same field, such as lines of different voltage levels, demonstrating the versatility of the technical solution.

[0088] S104: The risk assessment result is calculated based on the dynamic correlation characteristics.

[0089] In this embodiment, the risk assessment result is calculated based on dynamic correlation characteristics, including: Calculate the association strength of dynamic association features; The correlation strength is compared with a first preset threshold, and the potential failure risk level is generated based on the comparison result. Generate maintenance priorities based on the level of potential failure risk; The risk assessment results were calculated based on maintenance priorities and micrometeorological monitoring data.

[0090] In one specific embodiment, the correlation strength is compared with a first preset threshold, and a potential fault risk level is generated based on the comparison result, including: When the correlation strength is greater than the first preset threshold, it is determined to be at the first-level risk level; When the correlation strength is greater than the second preset threshold and less than or equal to the first preset threshold, it is determined to be at the level of Level 2 risk. When the correlation strength is less than or equal to the second preset threshold, it is judged as a level three risk level; Potential failure risk levels are generated based on Level 1, Level 2, and Level 3 risk levels.

[0091] For example, the first preset threshold is 0.7. When the correlation strength of the dynamic correlation feature is greater than 0.7, it is determined to be at level 1 risk (high risk). The correlation strength is usually calculated using the random forest algorithm, ranging from 0 to 1, representing the correlation strength between micrometeorological monitoring data and vulnerability indicator data. The second preset threshold is 0.5. When the correlation strength is greater than 0.5 but less than or equal to 0.7, it is determined to be at level 2 risk (medium risk); when the correlation strength is less than or equal to 0.5, it is determined to be at level 3 risk (low risk).

[0092] In this embodiment, the maintenance priority is generated based on the potential failure risk level as follows: Level 1 Risk: Equipment failure may lead to major safety accidents or large-scale power outages, requiring immediate maintenance, and is classified as a high priority.

[0093] Level 2 Risk: Equipment failure may cause partial power outages or affect equipment lifespan, requiring maintenance in the near future, and is classified as medium priority.

[0094] Level 3 risk: The risk of equipment failure is low and can be addressed during planned maintenance; therefore, it is classified as a low priority.

[0095] Furthermore, based on maintenance priorities and micrometeorological monitoring data, the risk assessment results were calculated as follows: The risk assessment results are calculated by a risk assessment model, which can be constructed by weighting and summing maintenance priorities and micrometeorological monitoring data. Specifically: Maintenance Priority (P): A level (1, 2, 3) based on the importance of the equipment and the risk of failure.

[0096] Micrometeorological monitoring data: wind speed (W); lightning intensity (L); temperature (T); humidity (H); rainfall (R).

[0097] The weighted summation process is as follows: Risk(t)=α*P+β*W(t)+γ*L(t)+δ*T(t)+ϵ*H(t)+ζ*R(t) Where α, β, γ, δ, ϵ, and ζ are weighting coefficients, which are adjusted according to the actual application scenario. Risk(t) is a risk assessment model used to determine the potential failure risk level of equipment.

[0098] For example: Maintenance priority (P): 2 (medium priority) Micro-meteorological monitoring data: Wind speed (W): 10 m / s; Lightning intensity (L): 50 joules / square meter; Temperature (T): 35 degrees Celsius; Humidity (H): 80%; Rainfall (R): 10 mm; Weighting coefficients: α = 0.3; β=0.2; γ=0.2; δ=0.1; ϵ=0.1; ζ=0.1; Calculation process Calculate the weighted value for each factor: Maintenance priority weighting value: 0.3 × 2 = 0.6; Wind speed weighted value: 0.2 × 10 = 2.0; Lightning intensity weighting: 0.2 × 50 = 10.0; Temperature weighted value: 0.1 × 35 = 3.5; Humidity weighted value: 0.1 × 80 = 8.0; Rainfall weighted value: 0.1 × 10 = 1.0; The risk assessment results are calculated using the Risk(t) risk assessment model: Risk(t)=0.6+2.0+10.0+3.5+8.0+1.0=25.1; Based on the risk assessment results, appropriate thresholds can be set to determine the potential failure risk level of the equipment. For example: Low risk: Risk(t) < 10; Medium risk: 10 ≤ Risk(t) < 20; High risk: Risk(t) ≥ 20; In this example, the risk assessment result is 25.1, which falls under the high-risk category. Therefore, the risk assessment model will output a high-risk signal, triggering subsequent fault prediction and emergency plan generation.

[0099] S105: When the risk assessment result is determined to be high risk, the spatiotemporally standardized dataset is processed by a long short-term memory network to generate a fault prediction result; the fault prediction result includes the fault probability and the possible fault location.

[0100] In this embodiment, when the risk assessment result determines the risk to be high, a fault prediction result is generated by processing the spatiotemporally standardized dataset through a long short-term memory network, including: When the risk assessment result is greater than the third preset threshold, it is judged as high risk; The spatiotemporal normalized dataset is processed using a long short-term memory network to generate normalized sequences; Calculate the failure probability based on the standardized sequence; Fault prediction results are generated based on the fault probability and location coordinate data in the spatiotemporally standardized dataset.

[0101] For example, the third preset threshold is 20. When the risk assessment result (calculated by weighted summation, such as Risk(t)=α*P+β*W(t)+γ*L(t)+δ*T(t)+ϵ*H(t)+ζ*R(t)) is greater than 20, it is judged as high risk, triggering the long short-term memory network to predict the fault. Referring to the example in the specification, the risk assessment result is classified as: low risk (<10), medium risk (10≤Risk(t)<20), and high risk (≥20).

[0102] Specifically, in one implementation, when the risk assessment model outputs a high-risk signal (i.e., when the risk assessment result determines it to be high-risk), the system activates a long short-term memory network to process the spatiotemporally standardized dataset to predict the probability and location of the failure in the short term. This processing is based on a pre-built risk assessment model trained on historical data, which is capable of identifying potential risk thresholds.

[0103] For example, a high-risk signal can be defined as a risk score (risk assessment result) calculated by the model exceeding a third preset threshold, such as a probability threshold of 0.8, at which point the subsequent prediction process is triggered. This mechanism ensures that the system responds promptly when anomalies are detected, avoiding unnecessary calculations in low-risk scenarios, thereby improving overall efficiency. Furthermore, the construction of a spatiotemporally standardized dataset is a crucial step. First, spatiotemporal data from the same domain are collected, such as sensor data distributed across different geographical locations in a power system, including timestamps, voltage fluctuations, and location coordinates. These data form raw sequences, which are then standardized.

[0104] Specifically, standardization involves transforming the data into a distribution with a mean of 0 and a standard deviation of 1, using a formula such as (x - μ) / σ, where x is the original value, μ is the mean, and σ is the standard deviation. This process eliminates data scaling differences, ensuring the sequence is suitable for input to a long short-term memory network. This is particularly relevant in power fault prediction scenarios.

[0105] For example, a sequence may contain voltage and current readings over a continuous 24-hour period, which, combined with latitude and longitude information, form a multidimensional spatiotemporal sequence.

[0106] It should be noted that Long Short-Term Memory (LSTM) networks are a variant of recurrent neural networks designed specifically for processing sequential data and capable of capturing long-term dependencies. In this embodiment, the network receives a spatiotemporally normalized dataset as input and manages the information flow through forget gates, input gates, and output gates. The specific process includes: first, the network's hidden layers process data at each time step of the sequence; the forget gate determines whether to retain or discard previous memories; the input gate updates the current information; and the output gate generates the hidden state. This structure is particularly suitable for spatiotemporal data because it can simultaneously consider temporal evolution and spatial correlations. For example, in power systems, the network can learn historical fault patterns, such as the propagation of voltage peaks at specific locations, to predict future trends.

[0107] In one possible implementation, the process of predicting the probability of a fault occurring in the short term is implemented through the output layer of a Long Short-Term Memory (LSTM) network. After processing the sequence, the network outputs a vector representing the probability value at each time step and location. Specifically, the output layer uses a softmax function to transform the hidden states into a probability distribution.

[0108] For example, for a prediction within the next hour, the network generates a probability matrix, where rows correspond to time and columns correspond to location. For instance, if the input sequence shows a persistent voltage anomaly in a certain area, the network might output a probability of 0.75 for a fault at that location within 30 minutes. This prediction is based on learning during the training phase, where the network optimizes parameters using labeled data and minimizes a loss function, such as cross-entropy loss, through backpropagation, thereby improving accuracy. In substation maintenance scenarios within the same domain, this method can help operators allocate resources in advance.

[0109] Preferably, the prediction of fault location incorporates a spatial attention mechanism to enhance the functionality of the Long Short-Term Memory (LSTM) network. When processing spatiotemporal sequences, the network integrates an attention layer to calculate the weights of different locations within the sequence. For example, correlation is evaluated using a dot-product attention formula. This mechanism allows the model to focus on high-risk areas, improving the accuracy of location prediction. In a specific implementation, the attention layer outputs a weighted sequence, which is then input into the LTM unit. For example, in power transmission line fault prediction, if the sequence shows data from multiple nodes, the attention mechanism can highlight the location of segments with high wind loads, predicting a breakage probability of 0.6, while simultaneously providing latitude and longitude coordinates. This extension supports the versatility of the claims and is applicable to power networks of different sizes. For example, in another embodiment, the spatiotemporally normalized dataset can be extended to include environmental factors such as temperature and humidity data. These additional dimensions are normalized after dimensionality reduction using principal component analysis and input into the LTM network. The prediction process is similar, but these factors are incorporated during model training to capture fault modes such as those caused by high temperatures. In the power sector, this implementation is suitable for predicting peak summer periods, ensuring system robustness. Furthermore, after obtaining the fault prediction results, the system can generate a visualization report, including a probability heatmap and location markers. This output, based on the final computation of a Long Short-Term Memory (LSTM) network, aids user decision-making. For example, the report displays a color gradient of high-probability areas, with their locations marked by coordinates. In implementation, the results are thresholded to highlight only fault points with a probability exceeding 0.5, thus achieving efficient response.

[0110] Understandably, the training process of Long Short-Term Memory (LSTM) networks involves batch gradient descent, using historical spatiotemporal datasets as the training set.

[0111] Specifically, the dataset is divided into training, validation, and test sets, and the network is iteratively optimized until the validation loss stabilizes. This training ensures that the model can reliably predict when high-risk signals are triggered. In power fault scenarios, training data can be derived from sensor records over the past year, covering various fault types such as short circuits or overloads.

[0112] In one embodiment, to enhance flexibility, the Long Short-Term Memory (LSTM) network can employ a bidirectional structure, processing both forward and backward information of the sequence. This variant is more accurate in predicting location because it takes contextual dependencies into account. For example, the bidirectional network outputs a combined probability, suitable for predicting fault propagation in complex power grids.

[0113] Specifically, the entire process begins with the risk assessment model outputting a high-risk signal and ends with processing by the Long Short-Term Memory (LSTM) network, forming a closed-loop prediction system. This method demonstrates versatility in the power sector, adapting to different sub-scenarios through various data input variations, such as urban vs. rural power grids, ultimately providing reliable fault early warnings.

[0114] S106: Test the distribution network equipment to be tested based on the fault prediction results.

[0115] In this embodiment, the method further includes using fault prediction results combined with a preset emergency response rule base to generate a sequence of preventive measures for specific equipment and areas, and determining a preliminary draft emergency plan, specifically as follows: Obtain abnormal indicator data from the fault prediction results; among which, abnormal indicator data includes equipment operation deviation values; When abnormal indicator data is determined to exceed the response threshold, a preliminary risk assessment report is generated; the response threshold is preset in the emergency rule base. Based on the preliminary risk assessment report, a risk distribution map is determined by combining the prediction results with specific parameters for each equipment area using correlation logic. The specific parameters cover the load distribution of the area, and the correlation logic is generated by matching prediction deviations with corresponding rules. By using a risk distribution map, the priority order of preventive measures is obtained, and preventive paths are constructed for specific regional equipment to determine a preliminary draft emergency plan. The priority order is sorted according to the risk level, and the preventive paths integrate equipment monitoring sequences.

[0116] Specifically, in one implementation, fault prediction results are obtained through the analysis of equipment operating data. For example, in a power transmission system, data on conductor temperature, wind speed, and load are collected to predict potential faults.

[0117] Specifically, the fault prediction model uses machine learning algorithms to process this data. First, it extracts features such as temperature change rate and wind force influence coefficient. Then, it trains the model based on historical fault datasets to output a fault probability distribution. This prediction provides a foundation for subsequent emergency response, ensuring the targeted nature of measures. Furthermore, the pre-built emergency response rule base is a pre-constructed database containing various standardized rules. For example, rules for high-voltage power line areas include wind pressure threshold checks and isolation operation guidelines. The rule base is designed considering equipment type and regional characteristics. For instance, power lines in mountainous areas are susceptible to strong winds; therefore, a wind load coefficient is defined in the rules. This coefficient is an empirical value calculated based on wind speed and power line diameter, used to assess the wind pressure the power line can withstand. The specific calculation involves multiplying the square of the wind speed by the load coefficient and then combining it with the power line cross-sectional area to obtain the wind pressure value. For example, if the wind speed is 20 meters per second, the load coefficient is 0.5, and the power line diameter is 0.02 meters, the wind pressure value is estimated using a formula as a reference for the power line's safety threshold. This explanation helps to understand how the rule base transforms abstract wind factors into actionable response standards, thereby triggering corresponding rules when a high wind pressure risk is predicted.

[0118] Preferably, when generating a sequence of preventive measures by combining the fault prediction results with the rule base, the predicted fault type is first matched to the corresponding entry in the rule base.

[0119] In one possible implementation, if the prediction indicates a conductor overload probability exceeding 30%, a sequence is extracted from the rule base, such as first reducing the load, then checking connection points, and finally deploying backup lines. This sequence generation emphasizes sequentiality, ensuring that measures are implemented progressively from prevention to response. In the power equipment field, this approach is applicable to substation areas, and the sequence can include real-time monitoring adjustments to reduce the probability of fault occurrence.

[0120] It's important to note that the innovation of generating preventative action sequences lies in its customization for specific equipment and regions. For example, for equipment in coastal areas, the sequence prioritizes corrosion prevention measures, such as coating inspections, while for mountainous areas, it emphasizes wind pressure mitigation steps. This differentiation is achieved through regional tags in a rule base, which are categorized based on geographic data to ensure the sequence's applicability. The detailed process includes: first, inputting the prediction results and equipment identification; second, querying the rule base to filter and match rules; and third, sorting the measures to form a sequence, such as an initial alarm notification, intermediate equipment isolation, and final maintenance scheduling. This step-by-step decomposition makes sequence generation more logical and enables rapid response in business operations, reducing equipment downtime.

[0121] In one embodiment, determining a preliminary draft emergency response plan involves integrating the generated sequence of preventative measures into a document. For example, the draft may include a list of measures, assignment of responsibilities, and a timeline.

[0122] Specifically, based on sequences, draft plans automatically populate templates. For example, a draft plan for conductor faults might list preventative inspections, emergency isolation, and recovery steps. This drafting process ensures the initial completeness of the plan, facilitating subsequent review. For instance, the emergency response rule base can be constructed using a modular structure, with each module corresponding to a fault type; for example, an electrical overload module might contain a subset of rules. Modules are connected via association keys. For example, a wind pressure rule module might link to a mechanical stress module, ensuring comprehensive rule invocation. In power systems, this structure allows for dynamic updates to the rule base; as new fault modes emerge, corresponding rules are added, maintaining the base's timeliness.

[0123] Specifically, obtaining fault prediction results involves data preprocessing steps.

[0124] For example, after filtering out noisy data, feature values ​​such as average wind speed and peak load are calculated. These feature values ​​are input into a prediction model based on a neural network architecture. The training process includes forward propagation and backward weight adjustment to minimize prediction error. This detailed training description highlights the accuracy of the predictions, enabling early identification of risks in operations, such as precursors to conductor breakage, thereby guiding the prioritization of preventative measures. Furthermore, prioritization algorithms can be introduced when generating a sequence of preventative measures.

[0125] For example, based on fault severity ranking measures, high-severity faults, such as short-circuit risks, are isolated first. This algorithm is implemented through weighted scores, calculated based on predicted probabilities and rule-based influence factors, ensuring sequence optimization. In transformer equipment scenarios within the same field, the sequence might include a cooling system check as the first step, demonstrating the versatility of the technology.

[0126] In one embodiment, the determination of the initial draft emergency plan also includes a verification step. For example, simulating the effectiveness of a test sequence. The feasibility of the draft is confirmed by comparing simulation results with historical data. This verification enhances the reliability of the plan and can effectively reduce the accident rate in power maintenance operations.

[0127] Understandably, the entire process, from fault prediction to contingency plan determination, ensures consistent response. For example, prediction results directly influence rule matching, thereby generating targeted sequences and ultimately forming a draft. This seamless integration provides an efficient fault management framework in the industrial equipment field.

[0128] Furthermore, by iteratively matching the preliminary draft emergency plan with real-time updated micrometeorological monitoring data, the sequence of measures is adjusted to adapt to dynamic changes, resulting in a final optimized emergency plan, including: Real-time updated data is obtained from micrometeorological monitoring data, and an iterative matching process is carried out on the draft emergency plan to obtain the adjusted sequence of measures. By adjusting the sequence of measures and integrating dynamic changes to adapt, when the risk level ranking exceeds the fourth preset threshold, a prevention path is determined to be constructed. By combining the prevention path construction with regional load distribution, the final optimized emergency plan can be obtained.

[0129] For example, the fourth preset threshold is set to 20. During the emergency response plan optimization process, when the risk level ranking (based on risk assessment results or maintenance priority) exceeds 20, a prevention path is determined to be constructed. This threshold is consistent with the third preset threshold, ensuring that optimized emergency response plans are generated in a timely manner in high-risk scenarios.

[0130] Specifically, by iteratively matching the initial draft emergency plan with real-time updated micrometeorological monitoring data, it is first necessary to understand the composition of the micrometeorological monitoring data. This data typically includes indicators such as local wind speed, temperature fluctuations, and humidity changes, which are collected in real time by sensors deployed near power transmission equipment.

[0131] Specifically, the micrometeorological monitoring system employs a distributed sensor network, such as installing anemometers and temperature and humidity probes in high-voltage power line areas, updating data at second-level intervals to capture dynamic environmental changes. This real-time data update provides the basis for iterative matching, ensuring timely adjustments to contingency plans. Furthermore, the iterative matching process involves comparing the sequence of measures in the initial draft emergency plan with the micrometeorological data. For example, in a power transmission system, if the initial draft includes a sequence of load reduction and isolation operations, and real-time data shows a sudden increase in wind speed, the matching algorithm will identify an increased risk of wind pressure.

[0132] It should be noted that the matching algorithm is based on a threshold comparison mechanism. First, it extracts key features from the data, such as peak wind speed, and then compares them with preset rule thresholds in the draft. If the wind speed exceeds the threshold of 15 meters per second, sequence adjustment is triggered. The principle behind this mechanism is to adapt to dynamic environments through continuous cyclical comparisons, avoiding the limitations of static contingency plans.

[0133] In one possible implementation, the number of iterations can be set to 3 to 5, and the sequence priority is updated after each match, for example, moving the wind pressure mitigation measure to the first position in the sequence.

[0134] Preferably, when adjusting the sequence of measures to adapt to dynamic changes, the focus is on reordering and supplementing the sequence. The specific process involves analyzing trends in micrometeorological data, such as using simple trend calculations like wind speed change rates, to predict short-term environmental evolution. Then, based on this prediction, the sequence is adjusted; for example, in a substation scenario, if humidity data indicates a potential corrosion risk, a coating inspection step is inserted into the sequence. This adjustment ensures the sequence transitions progressively from initial prevention to response, logically forming a closed loop. In power equipment maintenance operations, this approach can handle sudden weather changes, such as prioritizing equipment fixing operations during storm warnings, thereby maintaining system stability. For example, in an embodiment of a coastal power transmission line, an initial draft might establish a sequence of measures for salt spray corrosion, while if real-time micrometeorological data detects a sharp rise in humidity, iterative matching adjusts the sequence to first perform insulation cleaning, then deploy backup power. This scenario demonstrates the versatility of the technology, applicable to power systems in different regions but within the same sector.

[0135] Understandably, the final optimized emergency plan is the result of integrating and adjusting sequences through multiple iterations.

[0136] Specifically, the final contingency plan combines a list of measures with a timeline to create an executable document, including a complete chain of alarm triggering, isolation execution, and recovery verification. This optimization process provides greater adaptability to business operations and reduces failure response delays.

[0137] In one embodiment, for applications involving power line equipment in mountainous areas, micrometeorological monitoring data emphasizes wind direction changes, and iterative matching adjusts the sequence to first assess wind loads and then perform support reinforcement. This variant highlights the flexibility of the technology without altering the core matching logic. Furthermore, the seamless integration of the entire process lies in the synchronization of data updates and matching; for example, sensor data is refreshed every minute, and the matching algorithm runs accordingly, ensuring real-time performance. In power systems, this design supports continuous monitoring and adapts to dynamic environments such as seasonal storms.

[0138] like Figure 2 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided; An embodiment of the present invention provides a power distribution network equipment detection device, comprising: The module includes an acquisition module 201, a first processing module 202, a first calculation module 203, a second calculation module 204, a second processing module 205, and a detection module 206. The acquisition module 201 is used to acquire micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested in real time; The first processing module 202 is used to process micro-meteorological monitoring data and vulnerability index data through a spatiotemporal interpolation algorithm to obtain a spatiotemporally standardized dataset; The first calculation module 203 is used to input the spatiotemporally standardized dataset into the random forest algorithm to calculate the dynamic correlation characteristics between micro-meteorological monitoring data and vulnerability index data; The second calculation module 204 is used to calculate the risk assessment result based on the dynamic correlation characteristics; The second processing module 205 is used to process the spatiotemporally standardized dataset through a long short-term memory network to generate a fault prediction result when the risk assessment result determines that the risk is high. The fault prediction result includes the fault probability and the possible fault location. The detection module 206 is used to detect the distribution network equipment to be tested based on the fault prediction results. It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the distribution network equipment detection method provided by any of the above-described method embodiments of the present invention.

[0139] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0140] Based on the above embodiments of the distribution network equipment detection method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the distribution network equipment detection method of any embodiment of the present invention.

[0141] For example, in this embodiment, the computer program can be divided into one or more modules, one or more modules are stored in memory and executed by a processor to complete the present invention. One or more module elements can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a terminal device.

[0142] Terminal devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. Terminal devices may include, but are not limited to, processors and memory.

[0143] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0144] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the distribution network equipment detection method described in any of the above-described method embodiments of the present invention.

[0145] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0146] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for testing distribution network equipment, characterized in that, include: Real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested; The micrometeorological monitoring data and the vulnerability index data are processed by a spatiotemporal interpolation algorithm to obtain a spatiotemporally standardized dataset; The spatiotemporally standardized dataset is input into the random forest algorithm to calculate the dynamic correlation characteristics between the micrometeorological monitoring data and the vulnerability index data; The risk assessment result is calculated based on the dynamic correlation characteristics. When the risk assessment result is determined to be high risk, the spatiotemporal standardized dataset is processed by a long short-term memory network to generate a fault prediction result; wherein, the fault prediction result includes the fault probability and the possible fault location; The distribution network equipment to be tested is tested based on the fault prediction results.

2. The distribution network equipment testing method as described in claim 1, characterized in that, The step of inputting the spatiotemporally standardized dataset into a random forest algorithm to calculate the dynamic correlation features between the micrometeorological monitoring data and the vulnerability index data includes: The spatiotemporally standardized dataset is input into the random forest algorithm, so that the random forest algorithm performs decision tree ensemble on the spatiotemporal variables and meteorological variables in the spatiotemporally standardized dataset to obtain the dynamic correlation characteristics between the micro-meteorological monitoring data and the vulnerability index data.

3. The distribution network equipment testing method as described in claim 2, characterized in that, The risk assessment result calculated based on the dynamic correlation characteristics includes: Calculate the association strength of the dynamic association features; The correlation strength is compared with a first preset threshold, and a potential fault risk level is generated based on the comparison result. Based on the level of potential failure risk, a maintenance priority is generated; The risk assessment results were calculated based on the maintenance priority and the micrometeorological monitoring data.

4. The distribution network equipment testing method as described in claim 3, characterized in that, When the risk assessment result determines the risk to be high, the spatiotemporal standardized dataset is processed through a long short-term memory network to generate a fault prediction result, including: When the risk assessment result is greater than the third preset threshold, it is determined to be high risk; The spatiotemporal normalized dataset is processed by the Long Short-Term Memory network to generate normalized sequences; Calculate the failure probability based on the standardized sequence; The fault prediction result is generated based on the fault probability and the location coordinate data in the spatiotemporal standardized dataset.

5. The distribution network equipment testing method as described in claim 4, characterized in that, The step of comparing the correlation strength with a first preset threshold and generating a potential fault risk level based on the comparison result includes: When the correlation strength is greater than the first preset threshold, it is determined to be at level one risk. When the correlation strength is greater than the second preset threshold and less than or equal to the first preset threshold, it is determined to be at level two risk. When the correlation strength is less than or equal to the second preset threshold, it is determined to be a level three risk level; The potential failure risk level is generated based on the first-level risk level, the second-level risk level, and the third-level risk level.

6. The distribution network equipment testing method as described in claim 5, characterized in that, The real-time acquisition of micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment under test includes: The micro-meteorological monitoring data is acquired in real time through a micro-meteorological sensor network; wherein, the micro-meteorological monitoring data includes wind speed data, lightning intensity data, temperature data, humidity data, and rainfall data; The vulnerability index data is acquired in real time through equipment status sensors; wherein the vulnerability index data includes: equipment aging index data, equipment load index data, equipment operating status index data, equipment mechanical performance index data, equipment insulation performance index data, and equipment environmental adaptability index data.

7. The distribution network equipment testing method as described in claim 6, characterized in that, The process of processing the micrometeorological monitoring data and the vulnerability index data using a spatiotemporal interpolation algorithm yields a spatiotemporally standardized dataset, including: Based on the micrometeorological monitoring data and the vulnerability index data, a comprehensive vulnerability index set is calculated; The geographic location and timestamp of the comprehensive vulnerability index are added according to the spatiotemporal annotation method to generate an annotated comprehensive vulnerability index set; wherein, the geographic location is generated from the location of the device status sensor, and the timestamp is generated from the acquisition time; The labeled comprehensive vulnerability index set is adjusted through a calibration mechanism to generate the original spatiotemporally labeled dataset. After cleaning the original spatiotemporally labeled dataset, spatiotemporally heterogeneous data is processed using a spatiotemporally interpolation algorithm to obtain a spatiotemporally standardized dataset.

8. A power distribution network equipment testing device, characterized in that, include: The system comprises an acquisition module, a first processing module, a first calculation module, a second calculation module, a second processing module, and a detection module. The acquisition module is used to acquire micro-meteorological monitoring data and vulnerability index data of the power distribution network equipment to be tested in real time; The first processing module is used to process the micrometeorological monitoring data and the vulnerability index data through a spatiotemporal interpolation algorithm to obtain a spatiotemporally standardized dataset; The first calculation module is used to input the spatiotemporally standardized dataset into the random forest algorithm to calculate the dynamic correlation characteristics between the micrometeorological monitoring data and the vulnerability index data; The second calculation module is used to calculate the risk assessment result based on the dynamic correlation characteristics; The second processing module is used to process the spatiotemporally standardized dataset through a long short-term memory network to generate a fault prediction result when the risk assessment result determines that the risk is high; wherein, the fault prediction result includes the fault probability and the possible fault location; The detection module is used to detect the distribution network equipment to be tested based on the fault prediction results.

9. A terminal device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the distribution network equipment detection method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the distribution network equipment detection method as described in any one of claims 1-7.

Images