Low-voltage distribution box data real-time acquisition method and system

Abstract

The invention provides a low-voltage distribution box data real-time acquisition method and system, and relates to the technical field of data processing, and the method comprises the steps: respectively determining space correlation weighting coefficients corresponding to a standardized meteorological load vector, a microenvironment load vector and a distribution box state load vector; respectively mapping the standardized meteorological load vector, the microenvironment load vector and the distribution box state load vector to three vertexes of an equilateral triangle to obtain a spatial feature weight of each load vector; fusing each load vector and the corresponding spatial correlation weighting coefficient, and performing regularization processing on the linear combination process by using the spatial feature weight to form an environmental load matrix containing a spatial topological relation; and based on the geometrical characteristics of the environmental load matrix, establishing a spatio-temporal evolution early warning model fused with a diffusion convolution operator so as to predict abnormal points. According to the invention, abnormal points can be effectively predicted, and a coupling propagation path of a potential fault can be identified.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method and system for real-time data acquisition of low-voltage distribution boxes. Background Art

[0002] With the rapid development of modern power systems and smart grids, low-voltage distribution boxes play a crucial role in the power network. However, due to the complex and changeable working environment of the distribution boxes, their operating states are easily affected by multiple factors such as meteorological conditions, micro-environmental factors, and the states of the equipment themselves. Therefore, it is particularly important to perform real-time data acquisition and status monitoring on the distribution boxes.

[0003] In order to overcome the limitations of traditional low-voltage distribution box data acquisition methods, in recent years, the industry has begun to explore the use of advanced sensing technologies and data analysis methods for real-time data acquisition and status monitoring of low-voltage distribution boxes. However, some of the existing real-time data acquisition methods only focus on a single environmental factor or equipment state, while ignoring the spatial topological relationships and temporal evolution characteristics among multiple factors, which results in limitations in accuracy and reliability when predicting and identifying potential faults. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for real-time data acquisition of low-voltage distribution boxes, which can effectively predict abnormal points and timely identify the coupling propagation paths of potential faults, thereby improving the operating reliability and safety of low-voltage distribution boxes.

[0005] To solve the above technical problem, the technical solution of the present invention is as follows: In a first aspect, a method for real-time data acquisition of low-voltage distribution boxes, the method includes: According to the historical data of the distribution box, respectively determine the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, micro-environment load vector, and distribution box state load vector; Map the standardized meteorological load vector, micro-environment load vector, and distribution box state load vector to the three vertices of an equilateral triangle respectively to obtain the spatial feature weights of each load vector; fuse each load vector with its corresponding spatial correlation weighting coefficient, and apply regularization processing to the linear combination process using the spatial feature weights to form an environmental load matrix including spatial topological relationships; Based on the geometric features of the environmental load matrix, establish a spatio-temporal evolution early warning model integrating diffusion convolution operators to predict abnormal points; Perform multi-scale matching degree analysis on the predicted abnormal points and the equipment aging feature spectrum, and combine the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for composite fault modes to identify the coupling propagation paths of potential faults.

[0006] Further, before respectively determining the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector based on the historical data of the distribution box, it further includes: Extract key meteorological factors from the dynamic meteorological data to construct a meteorological load vector, where the key meteorological factors include wind speed, wave height, and salt fog deposition rate; Extract key microenvironment factors from the cabin microenvironment data to construct a microenvironment load vector, where the key microenvironment factors include temperature and humidity gradient distribution and condensate film thickness; Obtain key indicators reflecting the electrical performance and insulation status of the distribution box to construct a distribution box status load vector, where the key indicators include leakage current harmonic components and insulation dielectric loss factors.

[0007] Further, according to the historical data of the distribution box, respectively determining the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector includes: Calculate the square values of the wind speed, wave height, and salt fog deposition rate in the meteorological load vector respectively and accumulate them to calculate the comprehensive strength evaluation value of the meteorological load vector; Calculate the square values of the temperature and humidity gradient distribution and condensate film thickness in the microenvironment load vector respectively and accumulate them to calculate the comprehensive strength value of the microenvironment load vector; Calculate the square values of the leakage current harmonic components and insulation dielectric loss factors in the distribution box status load vector respectively and accumulate them to calculate the comprehensive strength value of the distribution box status load vector; Fuse the comprehensive strength value of the meteorological load vector, the comprehensive strength value of the microenvironment load vector, and the comprehensive strength value of the distribution box status load vector to obtain the total strength value; Obtain the spatial correlation weighting coefficient of the meteorological load vector according to the comprehensive strength value of the meteorological load vector and the total strength value; Obtain the spatial correlation weighting coefficient of the microenvironment load vector according to the comprehensive strength value of the microenvironment load vector and the total strength value; Obtain the spatial correlation weighting coefficient of the distribution box status load vector according to the comprehensive strength value of the distribution box status load vector and the total strength value.

[0008] Further, map the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector to the three vertices of an equilateral triangle respectively to obtain the spatial characteristic weights of each load vector, including: According to the physical location of the distribution box, take the center point of the installation position of the distribution box as the origin of the polar coordinate system, and determine a direction as the polar axis; mark the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector as the initial points in the polar coordinate system, and the direction of each load vector represents its pointing in the multi-dimensional space; Using the historical data of the distribution box, set corresponding constraint conditions for each load vector, and the constraint conditions include the maximum and minimum values of the vector modulus length and the vector direction; Calculate the rotation matrix according to the geometric characteristics of the equilateral triangle, and use the rotation matrix to perform a rotation operation on the multi-dimensional tensor product to obtain the positions of the load vectors mapped to the vertices of the equilateral triangle; Calculate the spatial feature weights of each load vector according to the positions of the load vectors mapped to the vertices of the equilateral triangle.

[0009] Furthermore, fuse each load vector with its corresponding spatial correlation weighting coefficient, and use the spatial feature weights to impose regularization on the linear combination process to form an environmental load matrix including spatial topological relationships, including: Multiply the meteorological load vector, microenvironment load vector, and distribution box status load vector by their corresponding spatial correlation weighting coefficients respectively to obtain the weighted load vectors; Fuse the weighted load vectors to form a new comprehensive vector; Fuse the spatial feature weights with the new comprehensive vector to obtain the environmental load matrix.

[0010] Furthermore, based on the geometric characteristics of the environmental load matrix, establish a spatio-temporal evolution early warning model integrating the diffusion convolution operator to predict abnormal points, including: Analyze the geometric characteristics of the environmental load matrix, including the shape, size, and element distribution of the matrix; Collect various data during the operation of the distribution box, including current, voltage, temperature, and humidity data; Construct an early warning model corresponding to the real distribution box according to various data during the operation of the distribution box and the geometric characteristics of the environmental load matrix; Integrate sensor data into the early warning model to reflect the operation status of the distribution box in real time; Incorporate the diffusion convolution operator into the digital twin model to capture the spatio-temporal evolution characteristics of the data. Through the diffusion convolution operator, the early warning model extracts the spatio-temporal information during the operation of the distribution box, including the change trends of current and voltage and the temperature distribution; Use the diffusion convolution operator to extract features from the operation data of the distribution box to obtain spatio-temporal feature vectors, and analyze the spatio-temporal feature vectors to identify abnormal points during the operation of the distribution box.

[0011] Further, perform multi-scale matching degree analysis on the predicted abnormal points and the device aging characteristic spectrum, and combine the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for the composite fault mode, and identify the coupled propagation path of potential faults, including: Collect the historical operation data of the distribution box, including data in normal and abnormal states; extract the aging characteristic spectrum of the distribution box from the original data through Fourier transform; Calculate the cosine similarity between the predicted abnormal points and the device aging characteristic spectrum; Map the three key state parameters of the distribution box to the three vertices of the equilateral triangle, and calculate the phase angle difference between the three state parameters; Use the cosine similarity as the x-axis and y-axis, and the phase angle difference parameter of the equilateral triangle as the z-axis. In the three-dimensional space, each point represents a fault mode; In the three-dimensional matching space, group similar fault modes through clustering, and analyze the spatial relationship and evolution trend between different fault mode groups to identify potential fault coupling propagation paths.

[0012] In a second aspect, a real-time data acquisition system for a low-voltage distribution box includes: A determination module for respectively determining the spatial association weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box state load vector according to the historical data of the distribution box; A calculation module for respectively mapping the standardized meteorological load vector, microenvironment load vector, and distribution box state load vector to the three vertices of the equilateral triangle to obtain the spatial characteristic weights of each load vector; fuse each load vector with its corresponding spatial association weighting coefficient, and apply regularization processing to the linear combination process using the spatial characteristic weights to form an environmental load matrix including spatial topological relationships; A prediction module for establishing a spatio-temporal evolution early warning model integrating a diffusion convolution operator based on the geometric characteristics of the environmental load matrix to predict abnormal points; An analysis module for performing multi-scale matching degree analysis on the predicted abnormal points and the device aging characteristic spectrum, and combining the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for the composite fault mode, and identifying the coupled propagation path of potential faults.

[0013] In a third aspect, a computing device includes: One or more processors; A storage device for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the described method.

[0014] Fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described above is implemented.

[0015] The above solution of the present invention has at least the following beneficial effects: By determining the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box state load vector based on the historical data of the distribution box, various influencing factors can be comprehensively considered, and the relative importance of each factor can be accurately quantified, which helps to more comprehensively understand the operating environment of the distribution box.

[0016] Mapping the standardized load vectors to the vertices of an equilateral triangle and regularizing the linear combination process using spatial feature weights can form an environmental load matrix that includes spatial topological relationships. This processing method not only considers the spatial relationships between the load vectors but also can effectively reduce the dimension and complexity of the data, improving the efficiency and accuracy of data processing. At the same time, the formation of the environmental load matrix helps to reveal the internal connections and interaction mechanisms between the factors.

[0017] Based on the geometric characteristics of the environmental load matrix, a spatio-temporal evolution early warning model integrating diffusion convolution operators is established, which can realize real-time monitoring of the operating state of the distribution box and prediction of abnormal points. This model combines the advantages of diffusion convolution operators, can effectively capture the spatio-temporal evolution characteristics of the data, improve the accuracy and reliability of the prediction. By predicting abnormal points in a timely manner, it can provide strong support for the maintenance and management of the distribution box and reduce the risk of faults.

[0018] Perform multi-scale matching degree analysis on the predicted abnormal points and the equipment aging characteristic spectrum, and combine the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for the composite fault mode. This method can comprehensively consider multiple factors such as equipment aging and operating environment, accurately identify the coupling propagation paths of potential faults. By constructing a three-dimensional matching space, the propagation process and influence range of the faults can be intuitively displayed, providing strong guidance for fault troubleshooting and repair. At the same time, this also helps to improve the reliability and safety of the distribution box operation and reduce the losses caused by faults. Description of the Drawings

[0019] Figure 1 is a schematic flowchart of a method for real-time data acquisition of a low-voltage distribution box provided by an embodiment of the present invention.

[0020] Figure 2 is a schematic diagram of a system for real-time data acquisition of a low-voltage distribution box provided by an embodiment of the present invention. Detailed Embodiments

[0021] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0022] As Figure 1 shown, an embodiment of the present invention provides a method for real-time data acquisition of a low-voltage distribution box, and the method includes the following steps: Step 1: According to the historical data of the distribution box, respectively determine the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector; Step 2: Map the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector to the three vertices of an equilateral triangle respectively to obtain the spatial feature weights of each load vector; perform a tensor product operation on each load vector and its corresponding spatial correlation weighting coefficient, and use the spatial feature weights to impose regularization processing on the linear combination process to form an environmental load matrix including spatial topological relationships; Step 3: Based on the geometric features of the environmental load matrix, establish a spatio-temporal evolution early warning model integrating a diffusion convolution operator to predict abnormal points; Step 4: Perform multi-scale matching degree analysis on the predicted abnormal points and the device aging feature spectrum, and combine the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for the composite fault mode and identify the coupling propagation path of potential faults.

[0023] In the embodiments of the present invention, by analyzing the historical data of the distribution box, the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector are determined, which can more accurately quantify the influence of various external and internal factors on the operating state of the distribution box. This standardization and weighting process helps to improve the pertinence and accuracy of data collection. Mapping the standardized load vectors to the vertices of an equilateral triangle and regularizing the linear combination process through spatial feature weights to form an environmental load matrix containing spatial topological relationships not only simplifies the data processing process but also reveals the spatial correlation and interaction between the load vectors, which helps to more deeply understand the complexity and dynamics of the operating state of the distribution box. Based on the geometric characteristics of the environmental load matrix, a spatio-temporal evolution early warning model integrating a diffusion convolution operator is established, which can monitor the operating state of the distribution box in real time and accurately predict abnormal points. This model combines the advantages of the diffusion convolution operator and can capture the evolution law of data in time and space, thereby improving the accuracy and timeliness of fault prediction. This helps to realize the intelligent monitoring of the operating state of the distribution box and reduce the risk of faults. By performing multi-scale matching degree analysis on the predicted abnormal points and the equipment aging feature spectrum and combining the phase angle difference parameters between the vertices of the equilateral triangle, a three-dimensional matching space for composite fault modes is constructed, which can comprehensively and accurately identify the coupling propagation path of potential faults. This step not only considers the influence of equipment aging but also combines multiple factors such as the operating environment, providing more accurate guidance for fault troubleshooting and repair. This helps to improve the reliability and safety of the operation of the distribution box and reduce the losses and risks caused by faults.

[0024] In a preferred embodiment of the present invention, before respectively determining the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector according to the historical data of the distribution box, it further includes: Extract key meteorological factors from dynamic meteorological data to construct a meteorological load vector. The key meteorological factors include wind speed, wave height, and salt fog deposition rate. Specifically, it includes: obtaining real-time meteorological data from public meteorological data APIs, weather stations, or satellite data sources, determining the time range and spatial range of the required data to ensure the comprehensiveness and pertinence of the data; checking and removing duplicate records, missing values, or outliers, using moving average, median filtering, or other statistical methods to smooth the data, remove noise, and perform normalization or standardization processing on the data to ensure that data with different dimensions can be analyzed on the same scale; from the cleaned data, screen out meteorological factors directly related to the operating state of the distribution box, such as wind speed, wave height, salt fog deposition rate, etc. For these key factors, further data processing can be carried out, such as calculating statistics such as mean, maximum, and minimum values to better describe their characteristics; integrating the extracted key meteorological factors according to a specific standardization method. For example, z-score standardization (i.e., dividing by the standard deviation after subtracting the mean) can be used to convert each factor to the same dimension, and the standardized factors are combined into a meteorological load vector, which will contain key information such as wind speed, wave height, and salt fog deposition rate.

[0025] Extract key microenvironment factors from the cabin microenvironment data to construct a microenvironment load vector. The key microenvironment factors include temperature and humidity gradient distribution and condensate film thickness. Specifically, it includes: arranging temperature and humidity sensors at key positions in the cabin (such as near the distribution box, ventilation openings, corners, etc.), setting the sensors to collect data in real time at a certain frequency (such as every minute or every hour), and transmitting the data to the central data processing unit; receiving and storing the original data collected by the sensors, establishing a database or data file for management, and performing quality checks on the original data, including data integrity verification (whether there are missing values) and validity verification (whether it is within a reasonable range); for missing or abnormal data, handle it according to the actual situation using methods such as interpolation, deletion, or replacement.

[0026] Screen out the temperature and humidity data containing timestamp and location information from the sorted data. According to the timestamp, group the data by time period (such as hours, days), and calculate the average temperature and humidity for each time period; according to the location information of the sensors, divide the cabin into different regions. Calculate the average temperature and humidity for each region, select adjacent sensor data points, calculate the temperature and humidity differences between them, and divide these differences by the distance between the sensors to obtain the temperature and humidity gradient. For example, the formula can be used: gradient = (parameter difference) / (distance). Plot the calculated temperature and humidity gradient as a vector diagram or arrow diagram to visually display the direction and magnitude of the gradient.

[0027] Conduct a detailed investigation of each surface inside the cabin, record the types of materials used in different areas, consult relevant literature to obtain key physical property data such as the thermal conductivity, surface roughness, and hydrophilicity / hydrophobicity (usually measured by the contact angle) of each material; establish a physical property database, organize and store the collected data. Install temperature and humidity sensors at different positions inside the cabin to ensure that the sensor distribution can comprehensively reflect the temperature and humidity conditions inside the cabin, set up a sensor data acquisition system to record the temperature and humidity data at each position in real time, and preprocess the collected data, including data cleaning, outlier detection and correction, etc., to ensure the accuracy and reliability of the data; according to the principle of condensation formation, understand the conditions for condensation to occur, that is, when the surface temperature is lower than the dew point temperature in the air, water vapor in the air will condense on the surface to form a liquid film; use the real-time collected temperature and humidity data, combined with the material temperature of the cabin surface (which can be directly measured), to calculate the dew point temperature at each position; compare the surface temperature and the dew point temperature to determine which positions meet the conditions for condensation to occur.

[0028] Screen out the positions that meet the condensation conditions from the previously collected data, that is, the areas where the surface temperature is lower than the dew point temperature. For these positions, organize the corresponding real-time temperature and humidity data and the physical property data of the materials (thermal conductivity, surface roughness, hydrophilicity / hydrophobicity); Materials with high thermal conductivity will transfer heat faster, which may lead to the faster formation and evaporation of the condensation liquid film. Analyze the formation rate and stability of the condensation liquid film under different thermal conductivities; Rough surfaces may provide more condensation nuclei to promote the condensation process, study the relationship between roughness and the nucleation and growth of condensation droplets. Hydrophilic surfaces are more likely to attract water molecules to form a continuous liquid film; Hydrophobic surfaces cause droplet dispersion, analyze the influence of hydrophilicity / hydrophobicity on the morphology and distribution of the condensation liquid film.

[0029] Select an empirical formula that comprehensively considers factors such as the condensation rate, material thermal conductivity, and condensation time to estimate the thickness of the condensation liquid film. The empirical formula is as follows: ; Where, is the thickness of the condensation liquid film (unit: meter); k is an empirical coefficient; is the condensation time (unit: second); is the thermal conductivity of the material (unit: W / (m·K)); RH is the relative humidity (%); is the thermal diffusivity of the material (unit: m 2 / s), which is related to thermal conductivity, specific heat capacity, and density; is the surface temperature (unit: K); is the dew point temperature (unit: K); the empirical coefficient k plays a crucial role in the empirical formula for estimating the condensate film thickness, comprehensively reflecting the influence of various complex factors during the condensation process. Since the condensation process is restricted by multiple factors such as material properties, environmental conditions, and the properties of the condensing fluid, the value of k is not a fixed number but depends on the specific application scenario. The following is an explanation of the value range of k: When on a smooth surface: If in a high-temperature and high-humidity environment, the value range of k is 0.9 - 1.1. Although it is a smooth surface, in a high-temperature and high-humidity environment, the condensation driving force is strong, making the condensation process relatively easy to start, so the k value increases slightly; if in a low-temperature and low-humidity environment, the value range of k is 0.7 - 0.9. Under low-temperature and low-humidity conditions, there are fewer condensation nuclei on the smooth surface, and the start of condensation is more difficult, resulting in a significant reduction in the condensation rate, so the k value is smaller.

[0030] When on a moderately rough surface: If in a high-temperature and high-humidity environment, the value range of k is 1.1 - 1.3. On a moderately rough surface in a high-temperature and high-humidity environment, the number of condensation nuclei is moderate and the condensation driving force is strong, making the condensation process relatively rapid; if in a low-temperature and low-humidity environment, the value range of k is 0.9 - 1.1. Under low-temperature and low-humidity conditions, although the surface roughness is moderate, the condensation driving force weakens, resulting in a relatively slow condensation process.

[0031] When on a highly rough surface: If in a high-temperature and high-humidity environment, the value range of k is 1.4 - 1.6. On a highly rough surface in a high-temperature and high-humidity environment, the combination of a large number of condensation nuclei and a strong condensation driving force makes the condensation very rapid, and the liquid film forms quickly; if in a low-temperature and low-humidity environment, the value range of k is 1.2 - 1.4. Although it is a low-temperature and low-humidity condition, the highly rough surface still provides sufficient condensation nuclei, making the condensation process relatively fast.

[0032] When on a hydrophilic surface: If in a high-temperature and high-humidity environment, the value range of k is 1.0 - 1.2. On a hydrophilic surface in a high-temperature and high-humidity environment, it is more conducive to the adsorption and spreading of water molecules, forming a stable and continuous liquid film; if in a low-temperature and low-humidity environment, the value range of k is 0.8 - 1.0. Under low-temperature and low-humidity conditions, the hydrophilic surface can still maintain a certain adsorption capacity, but the condensation rate slows down relatively.

[0033] When on a hydrophobic surface: If in a high-temperature and high-humidity environment, the value range of k is 1.3 - 1.5. In a high-temperature and high-humidity environment, the hydrophobic surface may cause the aggregation of water molecules and the rapid formation of droplets, and the condensation process is relatively complex; if in a low-temperature and low-humidity environment, the value range of k is 1.1 - 1.3. Although it is in low-temperature and low-humidity conditions, the characteristics of the hydrophobic surface may cause the unstable aggregation of condensate droplets, affecting the uniformity of the condensation process.

[0034] When calculating, the specific calculation process of the empirical formula is as follows: Convert the condensation time from seconds to hours, that is, divide the condensation time by 3600; if the original data of the surface temperature and the dew point temperature are in degrees Celsius, they need to be converted to Kelvin, that is, add 273.15 to the surface temperature and add 273.15 to the dew point temperature; calculate the difference between the surface temperature and the dew point temperature, that is, △T = Ts - Td , this difference reflects the temperature driving force for condensation to occur; normalize the relative humidity, normalize the relative humidity RH to the range of 0 to 1, that is, divide the relative humidity RH by 100, which represents the water vapor saturation in the air. Substitute the prepared data above into the unit-normalized empirical formula to obtain the thickness of the condensate film.

[0035] Substitute the real-time temperature and humidity data that meet the condensation conditions and the physical property data of the material into the above empirical formula, calculate for each position that meets the condensation conditions, and obtain the estimated value of the thickness of the condensate film. Conduct statistical analysis on the calculated condensate film thickness data, such as calculating the average value, standard deviation, maximum value, etc., and identify the key areas with a larger film thickness. These areas may be the places where condensation problems are most serious.

[0036] Standardize the extracted key microenvironment factors (such as temperature and humidity gradient distribution, condensate film thickness) to eliminate the dimension difference and facilitate comparison. Combine the standardized factors in a certain order to form a microenvironment load vector.

[0037] Obtain the key indicators reflecting the electrical performance and insulation status of the distribution box to construct a distribution box status load vector. The key indicators include the harmonic component of the leakage current and the insulation dielectric loss factor. Specifically, through the sensors or special monitoring equipment built in the distribution box, monitor the electrical performance and insulation status of the distribution box in real time, collect and record various data during the operation of the distribution box, such as current, voltage, temperature, etc.; extract the key indicators reflecting the electrical performance and insulation status of the distribution box from the collected data, such as the harmonic component of the leakage current and the insulation dielectric loss factor. These indicators can directly reflect the health status of the distribution box. Standardize the extracted key indicators and integrate them into a distribution box status load vector.

[0038] In the embodiments of the present invention, extracting key meteorological factors from dynamic meteorological data can accurately capture and quantify meteorological conditions that have a direct impact on the operating state of the distribution box, such as wind speed, wave height, and salt fog deposition rate. By extracting these key meteorological factors, the impact of the external environment on the performance of the distribution box can be evaluated more accurately.

[0039] Integrating the extracted key meteorological factors into a meteorological load vector helps simplify complex and variable meteorological conditions into an operable mathematical model. This not only facilitates data processing and analysis but also enables a more intuitive understanding of the specific impact of meteorological factors on the operation of the distribution box. The microenvironment inside the cabin is also crucial for the operation of the distribution box. By extracting key microenvironment factors such as temperature and humidity gradient distribution and condensate film thickness, the subtle changes in the environment where the distribution box is located can be understood more deeply, thereby more accurately predicting and addressing possible faults or performance degradation caused thereby. Integrating the key microenvironment factors into a microenvironment load vector helps transform complex microenvironment conditions into a mathematical form for analysis, which can simplify the data processing process and improve the accuracy and efficiency of analysis, providing valuable reference information for the operation and maintenance of the distribution box. Key indicators such as leakage current harmonic components and insulation medium loss factors are important bases for evaluating the electrical performance and insulation state of the distribution box. By obtaining these indicators, the health status of the distribution box can be monitored in real time, potential safety hazards can be detected and addressed in a timely manner, and the stable operation of the power system can be ensured. Integrating these key indicators into a distribution box status load vector helps comprehensively and systematically evaluate the overall performance of the distribution box. This not only facilitates real-time monitoring and early warning of the distribution box status but also improves the reliability and safety of the power system.

[0040] In a preferred embodiment of the present invention, in step 1 above, according to the historical data of the distribution box, the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector are determined respectively, including: Step 11: Square and add the wind speed, wave height, and salt fog deposition rate in the meteorological load vector respectively to obtain the sum of squares of the meteorological load vector; take the square root of the sum of squares to obtain the comprehensive intensity value of the meteorological load vector; Step 12: Square and add the temperature and humidity gradient distribution and condensate film thickness in the microenvironment load vector respectively to obtain the sum of squares of the microenvironment load vector; take the square root of the sum of squares to obtain the comprehensive intensity value of the microenvironment load vector; Step 13: Square and add the leakage current harmonic components and insulation medium loss factors in the distribution box status load vector respectively to obtain the sum of squares of the distribution box status load vector; take the square root of the sum of squares to obtain the comprehensive intensity value of the distribution box status load vector; Step 14: Add the comprehensive strength value of the meteorological load vector, the comprehensive strength value of the microenvironment load vector, and the comprehensive strength value of the distribution box status load vector to obtain the total strength value; Step 15: Divide the comprehensive strength value of the meteorological load vector by the total strength value to obtain the spatial correlation weighting coefficient of the meteorological load vector; Step 16: Divide the comprehensive strength value of the microenvironment load vector by the total strength value to obtain the spatial correlation weighting coefficient of the microenvironment load vector; Step 17: Divide the comprehensive strength value of the distribution box status load vector by the total strength value to obtain the spatial correlation weighting coefficient of the distribution box status load vector.

[0041] In the embodiment of the present invention, in Step 11, it can comprehensively reflect the influence of various meteorological factors on the distribution box, making the evaluation more comprehensive and accurate. Step 12 helps to understand the specific conditions of the microenvironment where the distribution box is located, providing an important reference for subsequent maintenance and protection. Step 13 can directly reflect the internal state of the distribution box, helping to detect potential safety hazards in a timely manner. Step 14 realizes the comprehensive quantification of the influence of different loads, providing a basis for determining the spatial correlation weighting coefficients of each load subsequently. Step 15, this coefficient reflects the relative importance of meteorological factors on the distribution box, helping to accurately locate the main influencing factors in a complex external environment. Step 16, calculating the spatial correlation weighting coefficient of the microenvironment load vector can quantify the influence degree of microenvironment factors on the distribution box status. Step 17 helps to evaluate the influence of the distribution box's own state on its overall performance. These steps not only consider external environmental factors but also take into account the operating state of the distribution box itself, providing a strong guarantee for the safe and stable operation of the distribution box.

[0042] In the embodiment of the present invention, meteorological conditions have a significant impact on the operating state of low-voltage distribution boxes. Meteorological factors such as wind speed, wave height (in coastal or offshore environments), and salt fog deposition rate can directly or indirectly affect the heat dissipation performance, electrical insulation performance, and physical stability of the distribution box. Wind speed affects the heat dissipation efficiency of the distribution box. High wind speed may accelerate equipment cooling, but it may also bring additional mechanical stress or introduce dust and debris. Wave height in offshore or coastal environments affects the installation foundation stability of the distribution box, thereby affecting the normal operation of the equipment; the salt fog deposition rate has a significant corrosive effect on electrical equipment, and long-term exposure may lead to problems such as decreased insulation performance and poor contact.

[0043] The temperature, humidity conditions and condensation phenomenon in the microenvironment where the distribution box is located (such as inside the cabin) have an important impact on the operation and lifespan of the equipment. The temperature and humidity gradient distribution and the thickness of the condensation liquid film are key indicators reflecting the status of the microenvironment. An uneven temperature and humidity distribution may cause thermal stress or moisture stress inside the equipment, affecting the performance and lifespan of the equipment. The formation of the condensation liquid film may lead to problems such as electrical short circuits and a decline in insulation performance, especially in a high-humidity environment.

[0044] The electrical performance and insulation status of the distribution box are crucial for evaluating its operating status and predicting potential faults. The harmonic components of the leakage current and the insulation dielectric loss factor are important indicators reflecting the electrical performance and insulation status of the distribution box. The existence of the leakage current may indicate insulation defects or grounding problems in the equipment, while the harmonic components may reflect nonlinear loads or faults inside the equipment. The insulation dielectric loss factor is an important indicator for measuring the aging degree of the insulation material and the insulation performance, and its change may predict the occurrence of insulation faults.

[0045] For example, the low-voltage distribution box of an offshore oil platform frequently malfunctioned during operation, resulting in production interruptions and safety hazards. In order to find out the cause of the faults and take measures, it was decided to use the above methods to conduct real-time monitoring and data analysis on this distribution box.

[0046] Deploy meteorological sensors around the distribution box to collect meteorological data such as wind speed, wave height, and salt spray deposition rate in real time. Arrange temperature and humidity sensors in the cabin where the distribution box is located to monitor the temperature and humidity gradient distribution, and measure the harmonic components of the leakage current and the insulation dielectric loss factor of the distribution box. According to the collected data, construct a meteorological load vector, a microenvironment load vector, and a distribution box status load vector, and use historical data to determine the spatial correlation weighting coefficients corresponding to each load vector. Through tensor product operation and regularization processing, form an environmental load matrix containing spatial topological relationships. Establish a spatio-temporal evolution early warning model to monitor the operating status of the distribution box in real time and predict abnormal points. According to the prediction results of the early warning model, combined with multi-scale matching degree analysis and phase angle difference parameter analysis, identify the coupling propagation path of potential faults, conduct a comprehensive inspection of the distribution box, and find that due to long-term exposure to a high-salt spray environment, some insulation materials are severely aged, resulting in an increase in leakage current and a decline in insulation performance. Replace the aged insulation materials and strengthen the sealing performance of the distribution box to prevent the intrusion of salt spray and moisture.

[0047] After the implementation of the above measures, the failure rate of the distribution box has been significantly reduced, the number of production interruptions has decreased, and potential safety hazards have been effectively controlled. At the same time, through real-time monitoring and data analysis, maintenance personnel can promptly discover and handle potential problems, improving the reliability and safety of the distribution box.

[0048] In a preferred embodiment of the present invention, in step 2, the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector are respectively mapped to the three vertices of an equilateral triangle to obtain the spatial feature weights of each load vector, including: Step 21: According to the physical location of the distribution box, use the center point of the installation location of the distribution box as the origin of the polar coordinate system, and determine a direction as the polar axis; mark the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector as initial points in the polar coordinate system. The direction of each load vector represents its pointing in the multi-dimensional space, specifically including: determining the physical location of the distribution box, and taking the center point of its installation location as the reference, setting this point as the origin O of the polar coordinate system, and selecting a direction convenient for calculation and visualization as the polar axis. For example, the due north direction or the main orientation of the distribution box can be selected as the polar axis; represent the already standardized meteorological load vector, microenvironment load vector, and distribution box status load vector as points M1, M2, and M3 respectively, and mark the initial positions of these points in the polar coordinate system; the direction of each load vector is determined by its multi-dimensional data, and this direction is represented by an angle in the polar coordinate system to ensure that the direction of each load vector accurately reflects its pointing in the multi-dimensional space.

[0049] Step 22: Use the historical data of the distribution box to set corresponding constraint conditions for each type of load vector. The constraint conditions include the maximum and minimum values of the vector modulus length and the vector direction, specifically including: collecting and analyzing the historical data of the distribution box, especially the data related to meteorology, microenvironment, and distribution box status, and setting constraint conditions for each type of load vector according to the range and change trend of the historical data. For example, the modulus length of the meteorological load vector may be restricted by the maximum and minimum values of wind speed, wave height, and salt fog deposition rate, and determine the direction constraints of each load vector. These direction constraints are based on the correlation or physical laws between various factors in the historical data.

[0050] Step 23: Calculate the rotation matrix according to the geometric properties of the equilateral triangle, and use the rotation matrix to perform a rotation operation on the multi-dimensional tensor product to obtain the positions of the load vectors mapped to the vertices of the equilateral triangle, specifically including: setting one vertex of the equilateral triangle as the origin O(0, 0), and assuming that the side length of the equilateral triangle is , in the two-dimensional plane, the properties of the equilateral triangle can be used to determine the coordinates of the other two vertices. Since the interior angles of an equilateral triangle are all 60 degrees, polar coordinates can be used to determine the vertex positions; with the origin O as the reference, the second vertex A can be on the positive x-axis, and the coordinates are A( a , 0); the third vertex B can be obtained by rotating the first side using the properties of the equilateral triangle, and its coordinates in the Cartesian coordinate system can be obtained through the rotation matrix; using polar coordinates, the angle of vertex B is 60 degrees, and the distance from the origin is also , so its polar coordinates are , converted to Cartesian coordinates, the coordinates of vertex B are ; To rotate the load vector to the vertices of the equilateral triangle, calculate the rotation matrix. In a two-dimensional plane, the rotation matrix is in the form of: ; Where is the angle of rotation; For rotating the load vector to vertex A, no rotation is required (since it is the vertex in the horizontal direction), so ; For rotating to vertex B, the vector needs to be rotated by 60 degrees, so , substituting into the rotation matrix, the rotation matrix for rotating to vertex B is: .

[0051] Step 24, According to the positions of the load vectors mapped to the vertices of the equilateral triangle, calculate the spatial feature weights of each load vector, specifically including: There are three load vectors , and , and they have been mapped to the three vertices A, B, and C of the equilateral triangle; According to the previous mapping process, the corresponding relationships can be directly determined: corresponds to vertex A; corresponds to vertex B; corresponds to vertex C; Use the distance-based weight assignment method to calculate the spatial feature weights. Calculate the distance of each load vector to the centroid of the triangle. The centroid G of the equilateral triangle can be obtained by the arithmetic mean of the vertex coordinates. The closer the load vector is, the relatively higher its weight can be set because it may have a greater impact on the overall performance. Use the inverse proportional function to assign weights according to the distance. For example, the weight can be set to an inverse relationship with the distance d: ; To ensure that the sum of all weights is 1, normalize the calculated weights to obtain the normalized weights ; Output the normalized spatial feature weights of each load vector, and these weights will be used to evaluate the comprehensive impact of each load on the performance of the distribution box. For example: The spatial feature weight of ; The spatial feature weight of ; The spatial feature weight of , where .

[0052] In an embodiment of the present invention, in step 21, by taking the center point of the installation position of the distribution box as the origin of the polar coordinate system and determining a direction as the polar axis, the initial positions of each load vector in space can be clarified; marking the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector in the polar coordinate system helps to intuitively understand and display the distribution and direction of these multi-dimensional data in space. In step 22, by using the historical data of the distribution box to set corresponding constraint conditions for each load vector, it can ensure that the analyzed load vectors are within a reasonable range, thereby enhancing the reliability and accuracy of the data. Setting the maximum and minimum values of the vector modulus length and the vector direction helps to prevent data anomalies or deviations from the actual situation. In step 23, calculating the rotation matrix according to the geometric characteristics of an equilateral triangle and using this matrix to perform a rotation operation on the multi-dimensional tensor product can ensure that the structure of the data remains consistent during the mapping process, avoiding information distortion or deformation; by performing the rotation operation to map the load vectors to the vertices of the equilateral triangle, the subsequent calculation process of the spatial feature weights can be simplified, improving the calculation efficiency. In step 24, calculating the spatial feature weights of each load vector according to the positions of the load vectors mapped to the vertices of the equilateral triangle can quantify the influence degree of each load vector on the performance of the distribution box in space; through the calculation of the spatial feature weights, the load vector with the greatest influence on the performance of the distribution box can be identified.

[0053] In a preferred embodiment of the present invention, performing a tensor product operation on each load vector and its corresponding spatial correlation weighting coefficient, and using the spatial feature weights to impose regularization processing on the linear combination process to form an environmental load matrix including spatial topological relationships, including: Multiplying the meteorological load vector, microenvironment load vector, and distribution box status load vector by their corresponding spatial correlation weighting coefficients respectively to obtain each weighted load vector; Adding the weighted load vectors to form a new comprehensive vector; Multiplying the spatial feature weights by the new comprehensive vector to obtain the environmental load matrix.

[0054] In the embodiment of the present invention, by performing a tensor product operation on the load vector and its corresponding spatial association weighting coefficient, the influence of spatial characteristics on environmental loads can be captured and reflected more accurately. This method fully considers the response differences of different spatial positions to environmental factors, thereby improving the accuracy and applicability of the model. Introducing spatial feature weights for regularization processing during the linear combination process can effectively prevent the model from overfitting, enhance the generalization ability and stability of the model. Regularization also helps to reduce the influence of outliers or noise data on the model, making the model more robust. By weighted integration of different data sources such as meteorological load vectors, micro-environment load vectors, and distribution box status load vectors, the influence of the environment on the distribution box status can be reflected more comprehensively. The fusion of such multi-source information helps to improve the prediction accuracy and reliability of the model. The finally formed environmental load matrix containing spatial topological relationships not only considers the influence of a single factor but also comprehensively considers the interactions between multiple factors, enabling the model to more accurately describe the state changes of the distribution box in the actual operating environment.

[0055] In a preferred embodiment of the present invention, step 3, based on the geometric features of the environmental load matrix, establishing a spatio-temporal evolution early warning model integrating diffusion convolution operators to predict abnormal points, may include: Step 31, analyzing the geometric features of the environmental load matrix, including the shape, size, and element distribution of the matrix, specifically including: analyzing the environmental load matrix, observing the shape of the matrix, recording its number of rows and columns, analyzing the size of the matrix, that is, the number of elements it contains, and statistically analyzing the distribution of elements in the matrix. According to the element distribution, identify possible peaks, outliers, or specific patterns.

[0056] Step 32, collecting various data during the operation of the distribution box, including current, voltage, temperature, and humidity data, specifically including: determining the types of data to be collected, including current, voltage, temperature, and humidity; installing corresponding sensors on the distribution box, such as ammeters, voltmeters, temperature and humidity sensors, etc., setting up a data acquisition system to ensure that data can be read from the sensors regularly or in real-time, and storing the collected data in a database for subsequent analysis.

[0057] Step 33, constructing an early warning model corresponding to the real distribution box according to various data during the operation of the distribution box and the geometric features of the environmental load matrix, specifically including: Randomly divide the processed dataset into a training set and a validation set according to a certain ratio (e.g., 70% training set, 30% validation set); design the neural network structure, including the number of neurons in the input layer, hidden layer, and output layer; initialize the weight and bias parameters of the neural network, and select activation functions (such as ReLU) and loss functions (such as mean squared error); use the training set data to train the neural network model, calculate the predicted values through forward propagation, and then calculate the gradients and update the weight and bias parameters of the model through the backpropagation algorithm; during the training process, adopt an optimization algorithm (such as gradient descent) to minimize the loss function, and set appropriate learning rates and the number of iterations; monitor the changes in the loss function value and accuracy during the training process to ensure that the performance of the model on the training set gradually improves and there is no overfitting phenomenon.

[0058] Use the validation set data to validate the trained neural network model. Calculate the predicted values on the validation set through forward propagation, compare the predicted values with the actual operation data, and calculate the accuracy of the model; analyze the model performance based on the accuracy. If the accuracy does not meet the expectations, adjust the parameters of the neural network model (such as increasing the number of hidden layers, changing the activation function, etc.), and retrain and validate again. Repeat model training, model validation, and adjustment until the performance of the model on the validation set reaches a satisfactory level to obtain the final trained neural network model as the early warning model.

[0059] Step 34, integrate sensor data into the early warning model to reflect the operating state of the distribution box in real time, specifically including: obtaining data from sensors in real time, preprocessing the obtained sensor data, such as denoising, normalization, etc.; inputting the preprocessed data into the early warning model so that the model can update its state in real time to reflect the latest operating conditions of the distribution box.

[0060] Step 35, incorporate diffusion convolutional operators into the digital twin model to capture the spatio-temporal evolution characteristics of the data. Through the diffusion convolutional operators, the early warning model extracts the spatio-temporal information during the operation of the distribution box, including the change trends of current and voltage and the temperature distribution, specifically including: Analyze the existing structure of the early warning model to understand the dimensions, shapes of the data after feature extraction, and the connection methods between existing layers; the feature extraction part is usually responsible for extracting useful information from the original data. After this process, the data will contain the key features of the distribution box operation, but may lack a deep expression of spatio-temporal relationships. Therefore, the diffusion convolutional layer should be placed immediately after the feature extraction part to be able to process these extracted features and further capture their spatio-temporal dependencies; determine the size of the convolutional kernel according to the characteristics of the distribution box data. A larger convolutional kernel can capture a wider range of spatial relationships, and usually, one can start trying common sizes such as 3×3 or 5×5; the stride determines the span at which the convolutional kernel moves on the input data. A stride of 1 means that the convolutional kernel will move point by point, which helps to capture more detailed features. Increasing the stride can reduce the dimension of the output but may lose some information. In the initial design, the stride is usually set to 1; after determining the specific parameters of the diffusion convolutional layer, insert it into the position after the feature extraction part in the early warning model to make the input and output of the diffusion convolutional layer compatible with other parts of the model. After integrating the diffusion convolutional layer into the model, use a small part of the training data for preliminary testing to ensure that the model can run normally and there are no obvious errors. According to the test results, fine-tune the parameters of the diffusion convolutional layer to achieve the best balance of performance and efficiency; Convert the data of the distribution box and its related components into graph-structured data. Each node represents a component, and the edges between nodes represent the connection relationships between components. Assign feature vectors to each node, and these feature vectors contain the spatio-temporal information of the component, such as current, voltage, temperature, etc.; in the diffusion convolutional layer, define the convolutional operation to capture the spatio-temporal dependencies between nodes, which usually involves the process of information propagation and aggregation on the graph. Through the training process, the model will learn how to update the feature representation of nodes according to the neighborhood information of nodes, so as to capture the spatio-temporal evolution features; set the diffusion coefficient and time step. The diffusion coefficient controls the propagation speed of information in the graph. A smaller diffusion coefficient will result in slower information propagation, while a larger diffusion coefficient will make the information spread to the entire graph faster; the time step determines the distance of information propagation in each time step. A smaller time step allows the model to capture spatio-temporal changes more finely; use the operation data of the distribution box containing spatio-temporal information as the training set to train the early warning model with the introduced diffusion convolutional layer. During the training process, adjust the weight and bias parameters of the model through optimization algorithms (such as gradient descent) to minimize the loss function; use the validation set data to evaluate the performance of the model, which can be done by calculating the accuracy of the model on the validation set. Adjust and optimize the model according to the validation results to improve its performance.

[0061] Step 36: Use a diffusion convolutional operator to extract features from the operation data of the distribution box to obtain spatio-temporal feature vectors, and analyze the spatio-temporal feature vectors to identify abnormal points during the operation of the distribution box, specifically including: Organize the preprocessed data in the format required by the model, usually in the form of a time series, and input the organized data into an early warning model that has incorporated a diffusion convolutional operator; when the data flows through the diffusion convolutional layer of the model, this layer will use its convolutional kernel to capture the spatio-temporal correlation information in the data. Through diffusion convolution calculation, the model extracts the spatio-temporal feature vectors during the operation of the distribution box, and these feature vectors encode the spatio-temporal change information of key parameters such as current, voltage, and temperature.

[0062] Perform post-processing and analysis on the extracted spatio-temporal feature vectors to reveal the potential patterns and abnormal behaviors therein. A clustering algorithm (such as K-means) can be used to group the feature vectors to identify different operation modes or states. By comparing the difference between the current feature vector and the clustering center or the historical average feature vector, abnormal changes during the operation of the distribution box can be detected; set a threshold to determine whether the change of the feature vector exceeds the normal range, and the threshold can be determined according to historical data; when the value (or its change amount) of a certain spatio-temporal feature vector exceeds the preset threshold, mark it as an abnormal point, and the abnormal point may correspond to a fault, performance degradation, or an impending problem in the distribution box.

[0063] In the embodiment of the present invention, by analyzing the geometric features of the environmental load matrix and combining the real-time data during the operation of the distribution box, an early warning model corresponding to the real distribution box can be constructed. This model can more accurately simulate the actual operation of the distribution box, thereby improving the accuracy of abnormal point prediction. By integrating sensor data, the early warning model can reflect the operation status of the distribution box in real time, which means that once an abnormal situation occurs in the distribution box, the model can quickly capture and issue a warning, helping to take timely measures to prevent the expansion of the fault. The early warning model incorporating a diffusion convolutional operator can capture the spatio-temporal evolution features of the data, which enables the model to more deeply understand the dynamic changes during the operation of the distribution box, such as the change trends of current and voltage and the temperature distribution, so as to more accurately predict abnormal points. By using a diffusion convolutional operator to extract features from the operation data of the distribution box to obtain spatio-temporal feature vectors, maintenance personnel can more conveniently monitor and analyze the operation status of the distribution box, which helps to timely discover and handle potential problems, improve the maintenance efficiency and the stability of the power distribution system. Through accurate prediction and real-time monitoring, unnecessary maintenance inspections and downtime can be reduced, thereby reducing the operating cost. At the same time, timely warning can also avoid the occurrence of major faults, further reducing the cost of equipment repair and replacement.

[0064] In a preferred embodiment of the present invention, in step 4, the predicted abnormal points are subjected to multi-scale matching degree analysis with the equipment aging characteristic spectrum, and combined with the phase angle difference parameters between the vertices of the equilateral triangle, a three-dimensional matching space for the composite fault mode is constructed to identify the coupling propagation path of potential faults, which may include: Step 41, collect the historical operation data of the distribution box, including data in normal and abnormal states; extract the aging characteristic spectrum of the distribution box from the original data through Fourier transform, specifically including: obtain the historical operation data from the monitoring system of the distribution box, and this data should include time series data of key parameters such as current, voltage, and temperature, ensuring that the collected data covers the operation of the distribution box in normal and abnormal states; clean the collected data, remove noise, outliers, and missing data, and perform normalization processing on the data to eliminate the dimensional differences between different parameters; apply Fourier transform (such as fast Fourier transform FFT) to the preprocessed time series data, and convert the time-domain signal into a frequency-domain signal through Fourier transform, so as to extract the aging characteristic spectrum of the distribution box, and these characteristic spectra can reflect the frequency components and periodic characteristics of the equipment during operation.

[0065] Step 42, calculate the cosine similarity between the predicted abnormal points and the equipment aging characteristic spectrum, specifically including: represent the data of the predicted abnormal points and the aging characteristic spectrum extracted from the historical data as feature vectors, and for each predicted abnormal point, calculate the cosine similarity between it and each aging characteristic spectrum. The cosine similarity measures the similarity in direction between two vectors, and its value ranges from -1 to 1. The closer the value is to 1, the more similar the two vectors are.

[0066] Step 43, map the three key state parameters of the distribution box to the three vertices of an equilateral triangle, and calculate the phase angle difference between the three state parameters, specifically including: select the three key state parameters of the distribution box (such as current, voltage, and temperature), map these three parameters to the three vertices of the equilateral triangle, and each vertex represents a state parameter; according to the values of the three state parameters, calculate their relative positions in the equilateral triangle, and then determine the phase angle between them. The phase angle reflects the relative relationship and change trend between the three state parameters, and calculate the difference between the phase angles of the three state parameters at the predicted abnormal point and the phase angles in the normal state.

[0067] Step 44: Using cosine similarity as the x-axis and y-axis, and the equilateral triangle phase angle difference parameter as the z-axis, in three-dimensional space, each point represents a fault mode, specifically including: using two cosine similarity values (which can be similarities with different aging characteristic spectra) as the coordinate values of the x-axis and y-axis respectively, and using the equilateral triangle phase angle difference parameter as the coordinate value of the z-axis. In this three-dimensional space, the coordinates of each point represent a specific fault mode, where the x-axis and y-axis reflect the similarity with the aging characteristic spectra, and the z-axis reflects the phase relationship between state parameters.

[0068] Step 45: In the three-dimensional matching space, group similar fault modes through K-means, and analyze the spatial relationships and evolution trends between different fault mode groups to identify potential fault coupling propagation paths, specifically including: According to actual needs or experience, select an appropriate number of clusters K, that is, it is desired to divide the fault modes into K groups; randomly select K points as the initial cluster centers. For each point in the space, calculate the distance between it and each cluster center (usually using the Euclidean distance), and assign each point to the group where the nearest cluster center is located; for each group, calculate the average coordinate value of all points within it, and use this average value as the new cluster center. Repeat the above steps of "iteratively assigning points to cluster centers" and "updating cluster centers" until the stopping condition is met (such as the cluster centers no longer change significantly, or the preset maximum number of iterations is reached); For each pair of fault mode groups, calculate the average distance or the center distance between them, which can be achieved by calculating the distance between the two cluster centers, such as calculating the average distance between all pairs of points in the two groups; create a matrix, where each element represents the distance between a pair of fault mode groups; according to the distance values in the proximity matrix, sort the fault mode groups, find the groups that are closest in space, and apply statistical methods (such as Pearson correlation analysis) to explore the potential structure and associations between the fault mode groups; if there is an obvious linear, curved or other geometric arrangement, this may imply a certain development or evolution law between the fault modes; based on the spatial proximity analysis, connect the closest fault mode groups, which can be achieved by drawing line segments or arrows in three-dimensional space to represent the possible fault propagation directions. Among the identified propagation paths, some fault mode groups may play the role of key nodes, that is, they are important hubs for fault propagation.

[0069] In the embodiment of the present invention, the "three-dimensional matching space" is an abstract space model for comprehensively analyzing and identifying potential fault coupling propagation paths of distribution boxes. In this space: The x-axis and y-axis are: taking two cosine similarity values between the predicted anomaly point and the device aging characteristic spectra as coordinates. The cosine similarity measures the similarity in direction between the feature vector of the predicted anomaly point and the feature vectors of different aging characteristic spectra, and its value ranges from -1 to 1. The closer the value is to 1, the more similar the two vectors are. Through these two similarity values, the correlation degree between the predicted anomaly point and different aging modes can be reflected.

[0070] The z-axis is: taking the equilateral triangle phase angle difference parameter as the coordinate. This parameter is obtained by mapping three key state parameters (such as current, voltage, temperature) of the distribution box to the three vertices of an equilateral triangle and calculating the difference between the phase angles of the three state parameters at the predicted anomaly point and the phase angles in the normal state. It reflects the relative relationship and change trend between the state parameters of the distribution box. The meaning of the point is that in this three-dimensional space, each point represents a specific fault mode, and the coordinates (x, y, z) of the point comprehensively reflect the similarity of this fault mode with the device aging characteristic spectra and the difference from the normal state phase relationship.

[0071] By collecting the historical operation data of the distribution box in the normal state and the abnormal state, and extracting the aging characteristic spectra, a rich fault feature library can be established. When a new predicted anomaly point appears, matching it with this feature library can more accurately identify potential fault types. The faults of the distribution box are often not isolated, but the result of the interaction of multiple factors. By calculating the cosine similarity between the predicted anomaly point and the aging characteristic spectra and representing these relationships in a three-dimensional space, the coupling relationship between different fault modes can be clearly revealed, which helps to deeply understand the fault propagation path and evolution mechanism. By analyzing the phase angle difference between the key state parameters of the distribution box, the subtle signs of device performance changes can be captured. Combining this information with the cosine similarity and representing it in a three-dimensional matching space can help the operation and maintenance personnel discover potential fault signs earlier, so as to achieve early warning and prevention of faults. By grouping similar fault modes through the K-means algorithm and analyzing the spatial relationship and evolution trend between different fault mode groups, targeted fault handling suggestions can be provided for the operation and maintenance personnel, which can not only improve the efficiency of fault handling, but also reduce the risk of secondary faults caused by improper handling.

[0072] The embodiment of the present invention also provides a real-time data acquisition system for a low-voltage distribution box, including: A determination module, configured to respectively determine the spatial association weighting coefficients corresponding to the standardized meteorological load vector, the microenvironment load vector, and the distribution box state load vector according to the historical data of the distribution box; A calculation module, configured to map the standardized meteorological load vector, microenvironment load vector, and distribution box status load vector to three vertices of an equilateral triangle respectively, so as to obtain the spatial feature weights of each load vector; perform a tensor product operation on each load vector and its corresponding spatial correlation weighting coefficient, and use the spatial feature weights to impose regularization processing on the linear combination process to form an environmental load matrix including spatial topological relationships; A prediction module, configured to establish a spatio-temporal evolution early warning model integrating a diffusion convolution operator based on the geometric features of the environmental load matrix to predict abnormal points; An analysis module, configured to perform multi-scale matching degree analysis on the predicted abnormal points and the device aging feature spectrum, and combine the phase angle difference parameters between the vertices of the equilateral triangle to construct a three-dimensional matching space for the composite fault mode and identify the coupling propagation path of potential faults.

[0073] It should be noted that this system corresponds to the above method, and all implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0074] An embodiment of the present invention further provides a computing device, including: a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0075] An embodiment of the present invention further provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is caused to execute the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

Claims

1. A method for real-time data collection of a low-voltage distribution box, characterized in that: The method comprises: According to the historical data of the distribution box, the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, the microenvironmental load vector and the distribution box state load vector are determined respectively; The standardized meteorological load vector, microenvironmental load vector and distribution box state load vector are mapped to the three vertices of an equilateral triangle to obtain the spatial characteristic weight of each load vector; each load vector is fused with its corresponding spatial association weight coefficient, and the linear combination process is regularized using the spatial characteristic weight to form an environmental load matrix containing spatial topological relationships; Based on the geometric characteristics of the environmental load matrix, a spatiotemporal evolution warning model integrating diffusion convolution operators is established to predict abnormal points. A multi-scale matching degree analysis is performed between the predicted abnormal points and the equipment aging characteristic spectrum. The phase angle difference parameters between the vertices of the equilateral triangle are combined to construct a three-dimensional matching space of the composite fault mode and identify the coupled propagation path of potential faults.

2. A method for real-time data collection of a low-voltage distribution box according to claim 1, characterized in that: Before respectively determining the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, the microenvironmental load vector and the distribution box state load vector according to the distribution box historical data, it also includes: Extract key meteorological factors from dynamic meteorological data to construct meteorological load vectors. Key meteorological factors include wind speed, wave height and salt spray deposition rate. Extract key microenvironmental factors from the cabin microenvironmental data and construct the microenvironmental load vector. The key microenvironmental factors include temperature and humidity gradient distribution and condensation film thickness. The key indicators reflecting the electrical performance and insulation status of the distribution box are obtained to construct the distribution box state load vector. The key indicators include the harmonic component of the leakage current and the insulation medium loss factor.

3. A method for real-time data collection of a low-voltage distribution box according to claim 2, characterized in that: According to the historical data of the distribution box, the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, microenvironmental load vector and distribution box state load vector are determined respectively, including: The wind speed, wave height and salt spray deposition rate in the meteorological load vector are calculated and added to their quadratic power values ​​respectively to calculate the comprehensive intensity assessment value of the meteorological load vector; The second power values ​​of the temperature and humidity gradient distribution and the condensation film thickness in the microenvironment load vector are calculated and accumulated to calculate the comprehensive intensity value of the microenvironment load vector; The quadratic power values ​​of the leakage current harmonic component and the insulation medium loss factor in the distribution box state load vector are calculated respectively and accumulated to calculate the comprehensive intensity value of the distribution box state load vector; The comprehensive strength value of the meteorological load vector, the comprehensive strength value of the microenvironmental load vector and the comprehensive strength value of the distribution box state load vector are integrated to obtain the total strength value; According to the comprehensive intensity value and the total intensity value of the meteorological load vector, the spatial correlation weight coefficient of the meteorological load vector is obtained; According to the comprehensive intensity value and the total intensity value of the microenvironment load vector, the spatial correlation weight coefficient of the microenvironment load vector is obtained; According to the comprehensive strength value and the total strength value of the distribution box state load vector, the spatial correlation weighted coefficient of the distribution box state load vector is obtained.

4. A method for real-time data collection of a low-voltage distribution box according to claim 1, characterized in that: The standardized meteorological load vector, microenvironmental load vector and distribution box state load vector are mapped to the three vertices of an equilateral triangle to obtain the spatial characteristic weight of each load vector, including: According to the physical location of the distribution box, the center point of the distribution box installation location is used as the origin of the polar coordinate system, and a direction is determined as the polar axis; the standardized meteorological load vector, microenvironmental load vector and distribution box state load vector are marked as the initial point in the polar coordinate system, and the direction of each load vector represents its direction in the multidimensional space; Using the historical data of the distribution box, set corresponding constraints for each load vector, including the maximum and minimum values ​​of the vector modulus and the vector direction; The rotation matrix is ​​calculated according to the geometric characteristics of the equilateral triangle, and the multi-dimensional tensor product is rotated using the rotation matrix to obtain the position of the load vector mapped to the vertices of the equilateral triangle; The spatial characteristic weight of each load vector is calculated according to the position of the load vector mapped to the vertices of the equilateral triangle.

5. A method for real-time data collection of a low-voltage distribution box according to claim 3 or 4, characterized in that: Each load vector is fused with its corresponding spatial correlation weight coefficient, and the linear combination process is regularized using the spatial feature weight to form an environmental load matrix containing spatial topological relationships, including: The meteorological load vector, the microenvironmental load vector and the distribution box state load vector are respectively multiplied by their corresponding spatial correlation weighting coefficients to obtain weighted load vectors; The weighted load vectors are merged to form a new comprehensive vector; The spatial feature weights are fused with the new comprehensive vector to obtain the environmental load matrix.

6. A method for real-time data collection of a low-voltage distribution box according to claim 5, characterized in that: Based on the geometric characteristics of the environmental load matrix, a spatiotemporal evolution early warning model integrating diffusion convolution operators is established to predict abnormal points, including: Analyze the geometric characteristics of the environmental load matrix, including the shape, size and element distribution of the matrix; Collect various data during the operation of the distribution box, including current, voltage, temperature and humidity data; According to various data during the operation of the distribution box and the geometric characteristics of the environmental load matrix, an early warning model corresponding to the real distribution box is constructed; Integrate sensor data into the early warning model to reflect the operating status of the distribution box in real time; The diffusion convolution operator is integrated into the digital twin model to capture the spatiotemporal evolution characteristics of the data. Through the diffusion convolution operator, the early warning model extracts the spatiotemporal information of the distribution box during operation, including the change trend of current and voltage and the distribution of temperature; The diffusion convolution operator is used to extract the features of the distribution box operation data to obtain the spatiotemporal feature vector, which is then analyzed to identify abnormal points in the operation of the distribution box.

7. A method for real-time data collection of a low-voltage distribution box according to claim 6, characterized in that: The predicted abnormal points are analyzed for multi-scale matching with the equipment aging characteristic spectrum, and the phase angle difference parameters between the vertices of the equilateral triangle are combined to construct a three-dimensional matching space of the composite fault mode and identify the coupling propagation path of potential faults, including: Collect historical operation data of the distribution box, including data in normal and abnormal states; extract the aging characteristic spectrum of the distribution box from the original data through Fourier transform; Calculate the cosine similarity between the predicted abnormal point and the equipment aging characteristic spectrum; Map the three key state parameters of the distribution box to the three vertices of an equilateral triangle and calculate the phase angle difference between the three state parameters; With cosine similarity as the x-axis and y-axis, and the equilateral triangle phase angle difference parameter as the z-axis, each point in the three-dimensional space represents a fault mode; In the three-dimensional matching space, similar fault modes are grouped by clustering, and the spatial relationship and evolution trend between different fault mode groups are analyzed to identify potential fault coupling propagation paths.

8. A real-time data acquisition system for a low-voltage distribution box, characterized in that: The system is used to perform the method according to any one of claims 1 to 7, comprising: A determination module, used to determine the spatial correlation weighting coefficients corresponding to the standardized meteorological load vector, the microenvironmental load vector and the distribution box state load vector respectively according to the distribution box historical data; The calculation module is used to map the standardized meteorological load vector, microenvironmental load vector and distribution box state load vector to the three vertices of an equilateral triangle respectively to obtain the spatial characteristic weight of each load vector; each load vector is merged with its corresponding spatial correlation weight coefficient, and the linear combination process is regularized using the spatial characteristic weight to form an environmental load matrix containing spatial topological relations; The prediction module is used to establish a spatiotemporal evolution warning model integrating diffusion convolution operators based on the geometric characteristics of the environmental load matrix to predict abnormal points; The analysis module is used to perform multi-scale matching analysis between the predicted abnormal points and the equipment aging characteristic spectrum, and to construct a three-dimensional matching space of the composite fault mode by combining the phase angle difference parameters between the vertices of the equilateral triangle to identify the coupled propagation path of the potential fault.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images