Power grid micrometeorological data calibration method, device and equipment and storage medium

Abstract

The invention discloses a power grid micrometeorological data calibration method, device and equipment and a storage medium, and belongs to the technical field of intelligent monitoring of power equipment. The method comprises the steps of obtaining image data of the surface of power grid equipment and micro-meteorological sensor data of an area where the power grid equipment is located, and performing space-time alignment and fusion on the image data and the micro-meteorological sensor data to form an associated data set; performing hierarchical denoising processing on the image data in the associated data set to obtain a preprocessed image; extracting and quantifying morphological state features of the surface coating of the power grid equipment based on the preprocessed image and the micro-meteorological sensor data; dynamically generating a parameterization rule for calibrating the data of the micrometeorological sensor based on the quantized morphological state characteristics; and calibrating the data of the micro-meteorological sensor by using the parameterization rule, and outputting the calibrated data. According to the embodiment of the invention, the accuracy of the micrometeorological monitoring data can be improved.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring technology for power equipment, and in particular to a method, device, equipment and storage medium for calibrating micro-meteorological data of power grids. Background Technology

[0002] Power grid micro-meteorological data refers to local, real-time environmental data that directly affects the safe operation of power grid equipment (such as transmission lines and insulators). Currently, related monitoring mainly relies on sensor networks deployed in towers or substations, such as anemometers, humidity sensors, and image monitoring devices. Common methods include calibration techniques based on multi-sensor data fusion, which aim to improve the reliability of data from individual sensors through algorithmic processing.

[0003] However, in practical applications, this type of method still has obvious limitations: most data fusion algorithms rely too much on the mathematical and statistical relationships between sensor readings, and fail to fully combine the actual microclimate environment and physical state changes of the equipment, resulting in calibration results that often deviate from the actual insulation degradation process and cannot accurately reflect the actual operating risks of the equipment. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, device, and storage medium for calibrating power grid micro-meteorological data, which can effectively improve the accuracy of micro-meteorological monitoring data.

[0005] To achieve the above objectives, a first aspect of this application provides a method for calibrating power grid micrometeorological data, comprising: Acquire image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and perform spatiotemporal alignment and fusion of the image data and the micro-meteorological sensor data to form a related dataset; The image data in the associated dataset is subjected to hierarchical denoising processing to obtain a preprocessed image; Based on the preprocessed image and the micro-meteorological sensor data, the morphological state characteristics of the surface coating of the power grid equipment are extracted and quantified. Based on the quantized morphological state characteristics, parameterization rules for calibrating the micro-meteorological sensor data are dynamically generated. The parameterization rules are applied to calibrate the micro-weather sensor data, and the calibrated data is output.

[0006] Compared with existing technologies, the power grid micro-meteorological data calibration method provided in this application has the following advantages: This method simultaneously acquires surface image data of power grid equipment and micro-meteorological sensor data of the area, performs spatiotemporal alignment and fusion, and constructs a correlation mapping foundation between equipment surface status and environmental meteorological data, effectively avoiding calibration deviations caused by the one-sidedness of single-type data information. Through layered denoising processing of image data, the integrity and accuracy of effective information related to the equipment surface coating in the image are ensured, providing high-quality data support for the accurate extraction of subsequent morphological state features. Based on the collaborative extraction and quantification of coating morphological state features from preprocessed images and micro-meteorological sensor data, an intrinsic correlation between the coating physical state and meteorological sensor data is established. This allows for the dynamic generation of parameterized rules adapted to the current coating state, overcoming the limitations of fixed calibration modes in adapting to complex and variable coating environments. Finally, through these parameterized rules, targeted calibration of micro-meteorological sensor data is achieved, significantly improving the measurement accuracy and reliability of micro-meteorological data, and providing more accurate meteorological data support for scenarios such as power grid equipment operation status assessment and fault early warning.

[0007] In some embodiments, acquiring image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and then spatiotemporally aligning and fusing the image data and the micro-meteorological sensor data to form a correlated dataset, includes: Construct a spatial coordinate system for the surface of power grid equipment; Acquire image data of the surface of the power grid equipment, the image data including high-resolution images and infrared temperature distribution maps; At the corresponding location points in the spatial coordinate system, micro-meteorological sensor data are collected synchronously. The micro-meteorological sensor data includes at least one of the following: dust particle density, dust particle size distribution, rainfall humidity distribution, mud film surface roughness, and mud film surface adhesion. Based on timestamps and spatial coordinates, the micro-meteorological sensor data is used as an environmental label and fused with the high-resolution image and infrared temperature distribution map to obtain the associated dataset.

[0008] In some embodiments, performing hierarchical denoising on the image data in the associated dataset to obtain a preprocessed image includes: The low-frequency background noise formed by uniform dust deposition in the image data is identified and filtered out. An edge-preserving filter is used to remove high-frequency stripe noise caused by rainfall erosion from the image data; Feature enhancement transformation is used to preserve and highlight the texture and spectral features in the image data that are related to the moisture content or adhesion of the coating.

[0009] In some embodiments, the surface coating of the power grid equipment is a mud film or an ice layer, and the morphological characteristics are cracked characteristics.

[0010] In some embodiments, the cracking state characteristics include at least one of crack density, width, depth, morphological classification, and network connectivity; the step of extracting and quantifying the morphological state characteristics of the surface coating of the power grid equipment based on the preprocessed image and the micro-meteorological sensor data includes: Extract the edge contour of the crack from the preprocessed image and calculate the crack width; The relative depth of the cracks was assessed by combining the dust particle size data from the micro-meteorological sensor data. The number and branching of cracks per unit area are statistically analyzed to determine crack density and crack morphology classification.

[0011] In some embodiments, the step of dynamically generating parameterization rules for calibrating the micro-weather sensor data based on quantized morphological state features includes: A correlation model is established between the cracking state characteristics and the data error of the micro-meteorological sensor. The correlation model is a statistical model or a graphical model constructed based on the cracking network connectivity characteristics and near-surface humidity gradient data. Based on the cracking state characteristics at the current moment, the sensor data error compensation coefficient is calculated and processed through the correlation model to obtain the error compensation coefficient that constitutes the parameterization rule.

[0012] In some embodiments, after extracting and quantifying the morphological state features of the surface coating of the power grid equipment, the method further includes: The acquisition frequency of the image data or the micro-meteorological sensor data is dynamically adjusted based on the quantified morphological state characteristics or their rate of change.

[0013] To achieve the above objectives, a second aspect of this application provides a power grid micro-meteorological data calibration device, the device comprising: The fusion module is used to acquire image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and to perform spatiotemporal alignment and fusion of the image data and the micro-meteorological sensor data to form a related dataset. The denoising module is used to perform layered denoising processing on the image data in the associated dataset to obtain a preprocessed image; The quantization module is used to extract and quantify the morphological state features of the surface coating of the power grid equipment based on the preprocessed image and the micro-meteorological sensor data. The generation module is used to dynamically generate parameterization rules for calibrating the micro-meteorological sensor data based on the quantized morphological state features. The calibration module is used to calibrate the micro-weather sensor data by applying the parameterization rules and output the calibrated data.

[0014] To achieve the above objectives, a third aspect of this application provides an electronic device, the electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method described in the first aspect.

[0015] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when executed, controls the device containing the computer-readable storage medium to perform the method described in the first aspect. Attached Figure Description

[0016] Figure 1 This is a flowchart of a power grid micro-meteorological data calibration method provided in an embodiment of this application; Figure 2 This is a schematic diagram of a power grid micro-meteorological data calibration device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0018] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0019] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0021] In the field of power grid operation and maintenance, the micro-meteorological environment is one of the key factors affecting the safe and stable operation of power grid equipment. Micro-meteorological parameters such as dust, rainfall, and humidity around power grid equipment are directly related to the formation and evolution of the coating on the equipment surface, which in turn affects the equipment's insulation performance, mechanical strength, and other indicators. Therefore, accurate acquisition of micro-meteorological data is of great significance for power grid equipment condition assessment, fault early warning, and operation and maintenance decision-making.

[0022] Power grid micro-meteorological data refers to local, real-time environmental data that directly affects the safe operation of power grid equipment (such as transmission lines and insulators). Currently, related monitoring mainly relies on sensor networks deployed in towers or substations, such as anemometers, humidity sensors, and image monitoring devices. Common methods include calibration techniques based on multi-sensor data fusion, which aim to improve the reliability of data from individual sensors through algorithmic processing.

[0023] However, in practical applications, this type of method still has obvious limitations: most data fusion algorithms rely too much on the mathematical and statistical relationships between sensor readings, and fail to fully combine the actual microclimate environment and physical state changes of the equipment, resulting in calibration results that often deviate from the actual insulation degradation process and cannot accurately reflect the actual operating risks of the equipment.

[0024] Based on this, embodiments of this application provide a method, apparatus, device, and storage medium for calibrating power grid micro-meteorological data, which can effectively improve the accuracy of micro-meteorological monitoring data.

[0025] Please see Figure 1 , Figure 1 This is an optional flowchart of the power grid micro-meteorological data calibration method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S105.

[0026] Step S101: Obtain image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located; perform spatiotemporal alignment and fusion of the image data and micro-meteorological sensor data to form a related dataset. Step S102: Perform layered denoising on the image data in the associated dataset to obtain a preprocessed image; Step S103: Based on the preprocessed image and micro-weather sensor data, extract and quantify the morphological state characteristics of the coating on the device surface; Step S104: Based on the quantized morphological state characteristics, dynamically generate parameterization rules for calibrating micro-meteorological sensor data. Step S105: Apply parameterization rules to calibrate the micro-weather sensor data and output the calibrated data.

[0027] Steps S101 to S105 of this embodiment illustrate the power grid micro-meteorological data calibration method. By simultaneously acquiring surface image data of power grid equipment and micro-meteorological sensor data of the area, and performing spatiotemporal alignment and fusion, a correlation mapping foundation between equipment surface status and environmental meteorological data is constructed, effectively avoiding calibration deviations caused by the one-sidedness of single-type data information. Through layered denoising processing of image data, the integrity and accuracy of effective information related to equipment surface coating in the image are ensured, providing high-quality data support for the accurate extraction of subsequent morphological state features. Based on the collaborative extraction and quantification of coating morphological state features from preprocessed images and micro-meteorological sensor data, an intrinsic correlation between coating physical state and meteorological sensor data is established. Then, parameterized rules adapted to the current coating state are dynamically generated, breaking through the limitations of fixed calibration modes in adapting to complex and variable coating environments. Finally, targeted calibration of micro-meteorological sensor data is achieved through these parameterized rules, significantly improving the measurement accuracy and reliability of micro-meteorological data, and providing more accurate meteorological data support for scenarios such as power grid equipment operation status assessment and fault early warning.

[0028] In step S101 of some embodiments, the image data of the power grid equipment surface refers to visual data captured on the outer surface of the power grid equipment (such as insulators, conductors, etc.), including high-resolution visible light images and infrared temperature distribution maps. The high-resolution visible light images are mainly used to reflect the intuitive morphology of the surface coating, while the infrared temperature distribution maps are mainly used to reflect differences in surface temperature distribution. The micro-meteorological sensor data refers to data deployed in the area where the power grid equipment is located, used to collect parameters of the surrounding micro-environment, including at least one of dust particle density, dust particle size distribution, rainfall humidity distribution, mud film surface roughness, and mud film surface adhesion. Spatiotemporal alignment and fusion can be a processing method that maintains synchronization in time based on the timestamps of data acquisition and establishes a correlation in space based on the corresponding location points in the spatial coordinate system of the power grid equipment surface, integrating the two types of data. The associated dataset can be an integrated dataset containing the power grid equipment surface image data and the corresponding location and time micro-meteorological sensor data after spatiotemporal alignment and fusion, where the micro-meteorological sensor data serves as environmental annotation and is associated with the image data.

[0029] In some embodiments, image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located are acquired. The image data and micro-meteorological sensor data are spatiotemporally aligned and fused to form a correlated dataset, including: Construct a spatial coordinate system for the surface of power grid equipment; Acquire image data of the surface of power grid equipment, including high-resolution images and infrared temperature distribution maps; At the corresponding location points in the spatial coordinate system, micro-meteorological sensor data are collected synchronously. The micro-meteorological sensor data includes at least one of the following: dust particle density, dust particle size distribution, rainfall humidity distribution, mud film surface roughness, and mud film surface adhesion. Based on timestamps and spatial coordinates, micro-meteorological sensor data is fused with high-resolution images and infrared temperature distribution maps as environmental annotations to obtain a correlated dataset.

[0030] Specifically, firstly, taking the power grid insulator as the target power grid equipment, a spatial coordinate system is constructed based on its physical structural characteristics. The center of the bottom flange of the insulator string is selected as the origin of the coordinate system. The direction perpendicular to the ground is set as the Z-axis, the direction along the extension of the insulator string is set as the Y-axis, and the X-axis is determined according to the right-hand rule. A unique spatial identifier containing three-dimensional coordinate values ​​and the corresponding insulator piece number is set for each sampling point to ensure that subsequent multi-source data accurately corresponds to the specific physical location.

[0031] Subsequently, an infrared thermal imager and a visible light camera, which are fixed in a constant relative position by a rigid bracket, are used to acquire image data of the equipment surface. The infrared thermal imager is equipped with an uncooled focal plane array detector with a working wavelength of 8-14 micrometers. It scans the surface of the insulator to obtain the temperature value corresponding to each pixel and a temperature distribution map that reflects the difference in thermal radiation after the mud film is damp. The visible light camera uses a high-speed CMOS sensor to acquire RGB images with a resolution of 4096×3072, which can capture the texture details and color changes of the mud film surface, i.e., high-resolution images.

[0032] Next, micro-meteorological sensor data are simultaneously collected at the corresponding sampling points in the spatial coordinate system. A laser particle size analyzer based on the Mie scattering principle, with an emission wavelength of 650 nm and a measurement range covering 0.1 to 500 micrometers, can be selected. The particle size is calculated by detecting the intensity distribution of scattered light through the emitted laser beam passing through the dust environment. At the same time, the particle concentration is determined by calculating the particle mass per unit volume based on the attenuation of transmitted light intensity using the optical concentration method. In conjunction with an ultrasonic wind speed sensor with a sampling frequency of 10 times per second, the wind direction and speed are recorded. Combining these wind direction and speed data, the dust deposition rate is calculated using a particle deposition formula that considers Stokes's law of settling and turbulent diffusion effects. For fine particles with a diameter of less than 50 micrometers, the deposition rate is directly calculated using Stokes's law. ,in, For particle settling velocity, The density of sand and dust particles, air density, It is the acceleration due to gravity. The particle size of the sand and dust is... The aerodynamic viscosity is used; for larger particles with a diameter of 50 micrometers or more, the Reynolds number is introduced for correction, and the corrected settling rate is... ,in The particle deposition Reynolds number is used. Combined with wind speed data, the particle deposition flux at different heights is calculated using a turbulent deposition model to form vertical particle stratification distribution data, reflecting the accumulation characteristics of dust at different heights of the insulator.

[0033] Subsequently, a capacitive humidity sensor array, employing polyimide film as the moisture-sensing medium and arranged at equal intervals on the insulator surface with a measuring point every 5 cm vertically to form a humidity gradient acquisition network, was used to collect the relative humidity values ​​of each layer. When the humidity difference between adjacent layers exceeded a preset threshold, an atomic force microscope (AFM) with a probe tip radius of less than 10 nm was activated to scan the mud film surface in contact mode. During the scanning process, the probe moved along the surface in a constant force mode, and the surface height change was calculated by the cantilever beam deflection, thereby obtaining two parameters: arithmetic mean roughness Ra and root mean square roughness Rq. The study collected roughness values ​​and stored microscopic morphology data in the form of three-dimensional height maps with nanometer-level resolution to identify microcracks and pore structures. At the same time, a piezoelectric force sensor with a quartz crystal as the sensing element, a measurement range of 0 to 100 Newtons, and a resolution of 0.01 Newtons was used to measure the mud film adhesion strength at the corresponding location. During the measurement, the sensor probe was in contact with the mud film surface and a vertical tensile force was applied. The peak force at the moment of mud film detachment was recorded to reflect the bonding strength between the mud film and the insulator substrate. It was also found that the adhesion strength data was correlated with the surface roughness, that is, rough surfaces usually have higher adhesion strength.

[0034] Finally, spatiotemporal alignment and fusion processing are performed, with strict control over timestamp alignment accuracy within 10 milliseconds and spatial coordinate matching accuracy within 2 millimeters. After aligning the sensor data and image data according to the acquisition timestamps, a Kalman filter algorithm is used for multi-source data fusion. This algorithm defines the system state vector as a vector containing multi-dimensional parameters such as position, temperature, humidity, and particle concentration. Through the prediction step, the state at the next moment is estimated based on the system dynamic model. Then, through the update step, the prediction result is corrected by combining the sensor measurements. Through iterative calculation, sensor noise is effectively suppressed and the optimal state estimate is provided. The fused data retains the effective information of each sensor while eliminating the influence of measurement errors and environmental interference. Subsequently, the numerical sensor data is mapped to the corresponding areas of the image to complete environmental annotation. Temperature values ​​are presented in the visible light image in the form of pseudo-color overlay, humidity data is displayed in the form of contour lines, and particle concentration is represented by transparency. This multi-layer annotation method ensures that each image area contains complete environmental parameter information, and the final output is a correlated dataset with rich environmental annotations.

[0035] In step S102 of some embodiments, the layered denoising process can be a processing method that employs targeted denoising strategies in layers according to the source and characteristics of noise in the image (different frequencies, different causes), including three layers: low-frequency background noise filtering, high-frequency stripe noise removal, and effective feature preservation. The preprocessed image can be image data that has undergone layered denoising processing, after removing invalid noise and retaining effective information. Low-frequency background noise can be interference signals with low frequency and uniform distribution generated by sand and dust uniformly deposited on the surface of the equipment or during image acquisition. High-frequency stripe noise can be interference signals with high frequency and stripe pattern generated by rain washing the surface of the equipment or during image transmission.

[0036] In some embodiments, hierarchical denoising processing is performed on image data in the associated dataset to obtain a preprocessed image, including: The low-frequency background noise formed by uniform dust deposition in image data is identified and filtered out. An edge-preserving filter is used to remove high-frequency stripe noise caused by rainfall erosion in image data; Feature enhancement transformation is used to preserve and highlight the texture and spectral features in image data that are related to coating moisture content or adhesion.

[0037] Specifically, the first step involves identifying and filtering low-frequency background noise formed by uniform dust deposition in image data. This is achieved by performing a Fast Fourier Transform (FFT) on the images in the associated dataset, converting the spatial domain image into a frequency domain representation. Utilizing the characteristic that low-frequency components in the spectrum are concentrated in the center, corresponding to slowly changing regions in the image, specific energy-concentrated frequency bands formed by changes in dust deposition thickness are identified in the spectrum. Based on the principle of optical scattering, there is a positive correlation between low-frequency amplitude and dust deposition thickness—the thicker the dust deposition layer, the stronger the incident light scattering, the more obvious the image grayscale gradient, and the larger the low-frequency amplitude in the frequency domain. Through statistical analysis, a mapping relationship between amplitude and thickness is established, converting the spectral amplitude into deposition thickness values ​​and constructing a thickness distribution map. Deposition noise regions with thicknesses exceeding a threshold are marked, laying the foundation for subsequent accurate noise filtering.

[0038] Next, an edge-preserving filter is used to remove high-frequency stripe noise caused by rainfall erosion in the image data. A bilateral filter is chosen as the edge-preserving filter, which balances spatial neighborhood and gray-level similarity to achieve a balance between noise removal and edge preservation. The spatial neighborhood weight is calculated using a Gaussian function; pixels closer to the center pixel have higher weights, and the standard deviation of the Gaussian function controls the weight decay rate. The gray-level similarity weight is calculated based on the gray-level difference between the center pixel and its neighbors; the smaller the difference, the higher the weight. The stripe noise caused by rainfall erosion manifests as local gray-level abrupt changes. Bilateral filtering can smooth these abrupt changes while effectively maintaining the sharpness of the true edges of the mud film, thus removing high-frequency stripe noise and obtaining a preliminary denoised image.

[0039] Finally, feature enhancement transformation was used to preserve and highlight the texture and spectral features related to the moisture content or adhesion of the coating in the image data. On one hand, the Sobel operator was used to calculate the texture gradient value of the preliminary denoised image—by detecting grayscale changes through horizontal and vertical convolution kernels, the gradient magnitude reflecting the complexity of local textures was obtained, where high gradient value regions correspond to texture-rich regions related to mud film adhesion. Anisotropic diffusion filtering was applied to these high gradient value regions to enhance details. This filter adaptively adjusts the diffusion coefficient according to the local features of the image, suppressing noise while preserving detailed textures, and also smoothing along the water infiltration boundary direction and preserving edges in the vertical direction, thus completely preserving the water infiltration morphology. On the other hand, discrete wavelet transform was performed on the texture-enhanced image, decomposing it into approximate coefficients containing the main structure and detail coefficients containing edge textures. The effective coefficients were retained by analyzing the correlation between sub-band coefficients and moisture content. At the same time, the RGB image was converted to the HSV color space, and a hue retention interval was set to filter out interference and retain the hue (H channel) and brightness (V channel) changes after the mud film absorbed water. Finally, inverse wavelet transform is performed on the processed coefficients to reconstruct the image, outputting a preprocessed image that removes noise while fully preserving the characteristics related to coating moisture content and adhesion.

[0040] This embodiment first identifies and filters out low-frequency background noise caused by uniform dust deposition in the image. Then, it uses an edge-preserving filter to remove high-frequency stripe noise caused by rain erosion. Finally, it uses feature enhancement transformation to retain and highlight effective texture and spectral features related to the coating, resulting in a preprocessed image. This process achieves precise layered noise removal while avoiding the loss of effective information, ensuring the integrity and accuracy of the image data, and providing high-quality data support for subsequent morphological feature extraction.

[0041] In step S103 of some embodiments, the equipment surface coating refers to the material layer attached to the surface of the power grid equipment, specifically a mud film (formed by a mixture of sand, rainwater, etc.) or an ice layer (formed by rainwater freezing in a low-temperature environment). Morphological state characteristics refer to characteristic parameters reflecting the physical morphology of the equipment surface coating. In this step, they specifically refer to cracking state characteristics, including at least one of the following: crack density (number of cracks per unit area), width (distance between the two edges of the crack), depth (the extent to which the crack extends perpendicular to the surface), morphological classification (such as linear, mesh, etc.), and network connectivity (the degree of connectivity between cracks).

[0042] In some embodiments, based on preprocessed images and micro-meteorological sensor data, the morphological state characteristics of the surface coating of power grid equipment are extracted and quantified, including: Extract the edge contour of the crack from the preprocessed image and calculate the crack width; The relative depth of the cracks was assessed by combining dust particle size data from micro-meteorological sensor data. The number and branching of cracks per unit area are statistically analyzed to determine crack density and crack morphology classification.

[0043] Specifically, firstly, the edge contours of the cracks are extracted from the preprocessed image, and the crack width is calculated. The Canny operator is used as the edge detection operator to extract the crack contours. This operator first smooths the image using Gaussian filtering, then calculates the gradient intensity and direction, performs non-maximum suppression to preserve edge details, and finally obtains the complete crack contour through double threshold detection and edge connection. The extracted edge contours are stored as a sequence of pixel coordinates, with each crack segment consisting of consecutive edge points. The crack width is calculated by traversing the corresponding edge points on both sides of the crack. For each crack segment, its left and right edge lines are first identified, and then the Euclidean distance between the corresponding edge points is calculated—this distance is obtained by the square root of the sum of the squares of the differences in the horizontal and vertical coordinates of adjacent edge points, which is the crack width at the corresponding position. The width value is calculated every 5 pixels along the crack extension direction, ultimately forming a width distribution data sequence.

[0044] Secondly, by combining dust particle size data from micro-meteorological sensor data, the relative depth of the cracks is evaluated and the depth distribution characteristics are output. Specifically, a residual network is first used to extract crack depth features. Local image patches of the crack area are input into the network, which retains the original image information through skip connections while learning depth-related feature representations. Then, the rate of change of pixel gray values ​​in the crack area along the vertical direction is calculated (by the difference in gray values ​​between adjacent rows of pixels). The larger the rate of change, the more drastic the change in depth. Furthermore, there is a negative correlation between the crack depth and the surface gray value; the deeper the crack, the less light is reflected from the bottom, and the lower the gray value. Based on this, the median particle size (representing the typical size of dust particles in the area) is extracted from the dust particle size distribution data. The crack depth is divided by this median particle size for dimensionless processing to eliminate the influence of dimensions. The resulting dimensionless ratio reflects the degree of cracking relative to the particle scale (a ratio greater than 10 indicates a deep crack, 3-10 indicates a moderate crack, and less than 3 indicates a shallow crack). Finally, the depth distribution characteristics are output.

[0045] Finally, the number and branching of cracks per unit area are statistically analyzed to determine crack density and crack morphology classification, thereby comprehensively determining the mud film cracking state. Crack density is statistically analyzed using a grid division method, dividing the image into 100×100 pixel grid units. The ratio of the number of crack pixels in each unit to the total number of pixels in the unit is the crack density of the corresponding unit. If the crack density exceeds a preset threshold, the crack morphology is further determined based on the number of crack branches and the length of the main trunk: dendritic cracking is characterized by multiple branches extending from the main crack, with more than 5 branches and branch angles between 30 and 60 degrees; mesh cracking is characterized by cracks intersecting to form closed areas, with more than 10 intersection points; locally cracked cracks have a main trunk length less than one-third of the image width, while through-cracks span the entire monitoring area. In summary, by integrating information on crack width, depth distribution, density, and morphology classification, the mud film cracking state is finally determined, providing a quantitative basis for subsequent dynamic generation of parameterized rules.

[0046] This embodiment extracts the edge contours of cracks from preprocessed images and calculates the crack width using image measurement algorithms. It then indirectly assesses the relative depth of cracks by combining dust particle size data from micrometeorological sensor data. Finally, it statistically analyzes the number and branching of cracks per unit area to determine crack density and crack morphology classification, thereby completing the extraction and quantification of the characteristics of the overburden cracking state. This process establishes an intrinsic correlation between the physical state of the overburden and micrometeorological data, providing a precise quantitative basis for subsequent parameterization rule generation.

[0047] In some embodiments, after extracting and quantifying the morphological state features of the surface coating of the power grid equipment, the method further includes: Based on the quantified morphological state characteristics or their rate of change, the acquisition frequency of image data or micro-meteorological sensor data is dynamically adjusted.

[0048] Specifically, the degree of cracking is first classified based on the quantified state of mud film cracking. The classification of cracking degree is based on the comprehensive characteristic parameters of the cracks, combined with the actual failure mode of the mud film in the power grid equipment and the differences in its impact on insulation performance. The standards are set as follows: mild cracking corresponds to cracks with a width of less than 2 mm and a depth of less than 5 mm; moderate cracking is cracks with a width between 2 and 5 mm and a depth between 5 and 15 mm; and severe cracking refers to cracks with a width exceeding 5 mm or a depth exceeding 15 mm.

[0049] Secondly, environmental data around the crack was collected, and environmental dynamic change indicators were calculated. Humidity gradient data was acquired using an array of humidity sensors deployed around the crack. The sensors were arranged in a ring around the crack with a radius of 50 cm, with a measuring point every 10 cm. The rate of humidity change was calculated by dividing the difference between two consecutive humidity values ​​by the time interval. When this rate of change exceeded 5% per hour, it indicated drastic fluctuations in environmental humidity, increasing the risk of mud film cracking. The environmental dynamic change indicator comprehensively considered both the rate of humidity change and the magnitude of temperature change, using a weighted summation method to convert it into a single numerical indicator, directly reflecting the impact of environmental dynamic changes on cracking.

[0050] Furthermore, image difference analysis is used to detect changes in the crack edge and determine whether the rapid cracking phase has begun. This process includes two steps: image registration and difference calculation. First, a feature point matching algorithm is used to register the current and previous images to eliminate interference from minor changes in the shooting angle. Then, the gray values ​​of corresponding pixels in the registered images are subtracted to obtain the difference image. Non-zero pixels in the difference image are mainly distributed in the crack edge change area. By statistically analyzing the spatial distribution range of these non-zero pixels, the crack edge expansion distance is calculated and divided by the time interval to obtain the edge expansion speed. When the edge expansion speed exceeds a preset threshold of 0.5 mm per hour, the system is determined to have entered the rapid cracking phase, at which point a high-frequency acquisition mode needs to be activated.

[0051] Finally, a data acquisition frequency adjustment scheme was determined and output. The sampling strategy in high-frequency acquisition mode employs an adaptive adjustment mechanism. The initial sampling interval is set at 5 minutes, shortened to 100 seconds after entering high-frequency mode. The system continuously monitors the stability of the cracking state. If the rate of change of crack characteristic parameters is lower than a set threshold within 10 consecutive sampling cycles, the sampling interval is extended by 50% and gradually restored to the initial interval. This ensures data density during the rapid cracking period while avoiding resource waste during the stable period. Furthermore, the adjustment scheme incorporates weather forecast information. When rainfall is forecast within the next 24 hours, the system enters an early warning state, increasing the sampling frequency by one level to prepare for potential rapid cracking. Through these dynamic adjustments, the acquisition frequency is ensured to adapt to the mud film cracking state and environmental changes, providing high-quality data support for the accurate generation of subsequent parameterized rules.

[0052] In step S104 of some embodiments, the parameterization rule refers to a standardized rule used to correct the deviation of micro-weather sensor data, with a quantified error compensation coefficient as the main parameter. The correlation model refers to a model that establishes the correspondence between crack state characteristics and micro-weather sensor data errors, specifically a statistical model or graphical model constructed based on the crack network connectivity characteristics and near-surface humidity gradient data.

[0053] In some embodiments, parameterization rules for calibrating micrometeorological sensor data are dynamically generated based on quantized morphological state characteristics, including: Establish a correlation model between cracking state characteristics and micro-meteorological sensor data errors. The correlation model is a statistical model or a graphical model constructed based on the cracking network connectivity characteristics and near-surface humidity gradient data. Based on the cracking state characteristics at the current moment, the error compensation coefficient of the sensor data is calculated and processed through the correlation model to obtain the error compensation coefficient that constitutes the parameterized rule.

[0054] Specifically, firstly, data acquisition optimization and multi-source data fusion were completed to lay the foundation for the construction of the correlation model. On the one hand, the sensor sampling mode and data caching mechanism were configured according to the frequency parameters in the adjustment plan: the frequency parameters were divided into three levels: basic sampling frequency, fast response frequency, and steady-state monitoring frequency. The basic sampling frequency was set to once every 5 minutes, the fast response frequency was increased to once every 30 seconds, and the steady-state monitoring frequency was reduced to once every 15 minutes; the data cache adopted a circular buffer structure with a capacity of the most recent 100 sampling points. When new data arrived, it automatically overwrote the oldest data to ensure data integrity during critical periods. On the other hand, the data on rainfall erosion traces, mud film moisture content, and surface adhesion within the power grid area are integrated: erosion traces are extracted from post-rainfall images, and the dominant flow direction in each local area is extracted using a gradient direction histogram algorithm. The consistency of flow direction is statistically analyzed as an indicator of erosion intensity. Then, a connected component labeling algorithm is used to identify continuous erosion-affected areas with gray values ​​below a threshold, generating a pseudo-color erosion intensity distribution map, where red represents strong erosion areas, yellow represents moderate erosion areas, and green represents slight erosion areas. The gradient of moisture content at different depths of the mud film is obtained using a dielectric constant measurement method. The near-surface humidity gradient data (i.e., the rate of change of humidity from the mud film surface to the sensitive area 20 mm below) is obtained through layered acquisition using a capacitive sensor. The sensor probe starts from the mud film surface and measures the dielectric constant value every 2 mm (the dielectric constant of water is about 80, and that of dry soil is about 3-5). After estimating the percentage of moisture content at each depth, the ratio of the humidity difference between adjacent depth measurement points to the depth interval is calculated to obtain the near-surface humidity gradient value. At the same time, a peel test method is used with a piezoelectric sensor to record the force value at the moment the mud film separates from the substrate to obtain adhesion strength data.

[0055] Based on this, the moisture content range of 0-100% was divided into 10 intervals of 10%, and the adhesion strength range of 0-50 Newtons was divided into 10 intervals of 5 Newtons. A two-dimensional correlation matrix was constructed with moisture content as the row index and adhesion strength as the column index. The matrix elements recorded the frequency of occurrence of the corresponding combinations. Statistical analysis showed that the adhesion strength reached its peak when the moisture content was 20-40%, and dropped sharply after exceeding 60%. This correlation provides a quantitative basis for subsequent crack analysis.

[0056] Secondly, a correlation model is established, which is a combination of a statistical model and a graph model based on the connectivity characteristics of the crack network and near-surface humidity gradient data. First, the connectivity of the crack network is analyzed using a shortest path algorithm: the crack network is modeled based on a graph structure, with each crack segment as a node and the spatial distance between cracks as the edge weight. Dijkstra's algorithm is used to calculate the shortest connecting path between any two crack segments (the path length takes into account both spatial distance and crack width weights, with wider cracks more likely to form connections). When the path length is less than a set threshold of 80% of the current monitoring area width, these crack segments are determined to belong to the same connected region. If a connecting path exists from one end of the monitoring area to the other, it is determined to be a through crack. Simultaneously, the evolution of crack morphology from isolated to connected is recorded, forming a morphological transformation sequence. Each state in the sequence includes a timestamp, number of cracks, average width, maximum depth, and connectivity index. By comparing the differences between adjacent states, key turning points in crack development can be identified (such as the critical moment when an isolated crack transforms into a network of cracks), and the environmental conditions corresponding to these turning points are recorded in detail. Finally, by combining the extracted crack network connectivity features with near-surface humidity gradient data, a correlation model combining statistical and graphical models was constructed, which accurately depicts the intrinsic relationship between crack state characteristics and sensor data errors.

[0057] Next, based on the quantitative morphological state characteristics at the current moment, the error compensation coefficient is calculated using an association model. First, a feature vector of the current cracking scene is extracted from the morphological transformation sequence. This vector integrates four dimensions: crack density (calculated by the total length of cracks per unit area), average width (the arithmetic mean of all crack widths), connectivity (the ratio of the number of cracks in the largest connected region to the total number of cracks), and branch number (the number of secondary cracks extending from each main crack), comprehensively describing the characteristics of the current cracking scene. Then, a calibration experiment method is used to calculate the error compensation coefficient: standard mud film samples with known physical parameters are prepared in a laboratory environment. Measurements are performed using sensors, and the difference between sensor readings and standard values ​​is recorded. The linear compensation coefficient and bias are obtained through least squares fitting, establishing a linear mapping relationship between sensor readings and actual physical quantities, thus obtaining the error compensation coefficient adapted to the current scene.

[0058] Finally, the error compensation coefficients are integrated to generate and store parameterized rules. These rules are stored in a lookup table format, containing key information such as sensor type, measurement range, compensation coefficient, and validity period. Specifically, they cover error compensation parameters for various sensors: for example, a dust particle density compensation coefficient of 1.08 and an offset of -0.5; Gaussian filtering parameters for mud film surface roughness data, i.e., a filter kernel size of 5×5 and a standard deviation of 1.2; and a linear regression fitting formula for crack depth and roughness, i.e., roughness value = 0.8 × crack depth + 0.3. The rules also possess scenario adaptability, allowing for scenarios where the crack density is greater than 5 cracks / cm². 2When high sensitivity parameters are enabled, the frequency of compensation coefficient correction is increased when connectivity is greater than 60%. The system can automatically load the corresponding parameter set according to the current cracking scene characteristics to achieve real-time correction of micro-meteorological sensor data.

[0059] This process enables dynamic adaptation of parameterized rules, overcoming the limitations of fixed calibration modes in adapting to complex and variable coating environments, and ensuring that parameterized rules match the real-time environmental conditions.

[0060] In step S105 of some embodiments, the calibrated data refers to the micro-meteorological data with higher accuracy and better fit to the actual environment obtained after correcting the deviation of the original micro-meteorological sensor data through parameterization rules.

[0061] Specifically, the original data of dust particle density is first calibrated based on the linear compensation coefficient and bias determined by laboratory calibration in the parameterization rules. The calibration process involves comparative testing with standard particle density samples. After recording the correspondence between the original sensor readings and the standard values, the least squares method is used to fit the linear relationship, where the slope is the compensation coefficient and the intercept is the bias. In practical applications, the linear transformation is completed by multiplying the original density value by the compensation coefficient of 1.08 and adding the bias of -0.5 to obtain the corrected particle density data.

[0062] Based on the corrected particle density data, noise was processed on the mud film surface roughness measurements according to the Gaussian filtering parameters in the parameterization rules. A 5×5 filter kernel was used to perform convolution operations. The filter kernel followed a two-dimensional Gaussian distribution design, with the largest weight at the center and gradually decreasing towards the edges, smoothing random noise while preserving roughness characteristics and eliminating errors caused by sensor jitter and environmental interference. Subsequently, a depth-roughness linear correlation model was established according to the parameterization rules. A scatter plot was drawn with crack depth as the abscissa and the corresponding roughness value as the ordinate. The study confirmed that crack depth and bottom roughness were positively correlated. Based on this, a straight line was fitted using the least squares linear regression method. The measured roughness values ​​were standardized to a uniform scale according to the fitted equation to obtain standardized roughness. The data is processed as follows: Next, for the crack edge contour associated with the standardized roughness data, the curvature value of the edge points is calculated to filter out abnormal data. Specifically, a vector is formed by the two adjacent points before and after each edge point, and the curvature value is obtained by calculating the cosine of the angle between the two vectors. The larger the angle change, the higher the curvature value, and the greater the degree of edge curvature. When the difference between adjacent curvature values ​​exceeds 0.3 radians, it is identified as an abnormal abrupt change point. These abnormal points are mostly caused by measurement errors or edge recognition errors and are removed. Finally, the calibrated particle density and standardized roughness data, combined with crack depth, edge curvature data, and the timestamps and spatial coordinates of each monitoring point, are integrated in a structured format to form a dataset containing six fields. The final output is the complete calibrated monitoring data, providing reliable data support for subsequent crack early warning.

[0063] This embodiment applies parameterization rules to the raw micro-meteorological sensor data, and uses an error correction algorithm to specifically calibrate the data deviations, ultimately outputting the calibrated data. This process directly improves the measurement accuracy and reliability of micro-meteorological data, providing accurate data support for the refined operation and maintenance of power grid equipment.

[0064] This application deeply integrates surface image data of power grid equipment with micro-meteorological sensor data. Layered denoising ensures data quality, and the correlation between environmental and data deviations is established through quantification of coating cracking characteristics. Adaptive parameterized rules are dynamically generated, ultimately achieving accurate calibration of the micro-meteorological sensor data. The entire process fully considers the dynamic changes in the complex outdoor environment of the power grid and the influence of equipment surface coatings, effectively solving the problems of poor adaptability and large data deviations in traditional fixed calibration modes. This significantly improves the reliability of micro-meteorological data, providing solid data support for power grid equipment operation status assessment, fault early warning, and maintenance decisions.

[0065] In some embodiments, based on a comprehensive analysis of the crack morphology diversity and the impact of rainfall and humidity distribution on the calibrated data, an early warning signal for a specific power grid equipment surface mud film cracking scenario is generated and output to the maintenance decision module to complete subsequent calibration.

[0066] The detailed implementation method is as follows: First, based on the calibrated data, the width, depth, length and number of branches of the cracks are statistically analyzed, and the cracks are classified into three types: dendritic, network and parallel to determine the distribution of crack morphology. Dendritic cracking is characterized by a main crack extending into multiple secondary cracks, with branch angles ranging from 30 to 60 degrees, resembling the forking of a tree branch. Network cracking involves cracks intertwining to form multiple closed or semi-closed polygonal regions, with the number of intersections exceeding one-third of the total number of cracks per unit area. Parallel cracking presents multiple approximately parallel cracks, with the variation in the distance between adjacent cracks being less than 20% of the average distance.

[0067] For the established morphological distribution, the crack propagation rate is calculated by comparing the crack length changes at adjacent times and dividing by the time interval. Specifically, a time-series comparison method is used, where the length of the same crack is measured at two consecutive monitoring times, and the propagation rate is obtained by dividing the length increment by the time interval. When the monitoring frequency is once every 5 minutes, if the crack length increases from 100 mm to 102 mm, the propagation rate is 0.4 mm per minute. This direct measurement method avoids complex image registration problems. Then, the Pearson correlation coefficient is calculated between rainfall and humidity data and the propagation rate. The Pearson correlation coefficient is used to quantify the linear correlation between rainfall / humidity and crack propagation rate. The calculation formula is as follows: , Where r is the Pearson correlation coefficient, n is the sample size, and Xi For the i-th set of rainfall and humidity data, Y is the average of all rainfall and humidity data. i For the i-th group of crack propagation rate data, The correlation coefficient is the average of all crack propagation rate data, ranging from -1 to 1. If the correlation coefficient exceeds the preset threshold of 0.7, it is determined to be a humidity-sensitive crack. A correlation coefficient between 0.3 and 0.7 indicates moderate correlation, and less than 0.3 indicates weak or no correlation. The risk assessment level is determined by combining the propagation rate and morphological distribution. Then, based on the risk assessment level, the ID3 decision tree algorithm is used to construct classification rules. Risk level, morphological type, and humidity sensitivity are used as input nodes. The root node represents the risk level, divided into high, medium, and low branches. Each branch has a morphological type sub-node, further subdivided into humidity sensitivity. Leaf nodes correspond to specific warning levels, increasing from level one to level five. The decision path judges layer by layer based on the actual input attribute values, ultimately reaching the corresponding warning level. The warning level is then encoded into a three-digit signal: the hundreds digit represents the risk level, the tens digit represents the morphology type, and the units digit represents the humidity sensitivity. The code 521 represents high risk, mesh-like cracking, and humidity sensitivity. This signal is then transmitted to the maintenance decision module via RS-485 serial communication protocol. This protocol has a baud rate of 9600, 8 data bits, 1 stop bit, and no parity bit, ensuring reliable data transmission. After receiving the warning signal, the maintenance decision module automatically parses the corresponding risk level, morphology type, and humidity sensitivity to generate maintenance priorities and suggested processing time limits. High priority is ≤24 hours, medium priority is ≤72 hours, and low priority is ≤168 hours. This information is then fed back to the power grid equipment operation and maintenance management system, achieving a closed-loop interaction between calibration and maintenance decisions.

[0068] Please see Figure 2 This application also provides a power grid micro-meteorological data calibration device, which can implement the above-mentioned power grid micro-meteorological data calibration method. The device includes: The fusion module 201 is used to acquire image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and to perform spatiotemporal alignment and fusion of the image data and micro-meteorological sensor data to form a related dataset. The denoising module 202 is used to perform layered denoising processing on the image data in the associated dataset to obtain a preprocessed image; Quantization module 203 is used to extract and quantify the morphological state characteristics of the surface coating of power grid equipment based on preprocessed images and micro-meteorological sensor data; The generation module 204 is used to dynamically generate parameterization rules for calibrating micro-meteorological sensor data based on the quantized morphological state characteristics. The calibration module 205 is used to calibrate the micro-weather sensor data by applying parameterized rules and output the calibrated data.

[0069] The specific implementation method of the power grid micro-meteorological data calibration device is basically the same as the specific implementation method of the power grid micro-meteorological data calibration method described above, and will not be repeated here.

[0070] Thirdly, embodiments of this application provide an electronic device, see [link to relevant documentation]. Figure 3 The diagram shown is a structural schematic of an electronic device provided in this application.

[0071] like Figure 3 As shown, the device includes: Memory 31 is used to store computer programs; Processor 32 is used to execute computer programs; The processor 32 executes a computer program to implement the power grid micro-meteorological data calibration method as described in any of the above embodiments.

[0072] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 31 and executed by processor 32 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in an electronic device.

[0073] The processor 32 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0074] The memory 31 can be used to store computer programs and / or modules. The processor 32 implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory 31 and calling the data stored in the memory 31. The memory 31 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 31 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0075] It should be noted that the aforementioned electronic devices include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 3 The structural diagram is merely an example of the electronic device described above and does not constitute a limitation on the electronic device. It may include more components than shown in the diagram, or combine certain components, or use different components.

[0076] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed, implements the power grid micro-meteorological data calibration method of any of the above embodiments.

[0077] It should be understood that the implementation of all or part of the above-described power grid micro-meteorological data calibration method can also be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described power grid micro-meteorological data calibration method. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the relevant jurisdiction. For example, in some relevant jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0078] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0079] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A method for calibrating micrometeorological data of a power grid, characterized in that, include: Acquire image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and perform spatiotemporal alignment and fusion of the image data and the micro-meteorological sensor data to form a related dataset; The image data in the associated dataset is subjected to hierarchical denoising processing to obtain a preprocessed image; Based on the preprocessed image and the micro-meteorological sensor data, the morphological state characteristics of the surface coating of the power grid equipment are extracted and quantified. Based on the quantized morphological state characteristics, parameterization rules for calibrating the micro-meteorological sensor data are dynamically generated. The parameterization rules are applied to calibrate the micro-weather sensor data, and the calibrated data is output.

2. The method according to claim 1, characterized in that, The process involves acquiring image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, then performing spatiotemporal alignment and fusion of the image data and the micro-meteorological sensor data to form a correlated dataset, including: Construct a spatial coordinate system for the surface of power grid equipment; Acquire image data of the surface of the power grid equipment, the image data including high-resolution images and infrared temperature distribution maps; At the corresponding location points in the spatial coordinate system, micro-meteorological sensor data are collected synchronously. The micro-meteorological sensor data includes at least one of the following: dust particle density, dust particle size distribution, rainfall humidity distribution, mud film surface roughness, and mud film surface adhesion. Based on timestamps and spatial coordinates, the micro-meteorological sensor data is used as an environmental label and fused with the high-resolution image and infrared temperature distribution map to obtain the associated dataset.

3. The method according to claim 1, characterized in that, The step of performing hierarchical denoising processing on the image data in the associated dataset to obtain a preprocessed image includes: The low-frequency background noise formed by uniform dust deposition in the image data is identified and filtered out. An edge-preserving filter is used to remove high-frequency stripe noise caused by rainfall erosion from the image data; Feature enhancement transformation is used to preserve and highlight the texture and spectral features in the image data that are related to the moisture content or adhesion of the coating.

4. The method according to claim 1, characterized in that, The surface coating of the power grid equipment is a mud film or an ice layer, and the morphological characteristics are cracked characteristics.

5. The method according to claim 4, characterized in that, The cracking state characteristics include at least one of crack density, width, depth, morphological classification, and network connectivity; the extraction and quantification of the morphological state characteristics of the surface coating of the power grid equipment based on the preprocessed image and the micro-meteorological sensor data includes: Extract the edge contour of the crack from the preprocessed image and calculate the crack width; The relative depth of the cracks was assessed by combining the dust particle size data from the micro-meteorological sensor data. The number and branching of cracks per unit area are statistically analyzed to determine crack density and crack morphology classification.

6. The method according to claim 5, characterized in that, The parameterization rules for calibrating the micro-meteorological sensor data, based on the quantized morphological state features, include: A correlation model is established between the cracking state characteristics and the data error of the micro-meteorological sensor. The correlation model is a statistical model or a graphical model constructed based on the cracking network connectivity characteristics and near-surface humidity gradient data. Based on the cracking state characteristics at the current moment, the sensor data error compensation coefficient is calculated and processed through the correlation model to obtain the error compensation coefficient that constitutes the parameterization rule.

7. The method according to claim 5, characterized in that, After extracting and quantifying the morphological features of the surface coating of the power grid equipment, the method further includes: The acquisition frequency of the image data or the micro-meteorological sensor data is dynamically adjusted based on the quantified morphological state characteristics or their rate of change.

8. A power grid micro-meteorological data calibration device, characterized in that, include: The fusion module is used to acquire image data of the surface of the power grid equipment and micro-meteorological sensor data of the area where the power grid equipment is located, and to perform spatiotemporal alignment and fusion of the image data and the micro-meteorological sensor data to form a related dataset. The denoising module is used to perform layered denoising processing on the image data in the associated dataset to obtain a preprocessed image; The quantization module is used to extract and quantify the morphological state features of the surface coating of the power grid equipment based on the preprocessed image and the micro-meteorological sensor data. The generation module is used to dynamically generate parameterization rules for calibrating the micro-meteorological sensor data based on the quantized morphological state features. The calibration module is used to calibrate the micro-weather sensor data by applying the parameterization rules and output the calibrated data.

9. An electronic device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power grid micro-meteorological data calibration method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the power grid micro-meteorological data calibration method as described in any one of claims 1 to 7.

Images