A method and system for intelligent monitoring of the operating state of a photovoltaic power plant

Abstract

The present application relates to the field of image analysis, and more particularly to a kind of photovoltaic power station operating state intelligent monitoring method and system.The method comprises the following steps: based on unmanned aerial vehicle to the photovoltaic array of photovoltaic power station carries out periodic inspection, acquires standard image sequence and thermal imaging image sequence;Standard image sequence is subjected to snow detection calculation, and the snow cover index variation curve is constructed;Thermal imaging image sequence is subjected to regional temperature rise rate calculation, and the thermal trend evolution track of photovoltaic array surface is obtained;According to the thermal trend evolution track and the snow cover index variation curve, component failure risk calculation and risk discrimination are carried out, and high-risk area is marked;High-risk area is executed in area photovoltaic component multi-cycle tracking analysis, and critical failure point identification is carried out, and critical failure component and critical time point are marked.The present application realizes efficient, accurate photovoltaic power station risk area positioning and evaluation, improves power station operating efficiency and reduces failure rate.

Description

Technical Field

[0001] This invention relates to the field of image analysis, and in particular to an intelligent monitoring method and system for the operating status of photovoltaic power plants. Background Technology

[0002] In the actual operation of photovoltaic power plants, photovoltaic modules are exposed to a complex natural environment for extended periods, making them susceptible to various factors such as snow cover, abnormal temperatures, dust obstruction, and localized hot spots. This can lead to decreased power generation efficiency or even partial module failure. Currently, common monitoring methods in the industry mainly rely on manual ground inspections or single electrical performance monitoring methods, which suffer from problems such as long inspection cycles, limited coverage, insufficient real-time performance, and difficulty in effectively identifying early-stage hidden faults. Especially in large-scale photovoltaic power plants, where the number of modules is vast and their distribution is dispersed, traditional methods struggle to achieve refined and continuous condition monitoring and risk warnings, resulting in unnecessary power generation losses and potential failure risks. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes an intelligent monitoring method and system for the operating status of photovoltaic power plants, thereby resolving at least one of the aforementioned technical issues.

[0004] To achieve the above objectives, the present invention provides an intelligent monitoring method for the operating status of a photovoltaic power station, comprising the following steps: Step S1: Conduct periodic inspections of the photovoltaic array of the photovoltaic power station using drones, and collect standard image sequences and thermal imaging image sequences; Step S2: Perform snow cover detection calculations on the standard image sequence and construct a snow cover index change curve; Step S3: Calculate the regional temperature rise rate of the thermal imaging image sequence to obtain the thermal trend evolution trajectory of the photovoltaic array surface; Step S4: Calculate and identify component failure risk based on the thermal trend evolution trajectory and snow cover index change curve, and mark high-risk areas; Step S5: Perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure modules and critical time points; Step S6: Perform dynamic early warning and adaptive operation control based on critical failure components and critical time points to complete intelligent monitoring operations.

[0005] In this invention, step S1 specifically involves the following steps: Set a fixed inspection cycle, drive the drone to perform low-altitude flight path scanning of the photovoltaic array of the photovoltaic power station, and simultaneously collect dual-modal images to extract visible light images and thermal imaging image sequences of different inspection cycles; The visible light image is used to identify and mark highly reflective areas of snow accumulation. Dynamic exposure correction is performed on the reflective areas, and super-resolution enhancement is performed on the remaining low-light areas to obtain the corrected image; The corrected image is filtered for snow and fog interference and texture details are restored to output a standard image sequence.

[0006] In this invention, step S2 specifically involves the following steps: Frame-by-frame depth visual inspection is performed on standard image sequences to extract the snow-covered areas on the surface of the photovoltaic array; Visual semantic segmentation is performed on the snow-covered area to extract the non-snow-covered area; Based on the snow-covered and non-snow-covered areas, the snow-covered area, the proportion of snow-covered area, and the spatial distribution parameters of snow accumulation are calculated to generate the snow cover index. The snow cover index is tracked and calculated over multiple inspection cycles based on standard image sequences to construct a snow cover index change curve.

[0007] In this invention, step S3 specifically involves the following steps: Noise filtering is applied to the thermal imaging image sequence to obtain filtered thermal images; The temperature values ​​of multiple surface regions are calculated and calibrated for filtered thermal imaging, generating temperature calibration values ​​for different locations. Based on the temperature calibration values, a temperature distribution is fitted to construct a temperature cloud map; Perform continuous inspection cycle time-series change analysis on temperature cloud maps to generate dynamic temperature cloud maps; Calculate the regional temperature rise rate from the dynamic temperature cloud map and extract the temperature rise rate of different regions; The thermal development trend evolution of the temperature rise rate in different regions was analyzed to obtain the thermal trend evolution trajectory of the photovoltaic array surface.

[0008] In this invention, step S4 specifically involves the following steps: Based on the thermal trend evolution trajectory, the heat distribution dispersion and gradient change rate of different regions are calculated to generate the temperature difference evolution characteristics of different regions. Based on the snow cover index change curve, the correlation between snow cover shading and local heat accumulation is identified for the temperature difference evolution characteristics, and low temperature failure correlation characteristics are generated. Component failure risk is calculated based on the low-temperature failure correlation characteristics to obtain failure risk values ​​for different regions; The failure risk value is judged according to the preset failure risk threshold, and high-risk areas are marked.

[0009] In this invention, step S5 specifically involves the following steps: For high-risk areas, perform multi-cycle tracking analysis of photovoltaic modules within the area to extract the duration of thermal anomalies of each target module, the spatial diffusion direction and diffusion rate of the abnormal temperature zone, and obtain module information; Obtain the component archive database and extract historical running status; Low-temperature fatigue calculations are performed on component information based on historical operating conditions to obtain the low-temperature fatigue coefficient; The component lifespan decay is calculated based on the low-temperature fatigue coefficient, and a lifespan decay curve is generated. Critical failure points are identified based on lifetime decay curves, and critical failure components and critical time points are marked.

[0010] In this invention, the historical operating status includes the number of years of operation, the cumulative number of hours of low-temperature operation, and the frequency of historical snow cover.

[0011] In this invention, step S6 specifically involves the following steps: Identify the location coordinates of the critical failure component; Based on the critical time points, risk levels are classified to obtain the risk levels of different components; Based on the location coordinates, the risk levels are spatially distributed statistically to construct a risk distribution map; Based on the risk distribution map, large-area failure propagation prediction is performed to generate failure propagation probability; Dynamic early warning is generated based on the failure propagation probability, resulting in graded early warning results. Based on the results of the graded early warning, adaptive operation and control processing is triggered to complete the intelligent monitoring operation.

[0012] This specification provides an intelligent monitoring system for the operating status of a photovoltaic power station, used to execute the intelligent monitoring method for the operating status of a photovoltaic power station as described above, including: The image acquisition module is used to conduct periodic inspections of photovoltaic arrays in photovoltaic power plants using drones, and to acquire standard image sequences and thermal imaging image sequences. The snow detection module is used to perform snow detection calculations on standard image sequences and construct snow cover index change curves. The temperature rise calculation module is used to calculate the regional temperature rise rate of thermal imaging image sequences to obtain the thermal trend evolution trajectory of the photovoltaic array surface. The risk assessment module is used to calculate and assess the component failure risk based on the thermal trend evolution trajectory and the snow cover index change curve, and to mark high-risk areas. The failure analysis module is used to perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure components and critical time points. The intelligent control module is used to perform dynamic early warning and adaptive operation control based on critical failure components and critical time points, and to complete intelligent monitoring operations.

[0013] The specific benefits of this invention are as follows: By using drones to conduct periodic multimodal inspections of photovoltaic arrays, combined with visible light and thermal imaging data, comprehensive perception and continuous monitoring of the photovoltaic module's operating status can be achieved, significantly improving the completeness and real-time performance of data acquisition. Based on this, by detecting snow accumulation in standard image sequences and constructing a snow cover index change curve, discrete visual information is transformed into continuous quantitative indicators, enabling dynamic expression of snow accumulation changes and improving the ability to identify environmental interference factors. Calculating the regional temperature rise rate and constructing a thermal trend evolution trajectory using thermal imaging data can effectively reflect the changing characteristics of module thermal distribution over time, enabling early identification and trend prediction of thermal anomalies. Further integrating snow cover index and thermal trend data for multi-dimensional joint analysis improves the accuracy and robustness of module failure risk assessment, achieving precise location of high-risk areas. Conducting multi-period tracking analysis based on high-risk areas enables the transformation from regional-level risk to module-level refined diagnosis, identifying critical failure points and their occurrence times, and improving the foresight and accuracy of fault prediction. Ultimately, by using dynamic early warning and adaptive operation control based on critical information, the photovoltaic system can achieve optimized operation management, effectively reduce the failure rate and power generation loss, and improve the overall operational safety and intelligence level. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the steps of an intelligent monitoring method for the operating status of a photovoltaic power station according to the present invention. Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a detailed flowchart illustrating the implementation steps of step S2; Figure 4 This is a schematic diagram of the snow cover index variation curve; Figure 5 This is a schematic diagram of the dynamic temperature cloud map for the 4th inspection cycle. Figure 6 This is a schematic diagram of the dynamic temperature cloud map for the 6th inspection cycle. Figure 7 This is a schematic diagram of the dynamic temperature cloud map for the 14th inspection cycle. Detailed Implementation

[0015] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0016] This application provides a method and system for intelligent monitoring of the operating status of a photovoltaic power station. The executing entities of the intelligent monitoring method and system for the operating status of the photovoltaic power station include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered as general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0017] Please see Figures 1 to 7 This invention provides an intelligent monitoring method for the operating status of photovoltaic power plants, comprising the following steps: Step S1: Conduct periodic inspections of the photovoltaic array of the photovoltaic power station using drones, and collect standard image sequences and thermal imaging image sequences; Step S2: Perform snow cover detection calculations on the standard image sequence and construct a snow cover index change curve; Step S3: Calculate the regional temperature rise rate of the thermal imaging image sequence to obtain the thermal trend evolution trajectory of the photovoltaic array surface; Step S4: Calculate and identify component failure risk based on the thermal trend evolution trajectory and snow cover index change curve, and mark high-risk areas; Step S5: Perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure modules and critical time points; Step S6: Perform dynamic early warning and adaptive operation control based on critical failure components and critical time points to complete intelligent monitoring operations.

[0018] In this embodiment, a 50MW photovoltaic power station in a high-altitude and cold region is used as an example. The photovoltaic array contains 12,000 photovoltaic modules. The drone performs low-altitude inspections at a fixed 2-hour inspection cycle, which is shortened to 30 minutes under blizzard conditions. The drone flies at an altitude of 10m and a speed of 4m / s. The visible light camera has a resolution of 8000×6000 pixels, and the thermal imaging camera has a temperature measurement range of -40℃ to 120℃ and a thermal sensitivity of 0.05℃. The system synchronously acquires visible light images and thermal imaging image sequences at a frequency of once every 2 seconds.

[0019] Image acquisition was performed on the area where the PV-23 component is located. Image grayscale analysis showed that the average brightness of the local area on the component surface reached 238, and the pixel saturation ratio reached 22%. Since the average brightness exceeded 230 and the pixel saturation ratio exceeded 18%, the area was determined to be a high-reflectivity area covered by snow. The system performed dynamic exposure correction on this area, setting the exposure reduction ratio to 15%. At the same time, the average brightness of the lower edge area of ​​the component was detected to be 72, which is lower than the low illumination threshold of 80. Therefore, super-resolution enhancement processing was performed on this area. After enhancement, the image resolution was increased to 1.5 times that of the original image. Subsequently, the air transmittance of the image was detected to be 0.61, which is lower than the snow and fog interference threshold of 0.65. Therefore, snow and fog filtering and texture restoration processing were performed to finally generate a standard image sequence.

[0020] Snow cover recognition was performed on a standard image sequence; the system identified a snow-covered area of ​​185,000 pixels for the PV-23 component, corresponding to a total component area of ​​420,000 pixels; after spatial resolution conversion, the actual snow-covered area of ​​the component was 1.12m². 2 The total area of ​​the components is 2.54m². 2 The formula for calculating the proportion of snow-covered area is:

[0021] Rs represents the snow-covered area ratio, As represents the snow-covered area, and At represents the total area of ​​the component. Substituting the data, we get:

[0022] Since the snow-covered area ratio exceeds 40%, the component is determined to be in a state of heavy snow cover. Further, combining the snow cover dispersion (0.71) and snow distribution uniformity parameter (0.82), a snow cover index is generated. To ensure dimensional consistency and comparability in the snow cover index calculation process, this embodiment uses a unified dimensionless normalization method for all parameters involved in the calculation, mapping physical quantities or statistics from different sources to the interval [0,1]. Specifically, the snow cover area ratio Rs, originally in percentage form (0%–100%), is converted to a 0–1 interval value using the normalization function Rs'=Rs / 100. The snow cover dispersion Ds is normalized using the coefficient of variation based on the ratio of standard deviation to mean, calculated as Ds=(σ / μ) / (σ / μ)max, limiting it to the range [0,1]. The snow distribution uniformity parameter Us is calculated based on the grid distribution entropy value and normalized using Us'=(H / Hmax), also falling within the [0,1] interval.

[0023] After standardization, the snow cover index calculation formula is actually expressed using a dimensionless weighted model as follows:

[0024] Si represents the snow cover index, Ds represents the snow cover dispersion, and Us represents the snow cover distribution uniformity parameter. Substituting the data yields the snow cover index:

[0025] After three consecutive inspection cycles, the snow cover index rose to 64, 71 and 78 respectively, forming a snow cover index change curve.

[0026] The thermal imaging image sequence was analyzed; the system divided the component into 6×10 temperature analysis units; the temperature of the central area of ​​the PV-23 component was detected to be 48℃, the average temperature of the surrounding area was 29℃, and the local temperature difference reached 19℃; the temperature rise rate was calculated based on the regional temperature change results; the formula for calculating the temperature rise rate is:

[0027] Vt represents the rate of temperature rise, T1 represents the initial temperature, and T2 represents the current temperature; If the temperature is 41℃ at 08:00 and 48℃ at 10:00, then:

[0028] If the thermal diffusion boundary expands by 20 cm in 6 hours, then the thermal diffusion rate is:

[0029] A thermal trend evolution trajectory is generated based on the temperature change results of multiple inspection cycles.

[0030] Thermal risk analysis of the module revealed that the standard deviation of the module's regional temperature reached 7.8℃, exceeding the thermal distribution dispersion threshold of 6℃; the gradient change rate reached 2.1℃ / cm, exceeding the threshold of 1.8℃ / cm; simultaneously, the snow cover index reached 78, and the local temperature difference reached 19℃; therefore, it was determined that the module exhibited a coupling phenomenon of snow cover shading and localized heat accumulation; the formula for calculating the module failure risk value is as follows:

[0031] Fr represents the failure risk value, Dt represents the thermal distribution dispersion, and Gt represents the gradient change rate. Substitute the normalized data:

[0032] Since the failure risk value of 79 exceeds the high risk threshold of 70, the area where PV-23 is located is marked as a high-risk area.

[0033] Lifetime analysis was performed on high-risk components; PV-23 data was retrieved from the components, identifying their 8-year operational lifespan, 18,500 hours of cumulative low-temperature operation, and 126 historical snow cover events; the low-temperature fatigue coefficient was calculated based on the thermal anomaly duration of 6 hours, the thermal diffusion rate of 3.3 cm / h, and historical operating data; the formula for calculating the low-temperature fatigue coefficient is as follows:

[0034] Cf represents the low-temperature fatigue coefficient, Ht represents the thermal anomaly factor, Lt represents the low-temperature operation factor, and Sf represents the snow cover factor. Substitute the normalized data:

[0035] Since the low-temperature fatigue coefficient exceeds 0.7, the component is determined to be in a high fatigue state. Further calculation of the life health value shows that the current life health value is 58, with an average annual life decline of 6.5%. It is predicted that the life health value will drop below 30 in 18 months, so 18 months is identified as the critical time point.

[0036] The spatial coordinates of component PV-23 are identified as X=125m, Y=48m; simultaneously, 12 primary high-risk components and 26 secondary high-risk components are identified in the surrounding area; the formula for calculating the high-risk cluster density is:

[0037] Dr represents the risk cluster density, Nh represents the number of high-risk components, and Nt represents the total number of components in the region. If the total number of components in the region is 90, then:

[0038] Because the risk cluster density exceeds 40%, the area is determined to have a concentrated failure risk. Further, combined with the thermal diffusion rate of 4.2 cm / h and the snow cover index of 78, it is predicted that the probability of forming a large-scale inefficient power generation area within the next 60 days reaches 68%. Because the failure diffusion probability exceeds 60%, the system outputs an advanced early warning result. At the same time, the drone inspection cycle is automatically shortened from 2 hours to 20 minutes, and local load isolation and energy storage compensation control are initiated to complete the intelligent monitoring and adaptive operation management of the photovoltaic power station.

[0039] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Set a fixed inspection cycle, drive the drone to perform low-altitude flight path scanning of the photovoltaic array of the photovoltaic power station, and simultaneously collect dual-modal images to extract visible light images and thermal imaging image sequences of different inspection cycles; The visible light image is used to identify and mark highly reflective areas of snow accumulation. Dynamic exposure correction is performed on the reflective areas, and super-resolution enhancement is performed on the remaining low-light areas to obtain the corrected image; The corrected image is filtered for snow and fog interference and texture details are restored to output a standard image sequence.

[0040] In this embodiment, a fixed inspection cycle is set, and the drone is driven to periodically scan the photovoltaic array of the photovoltaic power station at a low altitude. The drone performs full-coverage flight data collection of the component area according to the preset route. The inspection cycle is set to 2 hours, which is shortened to 30 minutes under snowfall, cold wave, and strong wind conditions. The drone's flight altitude is controlled within the range of 8m to 15m, and its flight speed is controlled within the range of 3m / s to 6m / s. The drone is equipped with a visible light camera and a thermal imaging camera, with the visible light camera having a resolution of 8000×6000 pixels to collect surface texture images of the components. The thermal imaging camera's temperature measurement range is set to -40℃ to 120℃, and its thermal sensitivity is set to 0.05℃. It is used to collect images of the thermal distribution on the surface of the components. The UAV simultaneously acquires visible light images and thermal imaging image sequences at a sampling frequency of once every 2 seconds, and establishes a dual-modal image dataset based on the timestamp, UAV positioning coordinates, and component number. The inspection cycle is the time interval corresponding to the UAV completing one complete inspection of the photovoltaic array. The dual-modal image is a sequence of visible light images and thermal imaging images acquired synchronously at the same time node. The thermal imaging image sequence is a temperature distribution image generated based on the infrared radiation intensity of the components.

[0041] The visible light image is used to identify highly reflective snow-covered areas. An image grayscale distribution matrix is ​​constructed, and the mean brightness, variance brightness, and pixel saturation ratio of different areas are calculated. The mean brightness is the average of the grayscale values ​​of all pixels in the area; the variance brightness is the degree of brightness fluctuation in the area; the pixel saturation ratio is the proportion of pixels with grayscale values ​​reaching the maximum threshold to the total number of pixels in the area. When the mean brightness of an area exceeds 230 and the pixel saturation ratio exceeds 18%, the area is determined to be a highly reflective snow-covered area. Local brightness gradient analysis is performed on the highly reflective areas to calculate the rate of brightness change between adjacent pixels, and the snow reflection boundary is extracted based on the rate of brightness change. The brightness gradient is the rate of brightness change within a unit pixel distance. Combining the component border position, component arrangement direction, and historical snow distribution data, spatial continuity analysis is performed on the bright areas to filter out metal bracket reflections, lens flare, and water reflection areas. Spatial continuity is the degree of continuous spatial distribution of the bright areas. The results of marking highly reflective snow-covered areas are output.

[0042] Dynamic exposure correction is performed on reflective areas, and super-resolution enhancement is applied to the remaining low-light areas to obtain a corrected image. Exposure compensation parameters are calculated based on the average brightness and pixel saturation ratio of highly reflective areas. When the average brightness of the area is within the range of 230–245, the exposure reduction ratio is set to 15%; when the average brightness exceeds 245, the exposure reduction ratio is set to 30%. These exposure compensation parameters are correction coefficients used to adjust the local brightness of the image. Local brightness compression processing is performed on highly reflective areas to restore the texture of the snow-covered areas. Areas with an average brightness below 80 are identified as low-light areas. The region is defined as an area where the image brightness is below a preset threshold due to overcast or snowy weather, low temperature, or insufficient sunlight. A multi-frame image fusion method is used to perform super-resolution enhancement on the low-light region. Feature registration is performed on continuous inspection images, and component border textures, crack textures, and snow boundary textures are extracted for pixel reconstruction. Super-resolution enhancement is a method to improve the spatial resolution of the image through multi-frame fusion and texture reconstruction. Feature registration is the process of aligning the positions of the same target regions between different images. After enhancement, the image resolution is increased to more than 1.5 times that of the original image, and the brightness of the low-light region is increased by more than 20%, resulting in the output of a corrected image.

[0043] Snow and fog interference filtering is applied to the corrected image, and texture details are restored to output a standard image sequence. Local contrast, edge sharpness, and atmospheric transmittance are calculated to evaluate the degree of snow and fog interference. Local contrast represents the difference in brightness between local areas of the image; edge sharpness represents the intensity of gradient changes at the image edges; atmospheric transmittance is the proportion of effective light transmitted through the image. When atmospheric transmittance is below 0.65, the image is considered to have snow and fog interference. A snow and fog filtering method based on dark channel estimation is used to compensate for brightness in low-frequency fogged areas and protect the edges of high-frequency textured areas. The dark channel is the image... The minimum brightness channel value in a local area is used; texture interpolation is performed on the component border, crack edge, and snow boundary based on the component texture template; texture interpolation is a method to restore missing details based on the trend of texture change in the neighborhood; the edge sharpness of the restored image is improved by more than 25%, and the component texture recognition rate is improved by more than 30%; the processed images are uniformly arranged according to the inspection time order, and a standard image index is established according to the component number, timestamp, and spatial coordinates to output a standard image sequence; the standard image sequence is a continuous image set after exposure correction, super-resolution enhancement, snow and fog filtering, and texture restoration.

[0044] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Frame-by-frame depth visual inspection is performed on standard image sequences to extract the snow-covered areas on the surface of the photovoltaic array; Visual semantic segmentation is performed on the snow-covered area to extract the non-snow-covered area; Based on the snow-covered and non-snow-covered areas, the snow-covered area, the proportion of snow-covered area, and the spatial distribution parameters of snow accumulation are calculated to generate the snow cover index. The snow cover index is tracked and calculated over multiple inspection cycles based on standard image sequences to construct a snow cover index change curve.

[0045] In this embodiment, standard image sequences are input into a depth vision recognition model according to the inspection time order. Component region detection and snow-covered target recognition are performed on each frame of the image. A photovoltaic module surface texture feature library is established, including module glass texture, cell texture, border texture, and snow texture. Snow-covered areas on the module surface are identified through image texture difference analysis. Snow-covered areas are regions on the module surface that are covered by snow, causing the original module texture features to disappear or weaken. The local grayscale mean, texture energy value, and edge density parameter are calculated. The texture energy value is the intensity of local texture change in the image; the edge density parameter is the proportion of edge pixels per unit area. When the local grayscale mean is higher than 210, When the texture energy value is below 0.35 and the edge density parameter is below 0.25, the area is determined to be a snow-covered area. Further spatial correlation analysis is performed on the snow-covered area by combining the component tilt direction, solar incidence angle, and historical snow remnant areas to identify snow accumulation at component edges, partially obscured snow, and overall snow coverage. Spatial correlation analysis is a method for jointly analyzing the continuity and spatial relationship of snow distribution in different areas. Snow cover dynamics are tracked based on changes in snow-covered areas in consecutive frame images to identify the snow diffusion direction and snow movement trend. The snow diffusion direction is the spatial change direction of the snow coverage area over continuous time. The snow-covered area distribution results for each frame image are output.

[0046] Based on the component border coordinates and component layout, component regions are segmented, and pixel-level semantic classification is performed on snow-covered and non-snow-covered regions. Non-snow-covered regions are areas on the component surface that are not covered by snow and can normally receive solar radiation. Contour extraction is performed on the edges of snow-covered regions, and the boundary continuity parameter is calculated. The boundary continuity parameter represents the degree of continuous change of the snow-covered edge curve. When the boundary continuity parameter is higher than 0.8, the snow-covered region is considered to have an intact edge. Texture restoration detection is performed on non-snow-covered regions to analyze component texture integrity, cell boundary clarity, and cell reflection characteristics. Texture integrity represents the original surface texture of the component. Texture preservation level; cell boundary clarity is the intensity of grayscale change at the cell edge; further, based on the differences in brightness distribution and texture change within the module area, the partially snow-covered area is locally segmented to accurately separate the partially snow-covered area from the exposed module area; the partially snow-covered area is the area where the module surface is partially obscured by snow but the visible module area still exists; temporal matching is performed on the snow-covered and non-snow-covered areas in consecutive frames to establish the region correspondence under different inspection cycles; temporal matching is the position mapping process of the same module area at different time nodes; output the semantic segmentation results of the snow-covered and non-snow-covered areas.

[0047] The pixel area corresponding to the snow-covered region is calculated based on the semantic segmentation results, and the actual snow-covered area is obtained by converting it according to the drone's flight altitude, lens focal length, and image spatial resolution. The snow-covered area is the actual physical area corresponding to the area on the component surface covered by snow. The snow-covered area ratio is further calculated based on the total component area. The snow-covered area ratio is the proportion of the snow-covered area to the total component area. When the snow-covered area ratio is less than 10%, the component is determined to be in a light snow-covered state; when the snow-covered area ratio is between 10% and 40%, the component is determined to be in a moderate snow-covered state; when the snow-covered area ratio exceeds 40%, the component is determined to be in a heavy snow-covered state. Simultaneously, spatial statistical analysis is performed on the distribution of snow-covered areas on the component surface. The system calculates the snow center location, snow dispersion, and snow distribution uniformity parameters. The snow center location is the coordinate of the geometric center of the snow area. The snow dispersion is the degree of dispersion of the snow area on the component surface. The snow distribution uniformity parameter is the degree of evenness of the snow distribution on the component surface. Furthermore, it identifies the spatial distribution of snow by combining the component tilt angle, wind direction, and snow thickness variation trends. A snow cover index is constructed based on the snow coverage area ratio, snow dispersion, and snow distribution uniformity parameters. The snow cover index is a comprehensive evaluation parameter used to characterize the degree of snow coverage and spatial distribution of the component. In this embodiment, the snow cover index ranges from 0 to 100, with a higher value indicating a higher risk of snow cover. The component snow cover index result is then output.

[0048] The snow cover index was arranged in time series according to the inspection time sequence under different inspection cycles, and a module-level snow cover change database was established. The snow cover change database is a data set that records the changes in snow cover status of the module at different time points. The snow cover growth rate, snow cover melting rate, and snow cover duration were calculated based on the snow cover index change results of continuous inspection cycles. The snow cover growth rate is the increase of the snow cover index per unit time; the snow cover melting rate is the decrease of the snow cover index per unit time; and the snow cover duration is the length of time the module is continuously in a snow-covered state. When the snow cover growth rate exceeds 15 per hour, the module is determined to have entered the rapid snow accumulation stage; when the snow cover duration exceeds 8 hours, the module is considered to have entered the rapid snow accumulation stage. When the snow cover index consistently exceeds 60, the module is deemed to have a long-term risk of being obstructed by snow accumulation. Further analysis of the snow cover trend is conducted by considering ambient temperature, wind speed, and solar irradiance. The snow cover trend represents the pattern of snow accumulation on the module over time. A snow cover index variation curve is constructed based on the snow cover index changes over multiple inspection cycles, and periodic fluctuation and anomalous change analyses are performed on this curve. Periodic fluctuation analysis examines the regular changes in the snow cover index over different time periods; anomalous change analysis examines the abnormal increases or decreases in the snow cover index within a short period. The module snow cover index variation curve is then output for subsequent snow risk assessment and operational status prediction analysis.

[0049] In this embodiment, step S3 includes the following steps: Noise filtering is applied to the thermal imaging image sequence to obtain filtered thermal images; The temperature values ​​of multiple surface regions are calculated and calibrated for filtered thermal imaging, generating temperature calibration values ​​for different locations. Based on the temperature calibration values, a temperature distribution is fitted to construct a temperature cloud map; Perform continuous inspection cycle time-series change analysis on temperature cloud maps to generate dynamic temperature cloud maps; Calculate the regional temperature rise rate from the dynamic temperature cloud map and extract the temperature rise rate of different regions; The thermal development trend evolution of the temperature rise rate in different regions was analyzed to obtain the thermal trend evolution trajectory of the photovoltaic array surface.

[0050] In this embodiment, infrared radiation data analysis is performed on the thermal imaging image sequence acquired by the UAV to establish a thermal radiation matrix on the component surface, and random thermal noise, environmental thermal interference, and low-temperature drift noise in the thermal imaging image sequence are identified. Random thermal noise is the irregular temperature fluctuation signal generated by the thermal imaging sensor in a low-temperature environment; environmental thermal interference is the interference caused by ground reflected heat sources, sunlight reflected heat sources, and thermal radiation from surrounding equipment on the component temperature measurement; low-temperature drift noise is the temperature offset error caused by changes in detector sensitivity in the thermal imaging device in extremely cold environments; thermal field smoothing is performed based on the continuity of temperature changes between adjacent pixels in the thermal imaging image sequence to suppress local abnormal high-frequency temperature fluctuations; when phase When the temperature difference between adjacent pixels exceeds 8℃ and there are no corresponding thermal diffusion characteristics in the surrounding area, the area is determined to be a noise interference area. Isolated thermal noise points are filtered using a median thermal distribution smoothing method, and temperature baseline compensation is performed on low-frequency thermal drift areas. Temperature baseline compensation is a method to correct the overall temperature shift of the thermal image based on the ambient reference temperature. Furthermore, by combining the component's spatial location, component number, and historical inspection temperature data, stability verification is performed on the thermal anomaly area. Stability verification is a method to analyze the continuous consistency of thermal anomalies at the same location across multiple inspection cycles. After filtering, random noise interference in the thermal imaging image sequence is reduced, making the true thermal distribution boundary of the component clearer, and outputting a filtered thermal image.

[0051] Based on the photovoltaic module arrangement structure, the filtered thermal imaging is divided into regional grids, dividing a single module into multiple independent temperature analysis regions. In this embodiment, each module is divided into 6×10 temperature analysis units, each corresponding to an independent region temperature value. The temperature analysis unit is the smallest region unit used for local temperature calculation. The corresponding temperature value is calculated based on the infrared radiation intensity of different regions in the thermal imaging image sequence, and temperature correction is performed in conjunction with the emissivity of the module material. The emissivity is a parameter of the module surface's ability to emit infrared radiation energy; the emissivity of the module glass surface is set to 0.92. Simultaneously, environmental compensation is performed on the temperature data in conjunction with ambient temperature, wind speed, and air humidity. Environmental compensation is a correction process to eliminate the influence of environmental factors on thermal imaging temperature measurement. Furthermore, spatial correction is performed on the temperature measurement error based on the UAV's flight altitude and shooting angle. When the UAV's tilt shooting angle exceeds 15°, angle compensation correction is performed on the temperature of the edge region. Subsequently, the temperature values ​​corresponding to all analysis units are uniformly calibrated, and a mapping relationship between the spatial coordinates of the module surface and the temperature value is established. The temperature calibration value is the true temperature value of the module region after environmental compensation and spatial correction. Finally, a dataset of temperature calibration values ​​at different locations is generated.

[0052] A spatial temperature distribution matrix is ​​established based on the temperature calibration values ​​of different regions on the component surface, and continuous temperature interpolation is performed based on the two-dimensional spatial coordinates of the component. The spatial temperature distribution matrix is ​​a two-dimensional data matrix used to describe the temperature distribution state of different regions on the component surface. A regional thermal continuity fitting method is used to smoothly connect the temperature changes between adjacent analysis units. Thermal continuity fitting is a method to process local temperature changes continuously based on the heat conduction law. Furthermore, local high-temperature accumulation regions and low-temperature accumulation regions are identified based on the temperature gradient changes. The temperature gradient is the rate of temperature change per unit spatial distance. When the regional temperature gradient exceeds 1.5℃ per centimeter, it is determined that there is an abnormal thermal change trend in the region. Different regions are color-mapped according to the component surface temperature, with low-temperature regions mapped to low-heat level colors and high-temperature regions mapped to high-heat level colors. The temperature cloud map is a heat distribution map that intuitively displays the temperature change state of the component surface in a color distribution manner. Furthermore, a whole-site-level temperature cloud map is established by combining the component arrangement direction and array spatial position to uniformly display the differences in heat distribution between different components. The photovoltaic array surface temperature cloud map is output.

[0053] Temperature cloud maps generated from different inspection cycles are arranged in a time series according to the inspection time sequence, and a component thermal distribution time series database is established. The thermal distribution time series database is a data set recording the changes in component thermal state at different time points. Based on the temperature change results of the same component in consecutive inspection cycles, regional temperature differences are calculated to identify the trend of component thermal state change. When the temperature of the same area rises by more than 5°C for two consecutive inspection cycles, it is determined that there is a continuous heating trend in the area. Furthermore, the local high-temperature areas in the temperature cloud map are dynamically tracked, and the changes in the area, positional shift, and heat diffusion direction of the high-temperature areas are analyzed. The heat diffusion direction is the spatial direction of heat propagation on the component surface. Thermal evolution is established based on temperature changes at different time points, and the thermal state of the component surface is dynamically fitted. The dynamic temperature cloud map is a time series thermal distribution map reflecting the continuous change of component surface temperature over time. Further analysis is conducted on temperature fluctuation cycles, diurnal thermal change patterns, and temperature change characteristics caused by snow melting. The temperature fluctuation cycle is the time pattern of repeated changes in component temperature within a certain time range. Finally, a dynamic temperature cloud map is generated. In this case, by analyzing the time series of continuous inspection cycles of the thermal imaging sequence, cloud maps of different cycles are obtained. One cycle corresponds to one temperature cloud map. By tracking the changes of multiple cloud maps, a dynamic temperature cloud map can be obtained.

[0054] The temperature rise rate is calculated based on the temperature change values ​​of different regions in the dynamic temperature cloud map at continuous time nodes; the temperature rise rate is the magnitude of the temperature increase in a region per unit time; in this embodiment, the unit of temperature rise rate is set to ℃ / min; the instantaneous temperature rise rate and the average temperature rise rate are calculated based on the regional temperature change curve; the instantaneous temperature rise rate is the rate of temperature change between two adjacent time nodes; the average temperature rise rate is the average rate of temperature change within a specified time range; when the average temperature rise rate of a region exceeds 0.8℃ / min, it is determined that there is an abnormal heat accumulation phenomenon in that region; the distribution differences of temperature rise rate in the component edge region, center region, and wiring region are further analyzed; the wiring region refers to the component junction box and wire connection area; the local hot spot formation region and the active heat diffusion region are identified based on the temperature rise rate change trend in different regions; the active heat diffusion region is the location where the heat propagation speed is significantly higher than that of the surrounding area; finally, the temperature rise rate results of different regions are extracted.

[0055] A regional thermal evolution time series was established based on the temperature rise rate of different regions, and the duration of temperature rise, direction of heat diffusion, and degree of heat accumulation were jointly analyzed. The degree of heat accumulation refers to the intensity of heat accumulation in a local area. The trend of thermal anomaly diffusion was identified based on the temperature rise rate changes during continuous inspection cycles. When the area of ​​the high-temperature region continuously expands and the temperature rise rate remains above 0.8℃ / min, the region is determined to have a risk of thermal anomaly diffusion. The heat propagation path was fitted with a trajectory. The thermal trend evolution trajectory is the continuous change path of the module's thermal anomaly in the time and spatial dimensions. The propagation characteristics of heat diffusion from the module edge to the center, from local hot spots to adjacent modules, and along the module wiring direction were analyzed. The source region of the thermal anomaly and the core region of heat diffusion were identified based on the thermal trend evolution trajectory. The source region of the thermal anomaly is the region where the abnormal temperature rise first occurs. The core region of heat diffusion is the location where heat continues to accumulate and has a significant thermal impact on the surrounding area. The thermal trend evolution trajectory of the photovoltaic array surface was output for subsequent low-temperature failure analysis and thermal risk prediction.

[0056] In this embodiment, step S4 includes the following steps: Based on the thermal trend evolution trajectory, the heat distribution dispersion and gradient change rate of different regions are calculated to generate the temperature difference evolution characteristics of different regions. Based on the snow cover index change curve, the correlation between snow cover shading and local heat accumulation is identified for the temperature difference evolution characteristics, and low temperature failure correlation characteristics are generated. Component failure risk is calculated based on the low-temperature failure correlation characteristics to obtain failure risk values ​​for different regions; The failure risk value is judged according to the preset failure risk threshold, and high-risk areas are marked.

[0057] In this embodiment, based on the thermal trend evolution trajectory generated by the dynamic temperature cloud map, a statistical analysis of the thermal distribution in different areas of the photovoltaic module surface is performed to establish a module regional thermal field distribution matrix. The thermal field distribution matrix is ​​a two-dimensional temperature data set reflecting the temperature change state in different areas of the module. The thermal distribution dispersion is calculated based on the temperature difference between each temperature analysis unit within the same module. The thermal distribution dispersion represents the degree of unevenness in the temperature distribution within the module. The regional temperature standard deviation is used as the calculation parameter for the thermal distribution dispersion. When the regional temperature standard deviation exceeds 6℃, it is determined that there is a significant thermal distribution anomaly in that region. Furthermore, the gradient rate of change is calculated based on the temperature changes between adjacent regions. The gradient rate of change is the rate of temperature change per unit spatial distance. In this embodiment, the unit of the gradient rate of change is set as follows: The value is ℃ / cm; when the gradient change rate of a region exceeds 1.8℃ / cm, it is determined that there is a rapid change in thermal stress in that region; based on the thermal trend evolution trajectory in the continuous inspection cycle, time series analysis is performed on the thermal distribution dispersion and gradient change rate of different regions to calculate the thermal anomaly growth rate and thermal diffusion direction change parameters; the thermal anomaly growth rate is the change amplitude of the regional thermal distribution dispersion per unit time; the thermal diffusion direction change parameter is the change trend of the heat propagation direction over a continuous time; further, combined with the temperature difference changes between the component edge region, wiring region and center region, the local temperature difference evolution characteristics of the component surface are extracted; the temperature difference evolution characteristics are the thermal change law parameters formed by the temperature difference of the component surface changing over time; the temperature difference evolution characteristic data of different regions are output.

[0058] Based on the snow cover index variation curves generated during different inspection cycles, a time-series correspondence analysis was performed on the snow cover status and thermal change status of the modules. The snow cover index variation curve is a continuous curve showing the change in the degree of snow cover over time. The snow cover area ratio, snow cover duration, and snow spatial distribution parameters were correlated and matched with the temperature difference evolution characteristics at the corresponding time points. The snow cover duration is the cumulative length of time the module remains in a snow cover state. The spatial overlap and temporal correspondence between the snow cover area and the high temperature area were analyzed. When the temperature rise rate of the area corresponding to the edge of the snow cover area exceeds 0.8℃ / min and the local temperature difference exceeds 10℃, it is determined that there is a local heat accumulation phenomenon caused by snow cover in this area. Local heat accumulation is an abnormal temperature rise state formed by the continuous accumulation of heat in a local area of ​​the module. Further analysis was conducted on the melting of the snow cover area. The process includes temperature abrupt changes and thermal diffusion trends; when the snow cover index drops by more than 20 in a short period and the corresponding temperature rises rapidly by more than 8°C, it is determined that there is a thermal stress change phenomenon caused by snow melting in the area; combining the component's operating years, low-temperature operating time, and diurnal temperature variation data, a comprehensive analysis is conducted on the coupling relationship between snow cover and heat accumulation; the low-temperature operating time is the cumulative operating time of the component in an environment below 0°C for a long period; the coupling relationship is the mutual influence relationship between changes in snow cover status and changes in thermal anomalies; further, characteristic parameters such as the duration of heat accumulation, the thermal gradient at the edge of snow cover, and the direction of thermal anomaly diffusion are extracted to construct low-temperature failure correlation features; the low-temperature failure correlation features are comprehensive characteristic parameters used to characterize the correlation state between snow cover, heat accumulation, and low-temperature thermal stress; the results of the low-temperature failure correlation features are output.

[0059] Failure risk is calculated for different regions of the component. Component failure risk refers to the degree of risk of functional degradation or structural damage caused by long-term low-temperature operation, snow cover, and heat anomaly accumulation. The cumulative value of regional thermal stress is calculated based on the heat distribution dispersion, gradient change rate, local heat accumulation intensity, and snow cover duration. The cumulative value of thermal stress is the degree of thermal fatigue accumulation caused by long-term temperature difference changes in the component. Furthermore, the thermal failure impact factor is calculated based on the duration of thermal anomaly, temperature rise rate, and heat diffusion range. The thermal failure impact factor is an evaluation parameter for the degree of impact of thermal anomaly on the structural stability of the component. Combining the component's operating years, historical hot spot records, and low-temperature fatigue coefficient, the low-temperature fatigue coefficient is a parameter of material fatigue formed by the component under long-term low-temperature and freeze-thaw cycle conditions. The failure risk value of the component region is calculated according to the weights corresponding to different characteristic parameters. In this embodiment, the failure risk value range is set to 0-100. When the failure risk value is below 30, the component region is determined to be in a low-risk state; when the failure risk value is between 30 and 70, the component region is determined to be in a medium-risk state; when the failure risk value exceeds 70, the component region is determined to be in a high-risk state. The failure risk value corresponding to different regions is output.

[0060] Risk classification is performed for different regions; the failure risk threshold is the risk value boundary used to divide the risk level of components; in this embodiment, the low-risk threshold is set to 30, and the medium-risk threshold is set to 70; the failure risk values ​​of all component regions are sorted and analyzed to identify risk concentration areas and risk diffusion areas; the risk concentration area is the area formed by the continuous distribution of multiple high-risk components; the risk diffusion area is the trend area formed by the diffusion of high-risk areas to surrounding components; the risk cluster density is calculated based on the spatial distribution of high-risk areas; the risk cluster density is the proportion of high-risk components per unit area; when the risk cluster density exceeds 40%, it is determined that there is a regional risk of inefficient power generation in the area; the diffusion direction and diffusion speed of high-risk areas are further analyzed by combining the thermal trend evolution trajectory; the diffusion speed is the expansion distance of the high-risk area boundary per unit time; the regional risk thermal distribution is processed according to the spatial relationship between high-risk areas and snow-covered areas, and different risk level areas are color-coded; the high-risk area is the component area where the failure risk value continuously exceeds the high-risk threshold and there is a clear trend of thermal anomaly evolution; the high-risk area marking results are output for subsequent intelligent early warning and operation control analysis.

[0061] In this embodiment, step S5 includes the following steps: For high-risk areas, perform multi-cycle tracking analysis of photovoltaic modules within the area to extract the duration of thermal anomalies of each target module, the spatial diffusion direction and diffusion rate of the abnormal temperature zone, and obtain module information; Obtain the component archive database and extract historical operating status; the historical operating status includes the number of years of operation, the cumulative number of hours of low-temperature operation, and the frequency of historical snow cover.

[0062] Low-temperature fatigue calculations are performed on component information based on historical operating conditions to obtain the low-temperature fatigue coefficient; The component lifespan decay is calculated based on the low-temperature fatigue coefficient, and a lifespan decay curve is generated. Critical failure points are identified based on lifetime decay curves, and critical failure components and critical time points are marked.

[0063] In this embodiment, multi-cycle tracking analysis of photovoltaic modules within a high-risk area is performed. Photovoltaic module PV-23 is continuously monitored and tracked for 12 hours. At 08:00, a local temperature of 41℃ is detected in the module, while the average temperature of the surrounding area is 29℃, resulting in a local temperature difference of 12℃. At 10:00, the abnormal area expands to 18% of the module area, with the temperature rising to 48℃. At 14:00, the abnormal area expands to 26%, with a boundary expansion distance of 20cm. The duration of the thermal anomaly is the cumulative time the abnormal temperature zone has existed continuously; in this embodiment, the duration is 6 hours. The diffusion rate is the expansion distance of the thermal anomaly boundary per unit time, which is approximately 3.3cm / h (20cm ÷ 6h ≈ 3.3cm / h). The module's thermal anomaly information is generated by combining the module number, the coordinates of the thermal anomaly center, and the diffusion direction.

[0064] Obtain the component archive database and extract historical operating status; retrieve the historical operating records of component PV-23, identify that the component has been running for 8 years, with a cumulative low-temperature operating hour of 18,500 hours and a historical snow cover frequency of 126 times; the cumulative low-temperature operating hour is the cumulative operating time of the component in an environment below 0℃; further extract the component's historical hot spot records, identify 5 local hot spot anomalies that occurred in the past two years; at the same time, identify that the component's historical power attenuation rate reached 13%; establish a set of long-term low-temperature operating characteristic parameters based on the component's historical operating status.

[0065] Low-temperature fatigue calculations were performed on the component information based on historical operating conditions to obtain the low-temperature fatigue coefficient. Low-temperature fatigue analysis was conducted based on the thermal anomaly duration of the PV-23 component (6 hours), cumulative low-temperature operating hours (18,500 hours), historical snow cover frequency (126 times), and diurnal temperature variation data. The weights for thermal anomaly (0.35), low-temperature operation (0.4), and snow cover (0.25) were set. After normalizing the parameters, the low-temperature fatigue coefficient was calculated as (0.82 × 0.35) + (0.76 × 0.4) + (0.71 × 0.25) = 0.768. The low-temperature fatigue coefficient is a parameter representing the degree of material fatigue formed under long-term low temperature and thermal stress. Since the low-temperature fatigue coefficient exceeds 0.7, the component is determined to be in a high-fatigue state.

[0066] The initial lifetime health value of the PV-23 module was set to 100. After 8 years of operation, the basic lifetime health value decreased to 58. Combining the low-temperature fatigue coefficient of 0.768 and the historical power decay rate of 13%, the average annual lifespan decrease of the module was calculated to be 6.5%. The lifetime health value is a comprehensive evaluation parameter of the current operational health of the module. The lifespan change trend for the next 3 years was further calculated. The lifetime health value decreased to 51 in the first year, 43 in the second year, and 35 in the third year. The lifetime decay curve was generated based on the changes in lifetime health value at continuous time points, and the module was identified as entering the rapid lifetime decay stage.

[0067] Based on the lifetime degradation curve of module PV-23, the slope of lifetime decline was analyzed, and its lifetime health value was identified as falling below 30 within the next 18 months. The lifetime decline slope represents the rate of decrease in lifetime health value per unit time. Combined with the low-temperature fatigue coefficient of 0.768, the thermal anomaly diffusion rate of 3.3 cm / h, and the local temperature difference of 12℃, it was determined that the module has a risk of continuous thermal damage. It was further predicted that the lifetime health value of the module will drop to 22 in the 22nd month, reaching a severe failure state. The critical time point is the predicted time node when the module is expected to experience severe performance degradation or functional failure. Finally, module PV-23 was marked as a critical failure module, and the corresponding critical time point was output.

[0068] In this embodiment, step S6 includes the following steps: Identify the location coordinates of the critical failure component; Based on the critical time points, risk levels are classified to obtain the risk levels of different components; Based on the location coordinates, the risk levels are spatially distributed statistically to construct a risk distribution map; Based on the risk distribution map, large-area failure propagation prediction is performed to generate failure propagation probability; Dynamic early warning is generated based on the failure propagation probability, resulting in graded early warning results. Based on the results of the graded early warning, adaptive operation and control processing is triggered to complete the intelligent monitoring operation.

[0069] In this embodiment, the position coordinates of the critically failed component are identified; spatial coordinate mapping data of the component is established based on UAV inspection positioning data, component array number, and component installation location information; the position coordinates are the two-dimensional or three-dimensional spatial position parameters of the component in the photovoltaic array; the critically failed component PV-23 is located and identified as being in the 5th row and 12th column, with corresponding spatial coordinates of X=125m and Y=48m; at the same time, the position is corrected by combining the UAV GPS positioning error and the component installation offset, and the positioning error is controlled within ±0.3m; the distribution information of adjacent components around the component is further extracted, and it is identified that there are 6 medium-risk components and 2 high-risk components around it; adjacent components are surrounding components that have a direct spatial connection with the target component; finally, a spatial location dataset of the critically failed component is established.

[0070] The critical time point is the predicted time when a component is expected to experience severe performance degradation or functional failure. Component PV-23 is predicted to enter a severe failure state after 22 months, and component PV-17 is predicted to enter a severe failure state after 36 months. The remaining lifetime is set as follows: less than 12 months is classified as Level 1 high risk, 12-24 months as Level 2 risk, 24-48 months as Level 3 risk, and more than 48 months as low risk. The risk level is a parameter for classifying the degree of future failure risk of the component. Based on the remaining lifetime of the component and the thermal anomaly diffusion trend, a comprehensive score is calculated, and PV-23 is finally classified as a Level 2 high-risk component, and PV-17 is classified as a Level 3 risk component.

[0071] Spatial clustering analysis was performed on components with different risk levels. Spatial clustering analysis is a method for statistically analyzing the spatial concentration of high-risk components. In the region from row 5 to row 8 of the photovoltaic array, a total of 12 first-level high-risk components and 26 second-level high-risk components were identified, with a high-risk cluster density of 42%. Risk cluster density is the proportion of high-risk components to the total number of components in a unit area. Furthermore, a risk-thermal mapping relationship was established based on the direction of heat diffusion between components and the snow cover distribution trend. High-risk areas were marked in red, medium-risk areas in orange, and low-risk areas in green. The risk distribution map is a visualization map used to display the spatial distribution of components with different risk levels. Finally, a photovoltaic array risk distribution map was generated.

[0072] Failure propagation is the process by which thermal anomalies and failure states of a component continuously spread to surrounding components. The regional failure propagation speed is calculated by combining the thermal trend evolution trajectory, thermal diffusion rate, and snow cover change trend. The thermal diffusion rate in high-risk areas reaches 4.2 cm / h, and the snow cover index continues to rise to 78. The failure propagation impact factor is calculated based on the low-temperature fatigue coefficient of surrounding components and spatial distance. The failure propagation impact factor is an evaluation parameter for the degree of failure impact of the target component on surrounding components. Furthermore, based on the environmental temperature forecast and snow cover weather forecast results for the next 30 days, the regional failure propagation trend is simulated. Finally, it is calculated that the probability of a large-scale inefficient power generation area forming in the region from row 5 to row 8 within the next 60 days reaches 68%. The failure propagation probability is the predicted probability of concentrated failure risk occurring in the region in the future.

[0073] Dynamic early warning is an intelligent early warning method that adjusts the warning level in real time according to the trend of risk changes. It sets the failure propagation probability to be below 30% as a low-level warning, 30% to 60% as a medium-level warning, and above 60% as a high-level warning. The failure propagation probability in the area from row 5 to row 8 reaches 68%, so it is judged as a high-level warning area. The risk growth index is calculated by combining the thermal anomaly propagation speed and the number of high-risk components. The risk growth index is the degree of change of the regional risk level over time. When the risk growth index continues to rise for three consecutive inspection cycles, the warning level is automatically upgraded. The output includes the warning area, warning level, and expected propagation time.

[0074] Based on the tiered early warning results, adaptive operation and control processing is triggered to complete intelligent monitoring operations. Corresponding operation and control strategies are automatically activated according to the tiered early warning results. Adaptive operation and control is an intelligent control mechanism that dynamically adjusts the operation mode of the photovoltaic power station based on the risk level. When a region enters a high-level early warning state, the system automatically shortens the drone inspection cycle from 2 hours to 20 minutes and initiates thermal imaging re-inspection of key areas. Simultaneously, local load isolation is implemented on the branches corresponding to high-risk components to reduce continuous thermal stress on the components. Further scheduling of the energy storage system is used for regional power compensation, controlling regional output power fluctuations within ±5%. Local load isolation involves partially disconnecting the branches where high-risk components are located. Energy storage compensation is a method of compensating for the decline in regional power generation capacity through energy storage devices. Simultaneously, maintenance work orders are generated and pushed to the operation and maintenance terminal, completing the intelligent monitoring and adaptive control of the photovoltaic power station's operating status.

[0075] In this embodiment, an intelligent monitoring system for the operating status of a photovoltaic power station is provided, used to execute the intelligent monitoring method for the operating status of a photovoltaic power station as described above, including: The image acquisition module is used to conduct periodic inspections of photovoltaic arrays in photovoltaic power plants using drones, and to acquire standard image sequences and thermal imaging image sequences. The snow detection module is used to perform snow detection calculations on standard image sequences and construct snow cover index change curves. The temperature rise calculation module is used to calculate the regional temperature rise rate of thermal imaging image sequences to obtain the thermal trend evolution trajectory of the photovoltaic array surface. The risk assessment module is used to calculate and assess the component failure risk based on the thermal trend evolution trajectory and the snow cover index change curve, and to mark high-risk areas. The failure analysis module is used to perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure components and critical time points. The intelligent control module is used to perform dynamic early warning and adaptive operation control based on critical failure components and critical time points, and to complete intelligent monitoring operations.

[0076] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0077] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent monitoring of the operating status of a photovoltaic power station, characterized in that, Includes the following steps: Step S1: Conduct periodic inspections of the photovoltaic array of the photovoltaic power station using drones, and collect standard image sequences and thermal imaging image sequences; Step S2: Perform snow cover detection calculations on the standard image sequence and construct a snow cover index change curve; Step S3: Calculate the regional temperature rise rate of the thermal imaging image sequence to obtain the thermal trend evolution trajectory of the photovoltaic array surface; Step S4: Calculate and identify component failure risk based on the thermal trend evolution trajectory and snow cover index change curve, and mark high-risk areas; Step S5: Perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure modules and critical time points; Step S6: Perform dynamic early warning and adaptive operation control based on critical failure components and critical time points to complete intelligent monitoring operations.

2. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S1 are as follows: Set a fixed inspection cycle, drive the drone to perform low-altitude flight path scanning of the photovoltaic array of the photovoltaic power station, and simultaneously collect dual-modal images to extract visible light images and thermal imaging image sequences of different inspection cycles; The visible light image is used to identify and mark highly reflective areas of snow accumulation. Dynamic exposure correction is performed on the reflective areas, and super-resolution enhancement is performed on the remaining low-light areas to obtain the corrected image; The corrected image is filtered for snow and fog interference and texture details are restored to output a standard image sequence.

3. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S2 are as follows: Frame-by-frame depth visual detection is performed on standard image sequences to extract the snow-covered areas on the surface of the photovoltaic array; Visual semantic segmentation is performed on the snow-covered area to extract the non-snow-covered area; Based on the snow-covered and non-snow-covered areas, the snow-covered area, the proportion of snow-covered area, and the spatial distribution parameters of snow accumulation are calculated to generate the snow cover index. The snow cover index is tracked and calculated over multiple inspection cycles based on standard image sequences to construct a snow cover index change curve.

4. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, Step S3 is as follows: Noise filtering is applied to the thermal imaging image sequence to obtain filtered thermal images; The temperature values ​​of multiple surface regions are calculated and calibrated for filtered thermal imaging, generating temperature calibration values ​​for different locations. Based on the temperature calibration values, a temperature distribution is fitted to construct a temperature cloud map; Perform continuous inspection cycle time-series change analysis on temperature cloud maps to generate dynamic temperature cloud maps; Calculate the regional temperature rise rate from the dynamic temperature cloud map and extract the temperature rise rate of different regions; The thermal development trend evolution of the temperature rise rate in different regions was analyzed to obtain the thermal trend evolution trajectory of the photovoltaic array surface.

5. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S4 are as follows: Based on the thermal trend evolution trajectory, the heat distribution dispersion and gradient change rate of different regions are calculated to generate the temperature difference evolution characteristics of different regions. Based on the snow cover index change curve, the correlation between snow cover shading and local heat accumulation is identified for the temperature difference evolution characteristics, and low temperature failure correlation characteristics are generated. Component failure risk is calculated based on the low-temperature failure correlation characteristics to obtain failure risk values ​​for different regions; The failure risk value is judged according to the preset failure risk threshold, and high-risk areas are marked.

6. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S5 are as follows: For high-risk areas, perform multi-cycle tracking analysis of photovoltaic modules within the area to extract the duration of thermal anomalies of each target module, the spatial diffusion direction and diffusion rate of the abnormal temperature zone, and obtain module information; Obtain the component archive database and extract historical running status; Low-temperature fatigue calculations are performed on component information based on historical operating conditions to obtain the low-temperature fatigue coefficient; The component lifespan decay is calculated based on the low-temperature fatigue coefficient, and a lifespan decay curve is generated. Critical failure points are identified based on lifetime decay curves, and critical failure components and critical time points are marked.

7. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 6, characterized in that, The historical operating status includes the number of years since commissioning, the cumulative number of hours of operation at low temperatures, and the frequency of historical snow cover.

8. The intelligent monitoring method for the operating status of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S6 are as follows: Identify the location coordinates of the critical failure component; Based on the critical time points, risk levels are classified to obtain the risk levels of different components; Based on the location coordinates, the risk levels are spatially distributed statistically to construct a risk distribution map; Based on the risk distribution map, large-area failure propagation prediction is performed to generate failure propagation probability; Dynamic early warning is generated based on the failure propagation probability, resulting in graded early warning results. Based on the results of the graded early warning, adaptive operation and control processing is triggered to complete the intelligent monitoring operation.

9. An intelligent monitoring system for the operating status of a photovoltaic power station, characterized in that, The method for performing intelligent monitoring of the operating status of a photovoltaic power station as described in claim 1 includes: The image acquisition module is used to conduct periodic inspections of photovoltaic arrays in photovoltaic power plants using drones, and to acquire standard image sequences and thermal imaging image sequences. The snow detection module is used to perform snow detection calculations on standard image sequences and construct snow cover index change curves. The temperature rise calculation module is used to calculate the regional temperature rise rate of thermal imaging image sequences to obtain the thermal trend evolution trajectory of the photovoltaic array surface. The risk assessment module is used to calculate and assess the component failure risk based on the thermal trend evolution trajectory and the snow cover index change curve, and to mark high-risk areas. The failure analysis module is used to perform multi-cycle tracking analysis of photovoltaic modules in high-risk areas, identify critical failure points, and mark critical failure components and critical time points. The intelligent control module is used to perform dynamic early warning and adaptive operation control based on critical failure components and critical time points, and to complete intelligent monitoring operations.

Images