Photovoltaic power station comprehensive management system and method

Abstract

The application discloses a kind of photovoltaic power station comprehensive management system and method, it is related to photovoltaic power generation field, the system is by several function modules composition, including: local temperature benchmark construction module, infrared thermography image is collected, according to photovoltaic cell string physical structure is divided into multiple regions, based on the local temperature benchmark of sub-region in multiple regions is constructed, adopts weighted median filter to combine neighborhood consistency to correct local temperature benchmark, obtains the temperature distribution concentration degree and temperature gradient in sub-region interior, simultaneously generates local temperature benchmark;Self-adapting abnormal preliminary screening module obtains comprehensive full image environment state, based on local temperature benchmark and sub-region interior temperature distribution concentration degree, dynamically generates temperature rise threshold for sub-region, according to temperature rise threshold to sub-region carries out preliminary abnormal screening, obtains candidate abnormal region;Heat flow topological discrimination module is based on temperature gradient and constructs heat flow vector field.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation, specifically to a comprehensive management system and method for photovoltaic power plants. Background Technology

[0002] In photovoltaic power plant operation and maintenance, infrared thermal imaging technology is widely used to detect module anomalies, such as hot spots, microcracks, and poor welding. In practical applications, infrared thermal imaging can sensitively detect a variety of typical faults in photovoltaic modules. For example, hot spots are usually caused by local shadows, cell damage, or current mismatch, and appear as obvious high-temperature areas in thermal images. Microcracks are tiny cracks that appear inside the cells, which may lead to increased resistance or localized heating, and appear as abnormal temperature gradients in infrared images. Welding problems, such as poor soldering, over-soldering, or weak contact of interconnects, can cause overheating at the connection points, which appear as dot-shaped or strip-shaped high-temperature areas in thermal imaging.

[0003] Traditional methods mainly rely on temperature threshold criteria, that is, when the temperature of a certain area is higher than the surrounding area or a set threshold, it is judged as a fault. This method has significant defects. In the event of electrical faults, such as bypass diode failure or micro-short circuit of battery cells, local active heating can indeed be caused, forming high-temperature hot spots. At the same time, external obstructions, such as bird droppings, dust, and leaves, can reduce the local power generation capacity and make the temperature of the area lower than the surrounding area. However, under certain lighting conditions, the edges may experience secondary temperature rise due to current concentration or be misjudged as high-temperature areas in image processing. The logic of judging faults solely by high temperature cannot distinguish between the real internal heat source and the passive heat distribution abnormality caused by obstruction, resulting in a high false alarm rate and serious missed detections. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: A photovoltaic power plant integrated management system includes: The local temperature benchmark construction module acquires infrared thermal imaging images, divides the photovoltaic cell string into multiple regions according to its physical structure, constructs a local temperature benchmark based on the sub-regions of these regions, and uses weighted median filtering combined with neighborhood consistency to correct the local temperature benchmark, the concentration of temperature distribution within the sub-regions, and the temperature gradient. The adaptive anomaly screening module obtains the overall environmental status of the map, dynamically generates a temperature rise threshold for the sub-region based on the local temperature benchmark and the degree of temperature distribution concentration within the sub-region, and performs preliminary anomaly screening on the sub-region based on the temperature rise threshold to obtain candidate anomaly regions. The heat flow topology discrimination module constructs a heat flow vector field based on the temperature gradient, analyzes the heat flow divergence in the central region based on the heat flow vector field, and analyzes the consistency of the heat flow divergence in the central region and the heat flow direction at the boundary to determine whether the candidate abnormal region is an electrical fault.

[0005] Furthermore, the process of dividing the photovoltaic cell string into multiple regions based on its physical structure is as follows: Based on the physical layout of the photovoltaic cell string and the pixel resolution of the infrared thermal imaging image, the pixel region corresponding to the physical structure of the photovoltaic cell string in the infrared thermal imaging image is determined. Each physical structure region of the photovoltaic cell string is divided into several sub-regions, and the low-temperature boundary between the sub-regions in the infrared thermal imaging image is extracted by the edge detection algorithm.

[0006] Furthermore, the process of constructing a local temperature reference is as follows: By removing invalid pixels from each sub-region, valid sub-regions are obtained. The temperature values ​​of all pixels in each valid sub-region are extracted, and their sample mean and standard deviation are calculated. Each pixel is assigned a weight, which decreases Gaussian as its temperature value deviates from the sample mean. The temperature value that minimizes the weighted absolute deviation is the weighted median temperature of the sub-region, which serves as its local temperature benchmark.

[0007] Furthermore, the process of correcting the local temperature reference using weighted median filtering combined with neighborhood consistency is as follows: The system checks whether there is a discontinuity between the local temperature reference of each sub-region and the local temperature reference of the adjacent sub-regions. If a discontinuity exists, the sub-region is merged with its adjacent sub-regions, the joint weighted median temperature is recalculated, and the local temperature reference of the atomic region is updated with the joint weighted median temperature. If there is no discontinuity, the local temperature reference of the atomic region is maintained.

[0008] Furthermore, the process of temperature distribution concentration and temperature gradient within the sub-region is as follows: By statistically analyzing the dispersion of temperature values, the degree of concentration of temperature distribution within a sub-region can be obtained. By comparing the temperature values ​​of each pixel with its neighbors above, below, left, and right, the rate and direction of heat change in the horizontal and vertical directions are estimated, and the temperature gradient is obtained.

[0009] Furthermore, the overall environmental status of the map is obtained based on the sample mean and standard deviation; The temperature rise threshold for a sub-region is dynamically generated by performing histogram statistics on the temperature values ​​of all pixels in the sub-region, dividing it into several temperature intervals of equal width, calculating the pixel ratio of each temperature interval, solving the normalized temperature entropy, and determining the basic threshold offset by using the sample mean and standard deviation as inputs through a pre-calibrated mapping relationship. The temperature rise threshold for the sub-region is then dynamically generated based on the normalized temperature entropy and threshold offset of the sub-region.

[0010] Furthermore, the process of obtaining candidate anomaly regions is as follows: The actual temperature rise of a sub-region is obtained, and the difference between the highest temperature and the local temperature reference is generated. If the difference exceeds the temperature rise threshold of the sub-region, and the absolute value of the difference between the actual temperature rise and the dynamic temperature rise threshold is not less than 2 degrees Celsius, the sub-region is marked as a candidate abnormal region when both conditions are met.

[0011] Furthermore, the process of constructing the heat flux vector field based on the temperature gradient is as follows: Based on the temperature gradient between adjacent pixels in the infrared thermal imaging image, the central difference method is used to estimate the temperature change rate in the horizontal and vertical directions, obtain the two-dimensional gradient vector of each pixel, invert the temperature gradient vector of each pixel to obtain the corresponding heat flow vector, and combine all the heat flow vectors to obtain the heat flow vector field.

[0012] Furthermore, the process of determining whether a candidate abnormal region is an electrical fault is as follows: For candidate anomalous regions in infrared thermal imaging images, a heat flow vector field is constructed; the heat flow divergence is calculated in the central sub-region of the candidate anomalous region to determine whether the heat is in a divergent or convergent state; at the same time, the heat flow vector is sampled on its boundary ring, the directional consistency coefficient and the average flow direction are calculated, and the fault and obstruction areas are determined based on the directional consistency coefficient and the average flow direction.

[0013] A comprehensive management method for photovoltaic power plants includes the following steps: Step 1: Acquire infrared thermal imaging images, divide the photovoltaic cell string into multiple regions based on its physical structure, construct local temperature benchmarks based on sub-regions within these regions, and use weighted median filtering combined with neighborhood consistency correction to adjust the local temperature benchmarks, the concentration of temperature distribution within sub-regions, and the temperature gradient. Step 2: Obtain the overall environmental status of the map. Based on the local temperature benchmark and the degree of temperature distribution concentration within the sub-region, dynamically generate the temperature rise threshold for the sub-region. Perform preliminary anomaly screening on the sub-region based on the temperature rise threshold to obtain candidate anomaly regions. Step 3: Construct a heat flow vector field based on the temperature gradient, analyze the heat flow divergence in the central region based on the heat flow vector field, and analyze the consistency of the heat flow divergence in the central region and the heat flow direction at the boundary to determine whether the candidate abnormal region is an electrical fault.

[0014] The photovoltaic power plant integrated management system and method provided by this invention have the following beneficial effects: (1) This invention constructs a heat flow vector field and analyzes its topology in depth, and realizes the physical discrimination of the nature of abnormal heat state under the condition of single-frame infrared image. By calculating the heat flow divergence in the central region, it determines whether the location is active heating caused by electrical fault or passive cooling caused by obstruction. At the same time, combined with the consistency analysis of the boundary heat flow direction, it verifies whether the heat presents a regular divergence or convergence pattern, avoids the empirical misjudgment logic of traditional high temperature as fault, effectively distinguishes the real electrical defects, such as poor soldering, bypass diode failure and external obstruction, such as bird droppings, dust accumulation caused by temperature anomalies, greatly reduces the false alarm rate and false detection rate, and improves the accuracy of diagnostic results.

[0015] (2) This invention accurately identifies the type of abnormality, outputs structured diagnostic results, distinguishes between electrical faults that require emergency handling and shielding pollution that can be handled in a planned manner, provides priority guidance for operation and maintenance personnel, triggers alarms and links shutdown or maintenance work orders, and avoids further damage to equipment or safety accidents. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a partial system schematic diagram of the present invention; Figure 3 This is a schematic diagram of the overall method of the present invention. Detailed Implementation

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

[0018] Example 1 Please see Figure 1 and Figure 2 Embodiment 1 of this application provides a photovoltaic power plant integrated management system, which includes: The local temperature benchmark construction module acquires infrared thermal imaging images, divides the photovoltaic cell string into multiple regions based on its physical structure, and constructs local temperature benchmarks based on sub-regions within these regions. It employs weighted median filtering combined with neighborhood consistency correction to adjust the local temperature benchmarks, considering the concentration of temperature distribution and temperature gradient within each sub-region. The module is divided into multiple regions. Based on the standard physical layout of photovoltaic modules, such as the common arrangement of 60 or 72 cell strings, and combined with the pixel resolution of infrared thermal imaging images, the pixel area corresponding to each cell string in the image is determined; geometric mapping is performed by pre-setting the module size parameters and the image ratio to achieve the structured division of the entire image.

[0019] Each complete battery string region is further subdivided into several sub-regions, typically small blocks of 3×3 to 5×5 pixels. Each sub-region corresponds to about 1 to 2 battery cells, ensuring that its spatial scale is similar to that of a single battery cell, thereby guaranteeing the local representativeness of temperature analysis.

[0020] During the segmentation process, the boundaries of sub-regions are made to fit as closely as possible to the physical seams between the battery cells to avoid temperature mixing distortion caused by connecting different battery cells. Edge detection algorithms, such as the Canny operator, are used to extract the low-temperature boundaries between battery cells in the image and use this as the basis for segmentation to improve the accuracy of region segmentation.

[0021] The entire infrared image is divided into a series of non-overlapping sub-regions that cover the entire battery string, and each sub-region has independent temperature statistical attributes.

[0022] Acquire infrared thermal imaging images: During normal operation of a photovoltaic power station, inspection equipment equipped with infrared thermal imagers, such as drones, fixed gimbals, or handheld terminals, scan and photograph the target photovoltaic array. Data collection is usually conducted during the daytime when solar irradiance is stable, such as between 10 a.m. and 3 p.m., to ensure that the components are in an effective power generation state, thereby stimulating the temperature rise characteristics of the electrical fault area.

[0023] Infrared thermal imagers need to operate in the long-wave infrared band of 8 to 14 μm. This band is an atmospheric window, which can effectively penetrate the air and accurately capture the thermal radiation of the object surface. After the equipment is in focus, it should be aimed at the surface of the photovoltaic module at a vertical or near-vertical angle, maintaining an appropriate distance, usually 5 to 20 meters, depending on the lens's field of view and resolution. This ensures that a single image completely covers one or more modules, and that each cell occupies at least 3×3 pixels in the image to meet the spatial resolution requirements.

[0024] During the acquisition process, the system automatically records the raw infrared data, including the temperature value corresponding to each pixel, i.e., the radiation temperature. After emissivity correction, a two-dimensional temperature matrix image is formed. Before operation, key parameters need to be set, including emissivity, ambient temperature, relative humidity, and measurement distance. The emissivity of the photovoltaic glass surface is set to 0.85 to 0.92, based on the material's actual measured value or standard value. Ambient temperature and relative humidity are used for atmospheric attenuation compensation. The measurement distance is obtained by inputting the actual shooting distance to assist the thermal imager in radiation correction.

[0025] After image acquisition is completed, the system synchronously records metadata such as timestamp, device location, and component number, and stores the infrared image in a radiometric format, such as RadiometricJPEG or a dedicated .raw format.

[0026] Construct local temperature benchmarks for each spatially divided sub-region: Based on the standard cell arrangement of photovoltaic modules, such as 6×10 or 6×12 arrays and known module size parameters, combined with the resolution and field of view of infrared images, the entire image is divided into several cell string regions. Each cell string region corresponds to a row or column of cells connected in series. Each cell string region is further subdivided into multiple sub-regions, each covering approximately 40 to 60 pixels and corresponding to 1 to 2 physical cell units. During the division, priority is given to cutting along the low-temperature gaps between cells using edge detection algorithms to ensure that the boundaries of sub-regions are aligned with the boundaries of actual cells, avoiding cross-cell mixing. Within each sub-region, exclude pixels that are obviously abnormal, such as those whose temperature is outside a reasonable range, such as those that are 5°C below the ambient temperature or 80°C above the ambient temperature and have no contextual support, or those that are located at the edge of the image and are affected by lens distortion. For all pixel temperature values ​​in each effective sub-region, first calculate their sample mean and standard deviation; then assign a weight to each pixel, which decreases Gaussianly as its temperature deviates from the sample mean, i.e., the closer the pixel is to the average temperature of the region, the higher its weight, and the weight of extreme values ​​is suppressed; based on this, find the temperature value that minimizes the weighted absolute deviation, which is the weighted median temperature of the sub-region, and serves as its local temperature benchmark. Check whether there is a sudden change in the local temperature reference of each sub-region compared with the reference of its adjacent sub-regions above, below, left and right. If the temperature difference exceeds 3℃, and there is a significant discontinuity, it is considered that it may be affected by local shading, reflection or division error. At this time, the sub-region is merged with the surrounding 9 sub-regions, the joint weighted median temperature is recalculated, and the reference of the original region is updated with this value to make the temperature of the space temperature field smooth.

[0027] Calibrate the local temperature reference: For each divided sub-region, the temperature values ​​of all pixels within it are extracted to avoid individual abnormally high or low temperatures, such as sunlight reflection or minor stains, from interfering with the baseline calculation. A weighted median filtering method is used to analyze the overall temperature distribution within the sub-region, assigning higher weights to pixels whose temperatures are close to the region's average level and lower weights to pixels that deviate significantly from it. Based on this, the number of weighted pixels on both sides is basically balanced, and this value is used as the initial local temperature baseline for the sub-region.

[0028] The initial reference is subjected to neighborhood consistency correction. The reference temperature of the current sub-region is checked for abrupt changes in its reference temperature with the reference temperatures of its adjacent sub-regions above, below, left, and right. If the temperature difference exceeds 3 degrees Celsius, it indicates that there may be local shading, edge effects, or segmentation errors. The current region and nine surrounding sub-regions are merged into a correction window, and weighted median filtering is re-executed to generate a new reference temperature. The reference temperature of the original region is then updated to ensure the spatial continuity of the temperature reference in the entire image and to conform to the actual heat distribution law of the photovoltaic module.

[0029] Obtain the degree of temperature distribution concentration and temperature gradient within the sub-region: While completing the baseline construction, the temperature distribution characteristics within the sub-region are analyzed simultaneously. By statistically analyzing the dispersion of temperature values, the degree of concentration is determined: if most pixel temperatures are similar, it indicates a concentrated distribution, which may contain hidden hotspots; if the temperature range is large and uniform, it is in a normal state and is quantified in the form of temperature entropy.

[0030] The spatial trend of temperature change within a sub-region is calculated, i.e., the temperature gradient. By comparing the temperature difference between each pixel and its neighbors above, below, left, and right, the rate and direction of heat change in the horizontal and vertical directions are estimated. The larger the gradient, the more drastic the temperature change in the region, and the more likely there is a heat source or heat sink boundary.

[0031] The process of generating a local temperature reference: By traversing all the sub-regions in the image, for each sub-region, a robust representative temperature value is calculated using a weighted median filtering method based on the temperature distribution of its internal pixels. This suppresses outlier interference and reflects the expected operating temperature of the region under fault-free conditions. Neighborhood consistency correction is then applied to this initial reference: if the reference of a sub-region differs significantly from its surrounding neighboring regions, it is merged with the neighborhood and recalculated to ensure spatial smoothness. After determining the reference for a single sub-region, the reference temperature value is filled into all pixel positions covered by the corresponding sub-region. For example, if a 5×5 pixel sub-region is calculated to have a reference temperature of 48.3℃, then all 25 pixels within that region are assigned a value of 48.3℃ in the reference temperature field.

[0032] Once all sub-region references are determined, each pixel in the infrared thermal imaging image is assigned a local temperature reference value, and the images are stitched together to form a complete local temperature reference.

[0033] The adaptive anomaly screening module obtains the overall environmental status of the map, dynamically generates a temperature rise threshold for the sub-region based on the local temperature benchmark and the degree of temperature distribution concentration within the sub-region, and performs preliminary anomaly screening on the sub-region based on the temperature rise threshold to obtain candidate anomaly regions. The process of dynamically generating temperature rise thresholds for sub-regions is as follows: The global average operating temperature is obtained by calculating the arithmetic mean of all valid pixels in the image, excluding non-battery areas such as borders and supports. This reflects the comprehensive thermal boundary conditions such as current solar irradiance, ambient temperature, and wind speed. The global temperature dispersion is obtained based on the standard deviation of the temperature across the entire image. This dispersion is used to quantify the uniformity of the heat distribution on the array surface. High dispersion usually corresponds to local shading, cloud shadow movement, or uneven component aging, while low dispersion is more common in clear, unshaded conditions. The global average operating temperature and the global temperature dispersion together constitute the comprehensive environmental state vector of the entire image. For each defined sub-region, the corresponding reference temperature value is read from the pre-built local temperature reference. Histogram statistics are performed on the temperature values ​​of all pixels within the sub-region, dividing it into 10 equally wide temperature intervals. The pixel percentage of each interval is calculated, and the normalized temperature entropy is calculated accordingly. This entropy value represents the concentration of temperature distribution within the sub-region. A value close to 0 indicates a highly concentrated temperature, such as sharp hot spots or low-temperature patches, while a value close to 1 indicates a flat and uniform temperature distribution. This indicator effectively characterizes the abnormal thermal tendencies within the sub-region. Electrical faults typically manifest as small-scale high-temperature concentrations, while normal operation or uniform shading shows a more gradual distribution. By integrating macroscopic environmental conditions and microscopic sub-region characteristics, a temperature rise judgment threshold specific to each sub-region is dynamically generated, with adaptive environmental adjustment: using the global average temperature and dispersion as input, the basic threshold offset is determined through a pre-calibrated empirical mapping relationship; through local concentration compensation, if the normalized temperature entropy of the sub-region is lower than 0.3, it indicates high concentration, suggesting the possible presence of a real heat source or cold source. In this case, the system appropriately relaxes the threshold for the sub-region, such as increasing it by 0.5 to 1.0℃, to avoid misjudging it as noise due to steep local temperature rise; conversely, if the entropy value is higher than 0.7, a stricter threshold is adopted to suppress false alarms caused by random fluctuations.

[0034] When constrained by the absolute temperature rise lower limit, regardless of how the dynamic threshold is adjusted, the system forcibly sets an absolute temperature rise lower limit, which is usually 2.0℃. That is, only when the difference between the maximum temperature of the sub-region and the local temperature reference exceeds this lower limit is it allowed to trigger an anomaly flag, preventing small fluctuations in low-temperature environments from being amplified. Each sub-region obtains a personalized temperature rise threshold that matches its macroscopic environment, local thermal structure and physical location.

[0035] The candidate anomaly region is obtained as follows: The process iterates through each sub-region of the image. For the current sub-region, it extracts the temperature values ​​of all valid pixels within it and determines the maximum temperature value. This maximum value represents the most likely hotspot location within the sub-region and is a key observation point for anomaly detection. The local temperature baseline temperature corresponding to the sub-region is read. This value has been obtained in the previous step through weighted median filtering and neighborhood consistency correction, reflecting the expected temperature under normal working conditions. At the same time, a temperature rise threshold is dynamically generated based on the overall image environment and the degree of temperature distribution concentration within the sub-region, and the actual temperature rise of the sub-region is calculated, i.e., the difference between the highest temperature and the local temperature baseline. To determine whether the difference meets two conditions simultaneously, it must exceed the dynamic temperature rise threshold of the sub-region; simultaneously, the absolute temperature rise must not be less than 2 degrees Celsius, so that the criterion can adapt to different weather conditions, such as a higher threshold on sunny days and a lower threshold on cloudy days, and to prevent minor fluctuations in low-temperature environments from being misjudged as faults; if only a single or very few pixels have significantly higher temperatures, while the surrounding pixels have normal temperatures, they are considered isolated noise, which may be caused by reflection, lens defects, or transient interference, and are therefore eliminated; then, it is checked whether the high-temperature phenomenon in the sub-region has spatial continuity; when the high-temperature area covers a certain range, for example, containing at least 3 adjacent high-temperature pixels, it is retained; when the temperature rise amplitude, absolute lower limit, and spatial continuity requirements are met, the sub-region is marked as a candidate abnormal region.

[0036] The heat flow topology discrimination module constructs a heat flow vector field based on the temperature gradient, analyzes the heat flow divergence in the central region based on the heat flow vector field, and then analyzes the heat flow divergence in the central region and the consistency of the heat flow direction at the boundary to determine whether the candidate abnormal region is an electrical fault. Temperature gradient constructs heat flux vector field: Taking the initially screened candidate abnormal regions as the analysis objects, the complete temperature distribution data of the region is extracted from the original infrared thermal imaging image. The two-dimensional temperature scalar field is in pixels, and each pixel corresponds to a real surface temperature value after radiation correction. Before performing heat flow modeling, the system performs edge preservation and noise reduction processing on the temperature field. Since infrared images are easily affected by atmospheric disturbances, lens noise or reflection interference, direct differentiation will amplify the error. Anisotropic diffusion filtering technology is used to smooth random noise while preserving the real temperature change edges, such as the temperature difference boundary at the junction of the battery cell and the solder ribbon. The system calculates the spatial trend of local temperature at each valid pixel location within the candidate region. By comparing the temperature difference between the pixel and its adjacent pixels above, below, left, and right, it estimates the rate of temperature change in the horizontal and vertical directions respectively. It adopts a center difference strategy, that is, using the temperature difference between the left and right points to reflect the horizontal gradient and the temperature difference between the top and bottom points to reflect the vertical gradient, thereby obtaining a two-dimensional directional indicator pointing to the direction of the fastest temperature rise. Based on the fundamental principles of thermodynamics, heat always spontaneously flows from high-temperature regions to low-temperature regions, and its flow direction is opposite to the direction of temperature rise. The calculated direction is reversed to obtain the heat flow direction at that pixel. At the same time, the more drastic the temperature change, that is, the greater the temperature difference between adjacent pixels, the more active the heat transfer, and the higher the corresponding heat flow intensity. Each pixel is assigned a heat flow vector with direction and relative intensity. Spatial smoothing is performed on all vectors, and a weighted neighborhood averaging method is used to suppress directional jumps caused by noise in individual pixels. Secondly, the direction of abrupt flow in uniform temperature regions is corrected to make it consistent with the surrounding mainstream heat flow trend. The lengths of all vectors are normalized to a uniform scale to preserve the relative strength relationship. The resulting heat flow vector field is a spatially continuous vector distribution map covering the entire candidate region. In this map, each location clearly indicates the direction of heat flow and the relative level of activity. If a region is the central heat source, the vector will radiate outwards. If it is the central cooling area, such as a shading point, the vector will converge inwards, reflecting the dynamic characteristics of heat transfer. Analyze the heat dissipation in its central region: By analyzing the spatial source or sink characteristics of heat, electrical faults are distinguished as active heating and passive cooling due to shielding contamination. The focus is on the central and boundary regions of candidate areas, calculating heat flow divergence and directional consistency respectively. Typically, a fixed-size window, such as 5×5 pixels, is selected at the geometric center of the candidate anomaly region to ensure coverage of the most likely core point of the anomaly. In a two-dimensional image, divergence reflects whether the heat flow around a point is divergent or convergent. The discrete flux method is used to calculate the sum of the projections of the heat flow vectors of the four neighborhoods (up, down, left, and right) in the normal direction for each pixel in the central region. If the overall flow is outward, the divergence is positive, indicating that the region is a heat source; if the flow is inward, the divergence is negative, indicating that the region is a heat sink. The divergence values ​​of all pixels in the central region are averaged to obtain a representative central average divergence index. Analysis of heat flux divergence in the central region and consistency of heat flux direction at the boundary: The boundary region is defined as a ring of pixels around the outer edge of the candidate anomaly region, typically a closed loop 1 pixel wide. Several equally spaced points are sampled, such as 8 or 12. The direction of the heat flow vector at each sampling point is extracted, and the average direction of all boundary vectors is calculated using a vector synthesis method. The angular deviation between each vector and the average direction is then calculated. If most vectors point in roughly the same direction, such as all outwards or all inwards, the directional consistency is high; if the directions are chaotic, the consistency is low. The system calculates the mean projection of unit vectors onto the average direction to obtain a directional consistency coefficient between 0 and 1: close to 1 indicates high consistency, and close to 0 indicates complete randomness.

[0037] Determine whether the candidate anomaly region is an electrical fault: The following criteria are used to make a comprehensive judgment based on the central average divergence and the consistency of the boundary direction: If the central divergence is positive and the boundary direction consistency is high with the vectors generally pointing outward, it is determined to be an electrical fault hotspot, where heat is generated internally and diffuses outward; if the central divergence is negative and the boundary direction consistency is high with the vectors generally pointing inward, it is determined to be a shielded or contaminated area with a low central temperature and surrounding heat converging towards it; if the divergence is close to zero or the direction consistency is low, it is considered an uncertain area, which may be caused by edge effects, transient interference, or atypical faults, and is marked as pending verification.

[0038] Example 2 Please see Figure 3 Based on Example 1, Example 2 of this application also provides a comprehensive management method for photovoltaic power plants, including the following specific steps: Step 1: Acquire infrared thermal imaging images, divide the photovoltaic cell string into multiple regions based on its physical structure, construct local temperature benchmarks based on sub-regions within these regions, and use weighted median filtering combined with neighborhood consistency correction to adjust the local temperature benchmarks, the concentration of temperature distribution within sub-regions, and the temperature gradient. Step 2: Obtain the overall environmental status of the map. Based on the local temperature benchmark and the degree of temperature distribution concentration within the sub-region, dynamically generate the temperature rise threshold for the sub-region. Perform preliminary anomaly screening on the sub-region based on the temperature rise threshold to obtain candidate anomaly regions. Step 3: Construct a heat flow vector field based on the temperature gradient, analyze the heat flow divergence in the central region based on the heat flow vector field, and analyze the consistency of the heat flow divergence in the central region and the heat flow direction at the boundary to determine whether the candidate abnormal region is an electrical fault.

[0039] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0040] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A comprehensive management system for photovoltaic power plants, characterized in that, The system includes: The local temperature benchmark construction module acquires infrared thermal imaging images, divides the photovoltaic cell string into multiple regions according to its physical structure, constructs a local temperature benchmark based on the sub-regions of these regions, and uses weighted median filtering combined with neighborhood consistency to correct the local temperature benchmark, the concentration of temperature distribution within the sub-regions, and the temperature gradient. The process of correcting the local temperature reference using weighted median filtering combined with neighborhood consistency is as follows: Check if there is a discontinuity between the local temperature reference of each sub-region and the local temperature reference of the adjacent sub-regions; if there is a discontinuity, merge the sub-region with its adjacent sub-regions, recalculate the joint weighted median temperature, and update the local temperature reference of the atomic region with the joint weighted median temperature; if there is no discontinuity, maintain the local temperature reference of the atomic region. The adaptive anomaly screening module acquires the overall environmental status of the map. Based on a local temperature benchmark and the degree of temperature distribution concentration within a sub-region, it dynamically generates a temperature rise threshold for that sub-region. Based on this threshold, it performs preliminary anomaly screening on the sub-region to obtain candidate anomaly regions. The process of determining the degree of temperature distribution concentration and temperature gradient within a sub-region is as follows: By statistically analyzing the dispersion of temperature values, the degree of concentration of temperature distribution within a sub-region can be obtained. By comparing the temperature value difference between each pixel and its neighbors above, below, left, and right, the rate and direction of heat change in the horizontal and vertical directions are estimated, and the temperature gradient is obtained. The overall environmental status is obtained based on the sample mean and standard deviation. The temperature rise threshold for a sub-region is dynamically generated by performing histogram statistics on the temperature values ​​of all pixels in the sub-region, dividing it into several temperature intervals of equal width, calculating the pixel ratio of each temperature interval, solving the normalized temperature entropy, and determining the basic threshold offset by taking the sample mean and standard deviation as inputs through a pre-calibrated mapping relationship; the temperature rise threshold of the sub-region is dynamically generated based on the normalized temperature entropy of the sub-region and the threshold offset. The heat flux topology discrimination module constructs a heat flux vector field based on the temperature gradient. It analyzes the heat flux divergence in the central region of this vector field, and examines the consistency of the heat flux divergence and boundary heat flux directions to determine whether candidate abnormal regions are electrical faults. Specifically, the determination of whether a candidate abnormal region is an electrical fault is based on a comprehensive judgment of the central average divergence and boundary direction consistency: if the central divergence is positive, and the boundary direction consistency is high with vectors generally pointing outwards, it is determined to be an electrical fault hotspot, where heat is generated internally and diffuses outwards; if the central divergence is negative, and the boundary direction consistency is high with vectors generally pointing inwards, it is determined to be a shading or contaminated area, with a low central temperature and surrounding heat converging towards it; if the divergence is close to zero or the direction consistency is low, it is considered an uncertain region and marked for further review.

2. The photovoltaic power station integrated management system according to claim 1, characterized in that, The process of dividing a photovoltaic cell string into multiple regions based on its physical structure is as follows: Based on the physical layout of the photovoltaic cell string and the pixel resolution of the infrared thermal imaging image, the pixel region corresponding to the physical structure of the photovoltaic cell string in the infrared thermal imaging image is determined. Each physical structure region of the photovoltaic cell string is divided into several sub-regions, and the low-temperature boundary between the sub-regions in the infrared thermal imaging image is extracted by the edge detection algorithm.

3. The photovoltaic power station integrated management system according to claim 2, characterized in that, The process of constructing a local temperature reference is as follows: By removing invalid pixels from each sub-region, valid sub-regions are obtained. The temperature values ​​of all pixels in each valid sub-region are extracted, and their sample mean and standard deviation are calculated. Each pixel is assigned a weight, which decreases Gaussian as its temperature value deviates from the sample mean. The temperature value that minimizes the weighted absolute deviation is the weighted median temperature of the sub-region, which serves as its local temperature benchmark.

4. The photovoltaic power station integrated management system according to claim 1, characterized in that, The process of obtaining candidate anomaly regions is as follows: The actual temperature rise of a sub-region is obtained, and the difference between the highest temperature and the local temperature reference is generated. If the difference exceeds the temperature rise threshold of the sub-region, and the absolute value of the difference between the actual temperature rise and the dynamic temperature rise threshold is not less than 2 degrees Celsius, the sub-region is marked as a candidate abnormal region when both conditions are met.

5. A photovoltaic power station integrated management system according to claim 1, characterized in that, The process of constructing a heat flux vector field based on the temperature gradient is as follows: Based on the temperature gradient between adjacent pixels in the infrared thermal imaging image, the central difference method is used to estimate the temperature change rate in the horizontal and vertical directions, obtain the two-dimensional gradient vector of each pixel, invert the temperature gradient vector of each pixel to obtain the corresponding heat flow vector, and combine all the heat flow vectors to obtain the heat flow vector field.

6. The photovoltaic power station integrated management system according to claim 1, characterized in that, The process of determining whether a candidate anomaly region is an electrical fault is as follows: For candidate anomalous regions in infrared thermal imaging images, a heat flow vector field is constructed; the heat flow divergence is calculated in the central sub-region of the candidate anomalous region to determine whether the heat is in a divergent or convergent state; at the same time, the heat flow vector is sampled on its boundary ring, the directional consistency coefficient and the average flow direction are calculated, and the fault and obstruction areas are determined based on the directional consistency coefficient and the average flow direction.

7. A comprehensive management method for photovoltaic power plants, characterized in that, The steps include the following: Step 1: Acquire infrared thermal imaging images and divide the photovoltaic cell string into multiple regions based on its physical structure. Construct local temperature benchmarks based on sub-regions within these regions. Use weighted median filtering combined with neighborhood consistency correction to adjust the local temperature benchmarks, the concentration of temperature distribution within sub-regions, and the temperature gradient. The process of using weighted median filtering combined with neighborhood consistency correction to adjust the local temperature benchmarks is as follows: Check for discontinuities between the local temperature benchmarks of each sub-region and those of adjacent sub-regions. If discontinuities exist, merge the sub-region with its adjacent sub-regions, recalculate the joint weighted median temperature, and update the local temperature benchmark of the sub-region using the joint weighted median temperature. If no discontinuities exist, maintain the local temperature benchmark of the sub-region. Step 2: Obtain the overall environmental status of the entire image. Based on the local temperature benchmark and the concentration of temperature distribution within the sub-region, dynamically generate a temperature rise threshold for the sub-region. Perform preliminary anomaly screening on the sub-region based on the temperature rise threshold to obtain candidate anomaly regions. The process of determining the concentration of temperature distribution and temperature gradient within the sub-region is as follows: Obtain the concentration of temperature distribution within the sub-region by statistically analyzing the dispersion of temperature values; estimate the rate and direction of heat change in the horizontal and vertical directions by comparing the temperature values ​​of each pixel with its upper, lower, left, and right neighbors to obtain the temperature gradient; obtain the overall environmental status of the entire image based on the sample mean and standard deviation; dynamically generate a temperature rise threshold for the sub-region by performing histogram statistics on the temperature values ​​of all pixels within the sub-region, dividing it into several equally wide temperature intervals, calculating the pixel proportion of each temperature interval, solving the normalized temperature entropy, and determining the basic threshold offset using the sample mean and standard deviation as inputs through a pre-calibrated mapping relationship; dynamically generate the temperature rise threshold for the sub-region based on the sub-region's normalized temperature entropy and threshold offset. Step 3: Construct a heat flow vector field based on the temperature gradient. Analyze the heat flow divergence in the central region of the heat flow vector field, and analyze the consistency of the heat flow divergence in the central region and the heat flow direction at the boundaries to determine whether candidate abnormal regions are electrical faults. Specifically, the determination of whether candidate abnormal regions are electrical faults is based on a comprehensive judgment of the central average divergence and the consistency of the boundary direction: if the central divergence is positive, and the boundary direction consistency is high with vectors generally pointing outwards, it is determined to be an electrical fault hotspot, where heat is generated internally and diffuses outwards; if the central divergence is negative, and the boundary direction consistency is high with vectors generally pointing inwards, it is determined to be a shading or contaminated area, with a low central temperature and surrounding heat converging towards it; if the divergence is close to zero or the direction consistency is low, it is considered an uncertain region and marked as requiring further verification.

Images