Photovoltaic power station monitoring method and system based on multi-source data analysis

Abstract

The present application relates to the technical field of control system, and more particularly to a photovoltaic power station monitoring method and system based on multi-source data analysis, comprising a group string partition unit, which divides a plurality of power generation group strings into a plurality of power generation blocks; an intra-zone comparison unit, which constructs a triggering light condition set for each power generation block based on the light condition differences of different power generation blocks, collects the operation data of a plurality of power generation group strings in the power generation block and performs data comparison when the light condition meets the triggering light condition set corresponding to the power generation block, and determines the abnormal power generation group string; and a visual recognition unit, which acquires the position data of the problem group string when the problem group string is identified, adjusts the orientation of the monitoring camera closest to the problem group string to the problem group string and captures the real-time image, predefines a reference image library, compares and analyzes the real-time image with the corresponding reference image in the reference image library, and determines the abnormal power generation panel.

Description

Technical Field

[0001] This invention relates to the field of control system technology, and specifically to a photovoltaic power plant monitoring method and system based on multi-source data analysis. Background Technology

[0002] Multi-source data analysis refers to the process of collecting information from different types and sources of data, integrating them for comprehensive mining and interpretation. Its core lies in the fact that it does not rely on a single data channel, but rather combines information from multiple sources such as databases, sensors, images, text, and even external systems to build a more comprehensive and accurate understanding. This enables the discovery of problems that cannot be revealed in isolated data, supporting smarter decision-making and operations.

[0003] In existing technologies, most monitoring cameras in photovoltaic power plants use fixed-angle shooting, resulting in a single monitoring direction and limited coverage. The video image data is isolated from operational monitoring data such as current, voltage, and irradiance, making it impossible to achieve intelligent linkage. This data isolation means that the camera can only act as a passive recording device, unable to automatically adjust its angle to focus on abnormal areas according to the power plant's operating status, and also difficult to conduct in-depth fault analysis in conjunction with other data. As a result, its comprehensive effectiveness is not fully utilized, and it has certain defects. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a photovoltaic power plant monitoring method and system based on multi-source data analysis, which can effectively solve the problem that the monitoring equipment is difficult to link with other equipment for anomaly identification in the existing technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention provides a photovoltaic power plant monitoring system based on multi-source data analysis, comprising at least:

[0007] A string partitioning unit refers to a string of solar panels consisting of multiple solar panels, which is then divided into multiple power generation blocks.

[0008] Meteorological monitoring data within the photovoltaic power station area was acquired and dust storm data was extracted. The impact of different dust storm data on dust deposition was analyzed, and the dust storm data with the greatest impact on dust deposition was selected and recorded as extreme simulation data.

[0009] Based on the installation location and angle of the power generation string, the dust accumulation index corresponding to each power generation string under extreme simulation data is calculated. Based on the dust accumulation index, the photovoltaic power station area is divided into multiple block areas with dust accumulation differences less than a set threshold. Each block area corresponds to a power generation block.

[0010] The intra-region comparison unit constructs a set of triggering illumination conditions for each power generation block based on the differences in illumination conditions in different power generation blocks. When the illumination conditions meet the set of triggering illumination conditions corresponding to the power generation block, it collects the operating data of multiple power generation strings within the power generation block and compares the data to identify abnormal power generation strings.

[0011] The visual recognition unit, when it identifies a problematic sequence of data, acquires the location data of the problematic sequence, adjusts the head of the monitoring camera closest to the problematic sequence to face the problematic sequence and captures a real-time image. It has a preset reference image library, and compares and analyzes the real-time image with the corresponding reference image in the reference image library to determine the abnormal power generation board.

[0012] Furthermore, the process of dividing the block area is as follows:

[0013] A dust accumulation impact threshold is set, and the divided block areas meet the following conditions:

[0014] Condition A: The difference between the average dust accumulation index of two adjacent blocks is greater than or equal to the dust accumulation impact threshold;

[0015] Condition B: The range of dust accumulation index within the same block area is less than the dust accumulation impact threshold.

[0016] Furthermore, the extreme simulation data selection process is as follows:

[0017] The area where the photovoltaic power station is located is designated as the target area. Dust storm data monitored in the target area over the past year are obtained. The dust storm data includes average wind speed, dust storm duration, and dust mass concentration. Multiple dust storms that occurred are designated as simulated dust storms, and each simulated dust storm is assigned a dust storm number. The maximum wind speed of the simulated dust storms is obtained and designated as the target wind speed. Based on the dust storm data, the simulated dust storms are filtered to obtain the simulated dust storms with the greatest dust accumulation effect. The dust storm data corresponding to this simulated dust storm is designated as extreme simulation data.

[0018] Furthermore, the process of screening simulated sandstorms is as follows:

[0019] Multiple simulated dust storm data are obtained. A function of dust deposition rate with respect to wind speed is preset and denoted as the wind speed influence function. The average wind speed of the simulated dust storm is substituted into the wind speed influence function to calculate the corresponding dust deposition rate, which is denoted as the target deposition rate. The product of the target deposition rate, dust storm duration, and dust mass concentration is calculated to obtain the simulated evaluation value. The dust storm data with the largest simulated evaluation value is selected as the extreme simulation data.

[0020] Furthermore, the dust accumulation index is calculated as follows:

[0021] A digital elevation model (DEM) of the photovoltaic power station area is constructed, denoted as the terrain model. The locations of each photovoltaic string are marked on the terrain model, and the corresponding elevation values ​​are obtained. The average elevation within the terrain model is calculated and denoted as the elevation baseline. The difference between the elevation values ​​of the photovoltaic strings and the elevation baseline is calculated to obtain the elevation assessment value. The horizontal tilt angle of each photovoltaic string within the target area is obtained. Based on the horizontal tilt angle of the photovoltaic strings, the horizontal projected area of ​​a single solar panel within the photovoltaic string is calculated and denoted as the deposition area. Substituting these values ​​into the formula... The dust accumulation index M is obtained by calculation, where:

[0022] S represents the deposition area, Vd represents the target deposition rate, C represents the dust mass concentration, z represents the height assessment value, θ represents the horizontal dip angle, and t is the preset deposition time.

[0023] Furthermore, the process of constructing the set of triggering lighting conditions is as follows:

[0024] The power generation block to be analyzed is denoted as the analysis block. The region outline corresponding to the analysis block is drawn in the digital elevation model. The horizontal tilt angle and installation orientation of all solar panels within the region outline are obtained. The rectangular plane corresponding to any solar panel in the digital elevation model is obtained.

[0025] Let the solar altitude angle at any time be β. Calculate the solar irradiance angle of any solar panel under this solar altitude angle and record it as the irradiance angle. Record the sine value corresponding to the irradiance angle as the irradiance. Calculate the average irradiance of all solar panels within the region outline and record it as the average efficiency.

[0026] A preset efficiency threshold is set, and the solar altitude angle intervals where the average efficiency is greater than or equal to the efficiency threshold are recorded as trigger angle intervals. The power generation rate of the solar panel under different light intensities is obtained, and a preset rate threshold is set. The light intensity intervals where the power generation rate is greater than or equal to the rate threshold are recorded as trigger intensity intervals. The trigger angle intervals and trigger intensity intervals are combined to form a trigger light condition set.

[0027] Furthermore, the data collection and comparison process is as follows:

[0028] The illumination intensity and solar altitude angle within the outline of the current analysis block are obtained and recorded as real-time illumination intensity and real-time solar angle. When both real-time illumination intensity and real-time solar angle are within the corresponding interval of the trigger illumination condition set, the output current, DC voltage and backplane temperature range of each generator string within the analysis block are obtained.

[0029] A temperature difference threshold is preset. Generator strings with a backplane temperature difference greater than or equal to the temperature difference threshold are selected as test strings. The average output current difference and average DC voltage difference between the test string and other generator strings are calculated and recorded as current difference value and voltage difference value, respectively.

[0030] There are preset current difference thresholds and voltage difference thresholds. When either the current difference value or the voltage difference value of the string under test is greater than or equal to the corresponding threshold, the string under test is recorded as a problem string.

[0031] Furthermore, the image analysis process is as follows:

[0032] The problem image is recorded as the real-time image of the problem sequence captured by the monitoring camera. The angle parameter of the problem image captured by the monitoring camera is recorded as the problem angle. A reference image library is preset, which contains multiple reference images corresponding to multiple angle parameters. The reference image corresponding to the problem angle is recorded as the comparison image. The problem image and the comparison image are compared to identify the abnormal power board.

[0033] Furthermore, the process for identifying the problematic solar panels is as follows:

[0034] The power generation string corresponding to the problem string in the comparison image is recorded as the comparison string. Based on image recognition technology, multiple solar power panels in the comparison image are identified, and multiple solar power panels corresponding to the comparison string are selected and recorded as the comparison power panels.

[0035] In the comparison image, draw multiple closed contours corresponding to the comparison solar panel. The image within the closed contour area is recorded as the initial region. The comparison image is overlaid with the problem image. Based on the closed contour, the corresponding region image is extracted from the problem image and recorded as the problem region. The initial region and its corresponding problem region are combined to form a comparison combination. Each comparison combination corresponds to a solar panel.

[0036] The similarity between two images in a comparison set is analyzed and calculated. A similarity threshold is preset. When the similarity between two images in any comparison set is less than the similarity threshold, the solar panel corresponding to that comparison set is marked as an abnormal solar panel.

[0037] The photovoltaic power plant monitoring method based on multi-source data analysis, applied to the aforementioned photovoltaic power plant monitoring system based on multi-source data analysis, includes the following steps:

[0038] Step 1: Collect historical dust storm data of the power plant area and filter out extreme simulation data. Combine the digital elevation model of the power plant to calculate the theoretical dust accumulation index of each power generation string under simulated extreme dust conditions. Based on the difference of the theoretical dust accumulation index and the dust accumulation impact threshold, divide all strings into multiple power generation blocks.

[0039] Step 2: A set of trigger lighting conditions is set for each block. When the lighting conditions meet the set of trigger lighting conditions, the system automatically collects the operating data of all power generation strings in the block and performs a horizontal comparison to identify the problematic string.

[0040] Step 3: When a problematic sequence of images is identified, the nearest surveillance camera to the problematic sequence is designated as the target camera. The target camera is then used to capture an image of the problematic sequence. The captured real-time image is then overlaid and compared with the corresponding baseline image in the baseline image library. Specifically:

[0041] The baseline image library was established in the early stages of the power plant's construction. The baseline image library includes the angle parameters and corresponding images of each string of surrounding objects captured by the target camera.

[0042] Step 4: Construct a corresponding comparison group for each solar panel in the problematic string, analyze the similarity between the two images corresponding to the comparison group, filter out abnormal solar panels based on the similarity threshold, assign an identification number to each comparison group, and ensure that the identification number of the comparison group is consistent with the installation number of the corresponding solar panel. Obtain the identification number of the abnormal solar panel and the corresponding solar panel string number and send them to the handheld mobile terminal of the maintenance personnel.

[0043] The technical solution provided by this invention has the following advantages compared with the known prior art:

[0044] 1. This invention can fully utilize the differences in the power generation of each power generation string. By comparing and analyzing these operating data horizontally, it can identify problematic strings whose operating data is significantly different from other power generation strings. Further problem analysis can then be conducted on the problematic strings, improving the efficiency of problem monitoring while enhancing the accuracy of anomaly identification.

[0045] 2. This invention uses a baseline image library from the initial installation stage to compare real-time problem images from the same angle, calculating the image similarity of the photovoltaic panel area to quickly and accurately locate abnormal photovoltaic panels. Compared with traditional manual inspections or fixed monitoring methods, this method significantly improves the efficiency and accuracy of fault detection and reduces operation and maintenance costs through automated processes. In addition, this invention integrates the adjustment of camera angle parameters with other multi-source data, so that the monitoring camera is no longer just a passive recording device, but can perform targeted image acquisition and comparison when photovoltaic panels exhibit abnormal power generation, thus demonstrating significant advantages in the operation and maintenance of photovoltaic power plants in complex environments. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0047] Figure 1 This is an overall module block diagram of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] The present invention will be further described below with reference to embodiments.

[0050] See Figure 1 A photovoltaic power plant monitoring system based on multi-source data analysis, suitable for photovoltaic power plants in Gobi and desert regions, includes at least:

[0051] The string partitioning unit refers to a string of solar panels consisting of multiple solar panels as a power generation string. Each power generation string is assigned a string number to distinguish different power generation strings, allowing for quick location of the specific power generation string based on the string number. Each power generation string is equipped with independent operation monitoring components, including a string current sensor, a string voltage sensor, and multiple temperature sensors, to monitor the output current, DC voltage, and temperature of the back panel of each solar panel of the power generation string, and to obtain the location conditions of each power generation string. Based on the location conditions, multiple power generation strings are divided into multiple power generation blocks. The power generation capacity of the power generation strings within the same power generation block is basically consistent in terms of the influence of location.

[0052] It's important to note that the power generation process of a photovoltaic (PV) power station mainly consists of two parts. First, on the DC side: numerous solar panels are connected in series to form "strings" to boost the voltage. Multiple strings are then connected in parallel to a "combiner box" to collect the current, ultimately transmitting the powerful DC power to the inverter. Therefore, the generating string is the smallest monitoring unit for current and voltage data (since multiple solar panels in the same generating string are connected in series, monitoring the output current and voltage of a single solar panel is not very meaningful). Then, on the AC side: the inverter, as the core, converts the DC power into AC power for everyday use. This voltage is then significantly boosted by a box-type transformer for long-distance transmission, and finally safely connected to the grid via a step-up substation for normal power supply.

[0053] Specifically, the process of dividing power generation blocks is as follows:

[0054] Meteorological monitoring data within the photovoltaic power station area is acquired, and dust storm data is extracted. The impact of dust deposition on different dust storm data is analyzed, and the dust storm data with the greatest impact on dust deposition is selected as extreme simulation data. This data is used to simulate the dust accumulation under extreme dust conditions. The dust storm corresponding to the extreme simulation data causes the most dust deposition. The dust accumulation index corresponding to each power generation string under the extreme simulation data is calculated based on the installation location and angle of the power generation string. This index is used to represent the dust accumulation situation in different areas. A dust accumulation impact threshold is set, and multiple division conditions are set based on the dust accumulation impact threshold. The photovoltaic power station area is divided into multiple blocks where the dust accumulation difference is less than the threshold (i.e., conditions A and B are met). Each block corresponds to a power generation block.

[0055] More specifically, the partitioning conditions (the conditions that the partitioned blocks must meet) include:

[0056] Condition A: The difference between the average dust accumulation index of two adjacent blocks is greater than or equal to the dust accumulation impact threshold;

[0057] Condition B: The dust index range (the maximum dust index minus the minimum dust index) within the same block area is less than the dust impact threshold.

[0058] By dividing the power generation strings into multiple power generation blocks, the difference in power generation of photovoltaic panels within the same power generation block due to dust cover is less than a preset threshold, thereby making the power generation basically consistent. This facilitates the comparison of power generation of multiple power generation strings within the same power generation block and identifies power generation strings with abnormal power generation.

[0059] More specifically, the extreme simulation data screening process is as follows:

[0060] The area where the photovoltaic power station is located is designated as the target area. Dust storm data monitored in the target area over the past year are obtained. The dust storm data includes average wind speed, dust storm duration, and dust mass concentration. Multiple dust storms that occurred are designated as simulated dust storms, and each simulated dust storm is assigned a dust storm number. The maximum wind speed of the simulated dust storms is obtained and designated as the target wind speed. Based on the dust storm data, the simulated dust storms are filtered to obtain the simulated dust storms with the greatest dust accumulation effect. The dust storm data corresponding to this simulated dust storm is designated as extreme simulation data.

[0061] It should be noted that by constructing a dust accumulation model from the extreme values ​​in the historical meteorological monitoring data of the target area, it is possible to simulate the impact of dust accumulation on the power generation of each solar panel in a photovoltaic power station under extreme weather conditions (mainly sandstorms). At this time, the difference in power generation caused by dust covering each solar panel is the most obvious. In other words, the power generation blocks obtained by dividing the area into blocks can cover the range of differences in solar panel power generation under normal meteorological conditions, which helps to compare and analyze multiple power generation strings in different power generation blocks.

[0062] Generally speaking, using different sandstorm weather conditions to simulate and obtain dust accumulation index and divide the area into blocks will produce a variety of division results. However, there is an obvious pattern that as the severity of the simulation conditions increases, the boundaries between different blocks will become more obvious. Therefore, in this invention, the block division scheme obtained by using historical extreme weather conditions for extreme simulation is the scheme with the smallest division unit and the clearest boundaries.

[0063] Furthermore, the process of screening simulated sandstorms is as follows:

[0064] Multiple simulated dust storm data are obtained. A function of dust deposition rate with respect to wind speed is preset and denoted as the wind speed influence function. The average wind speed of the simulated dust storm is substituted into the wind speed influence function to calculate the corresponding dust deposition rate, which is denoted as the target deposition rate. The product of the target deposition rate, dust storm duration, and dust mass concentration is calculated to obtain the simulated evaluation value. The larger the simulated evaluation value, the greater the impact of dust accumulation. The dust storm data with the largest simulated evaluation value is selected as the extreme simulation data.

[0065] It should be noted that after a sandstorm, the amount of dust particles deposited on the surface of a solar panel is mainly related to the dust concentration in the air, the speed at which dust settles to the surface, the duration of the sandstorm, and the surface area of ​​the solar panel. Therefore, under the same subject (the same solar panel), the key data for the difference in dust deposition caused by different sandstorms are the average wind speed (affecting the speed at which dust settles to the surface), the duration of the sandstorm, and the dust mass concentration (corresponding to the dust concentration in the air). By comparing these three sets of data, we can identify the sandstorms that cause the most severe dust deposition and obtain the corresponding extreme sandstorm data, which helps in constructing a dust accumulation model and dividing the area into blocks. A multiphysics model was constructed using the CFD-DEM (Computational Fluid Dynamics-Discrete Element Method) coupling method to simulate the particle collision, rolling, deposition, and resuspension processes under various operating conditions. This allowed us to determine the dust deposition rate under different wind speed conditions and thus fit a function of dust deposition rate with respect to wind speed.

[0066] More specifically, the dust accumulation index calculation process is as follows:

[0067] A digital elevation model (DEM) of the photovoltaic power plant area is constructed, denoted as the terrain model. The locations of each photovoltaic string are marked in the terrain model, and the corresponding elevation values ​​are obtained. The average elevation within the terrain model is calculated and denoted as the elevation baseline. The difference between the elevation values ​​of the photovoltaic strings and the elevation baseline is calculated to obtain the elevation assessment value. The horizontal tilt angle of each photovoltaic string within the target area is obtained (the angle with the horizontal plane, obtained by combining the slope of the installation location and the installation angle of the solar panels). Based on the horizontal tilt angle of the photovoltaic strings, the horizontal projected area of ​​a single solar panel within the photovoltaic string is calculated and denoted as the deposition area. This area is then substituted into the formula. The dust accumulation index M is obtained by calculation, where:

[0068] S represents the deposition area (in square meters), Vd represents the target deposition rate (in meters per second), C represents the dust mass concentration (in grams per cubic meter), z represents the height assessment value (a dimensionless value), θ represents the horizontal inclination angle, and t is the preset deposition time (set by staff based on experience).

[0069] This formula reflects the relationship between dust deposition and various influencing factors. It can be used to estimate the amount of dust deposition caused by sandstorms on the surface of solar panels, and then differentiate the solar panels of different strings according to the impact of dust deposition (on power generation).

[0070] It should be noted that the construction of digital elevation models is an existing technology that can be generated through UAV ground scanning, and will not be elaborated further here. Due to the influence of topography and soil desertification, solar panels in photovoltaic power stations located in Gobi and desert areas are particularly susceptible to dust accumulation, especially after sandstorms. Dust accumulation affects the power generation of photovoltaic panels, and the degree of impact varies depending on the terrain and location of the photovoltaic panels.

[0071] It is worth noting that during a sandstorm, strong winds carry high concentrations of sand and dust particles across the surface of the solar panels. This process primarily involves two settling mechanisms: turbulent diffusion and inertial settling. High-speed airflow generates complex turbulence on the panel surface, and these vortices transport particles from the mainstream area to the vicinity of the panel surface. Simultaneously, particles with a certain mass, due to inertia, cannot completely follow the airflow around the panel surface and will directly impact and adhere to it. Gravity settling: For larger particles, their own weight overcomes air resistance and they fall naturally, eventually depositing on the inclined panel surface.

[0072] The intra-region comparison unit constructs a set of triggering illumination conditions for each power generation block based on the differences in illumination conditions in different power generation blocks. If the current illumination conditions meet the set of triggering illumination conditions within the power generation block, the corresponding power generation block is recorded as the target block (referring to the power generation block whose illumination conditions meet the set of triggering illumination conditions). The unit collects and compares the operating data of multiple power generation strings within the target block to identify abnormal power generation strings.

[0073] Specifically, the process of constructing the set of triggering lighting conditions is as follows:

[0074] The power generation area to be analyzed is designated as the analysis area. The corresponding region outline of the analysis area is drawn in the digital elevation model (DEM). The horizontal tilt angle and installation orientation of all solar panels within the region outline are obtained, thus obtaining the rectangular plane corresponding to any solar panel in the DEM. Let the solar altitude angle at any given time be β. Calculate the angle of sunlight irradiation for any solar panel under this solar altitude angle, denoted as the illuminance angle, which is the angle between the sunlight and the rectangular plane (obtainable through geometric analysis given the direction of the sunlight and the position of the rectangular plane). The sine value corresponding to the illuminance angle is denoted as the illuminance. The average illuminance of all solar panels within the region outline is calculated and denoted as the average efficiency. A preset efficiency threshold is established. The solar altitude angle interval where the average efficiency is greater than or equal to the efficiency threshold is denoted as the trigger angle interval. The power generation rate of the solar panels under different light intensities is obtained. A preset rate threshold is established. The light intensity interval where the power generation rate is greater than or equal to the rate threshold is denoted as the trigger intensity interval. The trigger angle interval and the trigger intensity interval form the trigger illuminance condition set.

[0075] It should be noted that by constructing a set of triggering illumination conditions, the timing for the monitoring system to collect and analyze the operating data of the power generation strings within the block can be determined. This ensures that the operating data collected under the specified conditions can fully reflect the differences in power generation between different power generation strings, thereby identifying abnormal strings. Data collected outside the time conditions corresponding to the triggering illumination condition set is prone to errors due to poor illumination conditions of the solar panels.

[0076] Specifically, the data collection and comparison process is as follows:

[0077] The illuminance and solar altitude angle within the outline of the current analysis block are recorded as real-time illuminance and real-time solar angle. When both real-time illuminance and real-time solar angle are within the corresponding interval of the trigger illuminance condition set, the output current, DC voltage, and backplane temperature range (i.e., the backplane temperature range of each solar panel within the same power generation string) of each power generation string in the analysis block are obtained. A temperature difference threshold is preset, and power generation strings with a backplane temperature range greater than or equal to the temperature difference threshold are selected as strings to be inspected. The average output current difference and average DC voltage difference between the strings to be inspected and other power generation strings are calculated and recorded as current difference value and voltage difference value, respectively. A current difference threshold and a voltage difference threshold are preset (corresponding to current difference value and voltage difference value, respectively). When either the current difference value or the voltage difference value of the strings to be inspected is greater than or equal to the corresponding threshold, the strings to be inspected are recorded as problem strings.

[0078] It should be noted that the current difference value The calculation formula is ,in This indicates the output current of the string under test. This represents the output current of other power generation strings, where n is the sequence number of the other power generation strings, n = 1, 2, ..., m, and m is the total number of other power generation strings. Abnormal backsheet temperature is a significant indicator of power generation failure in solar panels. For example, particulate matter can form a shading layer on the panel surface (hot spot effect). If any solar panel in a power generation string exhibits an abnormal backsheet temperature, further analysis is needed to determine if there are other abnormal operating data for that solar panel.

[0079] The operational data collected under the triggering illumination condition set can fully reflect the differences in the power generation of each generator string. By comparing and analyzing this operational data horizontally, it is possible to identify problematic generator strings whose operational data is significantly different from other generator strings. Further analysis can then be conducted on these problematic generator strings, improving both the efficiency of problem monitoring and the accuracy of anomaly identification.

[0080] When a problematic sequence is detected, the visual recognition unit acquires the location data of the problematic sequence, adjusts the monitoring camera closest to the problematic sequence so that the monitoring camera faces the problematic sequence, and performs image analysis on the real-time image captured by the monitoring camera to identify abnormal power panels (i.e., power panels with obvious visual abnormalities, such as power panels that are broken, tilted, or overturned).

[0081] It should be noted that the adjustment of the monitoring camera is an existing technology, which will not be elaborated on here. It can be adjusted manually or by setting a set of corresponding shooting angle parameters for each generator string. When the shooting angle parameters corresponding to the generator string are adjusted, the monitoring camera can capture a complete image of the generator string.

[0082] Specifically, the image analysis process is as follows:

[0083] Real-time images of problematic sequences captured by surveillance cameras are recorded as problematic images. The angle parameters of the surveillance cameras when capturing problematic images (i.e., the angle between the camera and the horizontal plane and the axis; each set of angle parameters determines a unique shooting direction) are recorded as problematic angles. A baseline image library is preset, containing multiple baseline images corresponding to multiple angle parameters (the baseline images are collected in the early stages of solar panel installation). The baseline image corresponding to the problematic angle is obtained and recorded as a comparison image. The problematic images are compared with the comparison images to identify abnormal solar panels.

[0084] It should be noted that since the installation position of the generator string is fixed, the angle parameters for taking a complete image of the generator string remain unchanged. In other words, the shooting angle parameters of the problematic string should have been determined in the early stage of building the reference image library, and there should also be a corresponding reference image for these shooting angle parameters.

[0085] More specifically, the process for identifying problematic solar panels is as follows:

[0086] The solar panels corresponding to the problematic string in the comparison image are denoted as the comparison string. Multiple solar panels in the comparison image are identified based on image recognition technology, and the solar panels corresponding to the comparison string are selected and denoted as the comparison panels. Multiple closed contours corresponding to the comparison panels are drawn in the comparison image, and the images within the closed contour areas are denoted as the initial regions. The comparison image is overlaid with the problem image, and the corresponding regions are cropped from the problem image based on the closed contours and denoted as the problem regions. The initial regions and their corresponding problem regions are combined into comparison combinations, and each comparison combination corresponds to one solar panel. The similarity between the two images in the comparison combination is analyzed and calculated. A similarity threshold is preset. When the similarity between the two images in any comparison combination is less than the similarity threshold, the solar panel corresponding to the comparison combination is marked as an abnormal solar panel.

[0087] It should be noted that image recognition technology and image similarity algorithms are existing technologies, and will not be elaborated on here. It is worth noting that the image similarity algorithm used in this system needs to consider the texture features of the image, rather than just the color of the pixels, so as to avoid similarity errors caused by dust covering (in which case the overall brightness of the two images is different, but the main texture and graphic shape are consistent).

[0088] By comparing the baseline image library established during the initial installation phase with real-time problem images from the same angle, the similarity of images in the photovoltaic panel area is calculated to quickly and accurately locate abnormal photovoltaic panels. Compared with traditional manual inspections or fixed monitoring methods, this method significantly improves the efficiency and accuracy of fault detection and reduces operation and maintenance costs through automated processes. Compared with existing technologies, this solution integrates the adjustment of camera angle parameters with other multi-source data, enabling the monitoring camera to move beyond being a passive recording device and perform targeted image acquisition and comparison when photovoltaic panels exhibit abnormal power generation. This demonstrates significant advantages in the operation and maintenance of photovoltaic power plants in complex environments.

[0089] It is worth noting that when the surveillance camera cannot identify a specific abnormal power panel, it should be reported that there is an abnormal power panel in the area, guiding staff to conduct further manual investigation, thus making up for the blind spots that exist when relying solely on surveillance cameras for anomaly identification.

[0090] A photovoltaic power plant monitoring method based on multi-source data analysis includes the following steps:

[0091] Step 1: Collect historical dust storm data for the power plant area, select the dust storm data with the greatest impact of dust deposition and record it as extreme simulation data. Combine the extreme simulation data with the power plant's digital elevation model to calculate the theoretical dust accumulation index of each generating unit under simulated extreme dust conditions. Based on the difference in theoretical dust accumulation index and the dust accumulation impact threshold, divide all units into multiple generating blocks so that the degree of dust impact on the units in the same block is basically the same in future real dust weather.

[0092] Step 2: A set of trigger lighting conditions is set for each block. The set of trigger lighting conditions includes the solar altitude angle and the range of light intensity. When the current lighting conditions meet the trigger setting, the system automatically collects the backplane temperature data of all generator strings in the block and performs a horizontal comparison. The generator strings with abnormal backplane temperatures are screened out, and the current and voltage of the generator string are compared and analyzed with those of other generator strings to identify the problematic string.

[0093] Step 3: When a problematic sequence of images is identified, obtain the location data of the problematic sequence, designate the surveillance camera closest to the problematic sequence as the target camera, call the target camera to capture an image of the problematic sequence, and compare the captured real-time image with the corresponding baseline image in the baseline image library, wherein:

[0094] The baseline image library was established in the early stages of the power plant's construction. The baseline image library includes the angle parameters and corresponding images of each string of surrounding objects captured by the target camera.

[0095] By using a pre-built baseline image library, the camera is transformed from a passive recording device into an active diagnostic tool, capable of automatically reproducing shooting angles that are completely consistent with historical archives, eliminating image differences caused by perspective and angle distortion, and facilitating subsequent automated image analysis.

[0096] Step 4: Overlay and compare the real-time images collected in Step 3 with the baseline images to construct a corresponding comparison group for each solar panel in the problem string. Analyze the similarity between the two images in the comparison group, filter out abnormal solar panels based on the similarity threshold, assign an identification number to each comparison group, and ensure that the identification number of the comparison group is consistent with the installation number of the corresponding solar panel. Obtain the identification number of the abnormal solar panel and the corresponding power generation string number and send them to the handheld mobile terminal of the maintenance personnel.

[0097] By identifying the abnormal power generation board, the system achieves precise location from "problem string" to "faulty single board," narrowing the scope of on-site inspections for maintenance personnel from dozens of boards to a single board, greatly improving maintenance efficiency. Compared to existing technologies that rely on maintenance personnel to inspect each board in the string one by one on-site, using the naked eye or handheld infrared devices to find hot spots or damage, this method is more efficient and intelligent. By fully utilizing computer vision technology, it can effectively eliminate interference from uniform dust, directly identify fault conditions such as cracks, damage, or tilting, and guide maintenance personnel to carry out targeted repairs.

[0098] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A photovoltaic power plant monitoring system based on multi-source data analysis, characterized in that, include: A string partitioning unit refers to a string of solar panels consisting of multiple solar panels, which is then divided into multiple power generation blocks. Meteorological monitoring data within the photovoltaic power station area was acquired and dust storm data was extracted. The impact of different dust storm data on dust deposition was analyzed, and the dust storm data with the greatest impact on dust deposition was selected and recorded as extreme simulation data. Based on the installation location and angle of the power generation strings, the dust accumulation index corresponding to each power generation string under extreme simulation data is calculated. Based on the dust accumulation index, the photovoltaic power station area is divided into multiple blocks with dust accumulation differences less than a set threshold. Each block corresponds to a power generation block. The dust accumulation index calculation process is as follows: A digital elevation model (DEM) of the photovoltaic power station area is constructed and recorded as a terrain model. The positions of each power generation string are marked in the terrain model, and the altitude values ​​corresponding to the positions of the power generation strings are obtained. The average altitude within the terrain model is calculated and recorded as an altitude benchmark value. The difference between the altitude values ​​of the power generation strings and the altitude benchmark value is calculated to obtain the altitude assessment value. The horizontal tilt angle of each photovoltaic string in the target area is obtained. Based on the horizontal tilt angle of the photovoltaic strings, the horizontal projected area of ​​a single solar panel within the photovoltaic string is calculated and recorded as the deposition area. A deposition time is preset. The dust accumulation index is defined by combining the deposition area, target deposition rate, dust mass concentration, altitude assessment value, and horizontal tilt tangent. The formula for calculating the dust accumulation index M is as follows: ,in: S represents the deposition area, Vd represents the target deposition rate, C represents the dust mass concentration, z represents the height assessment value, θ represents the horizontal dip angle, and t represents the deposition time. The process of dividing the block area is as follows: A dust accumulation impact threshold is set, and the divided block areas meet the following conditions: Condition A: The difference between the average dust accumulation index of two adjacent blocks is greater than or equal to the dust accumulation impact threshold; Condition B: The range of dust accumulation index within the same block area is less than the dust accumulation impact threshold; The intra-region comparison unit constructs a set of triggering illumination conditions for each power generation block based on the differences in illumination conditions in different power generation blocks. When the illumination conditions meet the set of triggering illumination conditions corresponding to the power generation block, it collects the operating data of multiple power generation strings within the power generation block and compares the data to identify the problematic string. The visual recognition unit, when it identifies a problematic sequence of data, acquires the location data of the problematic sequence, adjusts the monitoring camera closest to the problematic sequence to face the problematic sequence and captures a real-time image. It has a preset reference image library, and compares and analyzes the real-time image with the corresponding reference image in the reference image library to determine the abnormal power generation board.

2. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 1, characterized in that, The extreme simulation data selection process is as follows: The area where the photovoltaic power station is located is designated as the target area. Dust storm data monitored in the target area over the past year are obtained. The dust storm data includes average wind speed, dust storm duration, and dust mass concentration. Multiple dust storms that occurred are designated as simulated dust storms, and each simulated dust storm is assigned a dust storm number. The maximum wind speed of the simulated dust storms is obtained and designated as the target wind speed. Based on the dust storm data, the simulated dust storms are filtered to obtain the simulated dust storms with the greatest dust accumulation effect. The dust storm data corresponding to this simulated dust storm is designated as extreme simulation data.

3. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 2, characterized in that, The process of screening simulated sandstorms is as follows: Multiple simulated dust storm data are obtained. A function of dust deposition rate with respect to wind speed is preset and denoted as the wind speed influence function. The average wind speed of the simulated dust storm is substituted into the wind speed influence function to calculate the corresponding dust deposition rate, which is denoted as the target deposition rate. The product of the target deposition rate, dust storm duration, and dust mass concentration is calculated to obtain the simulated evaluation value. The dust storm data with the largest simulated evaluation value is selected as the extreme simulation data.

4. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 3, characterized in that, The process of constructing the set of lighting conditions to trigger the illumination is as follows: The power generation block to be analyzed is denoted as the analysis block. The region outline corresponding to the analysis block is drawn in the digital elevation model. The horizontal tilt angle and installation orientation of all solar panels within the region outline are obtained. The rectangular plane corresponding to any solar panel in the digital elevation model is obtained. Let the solar altitude angle at any time be β. Calculate the solar irradiance angle of any solar panel under this solar altitude angle and record it as the irradiance angle. Record the sine value corresponding to the irradiance angle as the irradiance. Calculate the average irradiance of all solar panels within the region outline and record it as the average efficiency. A preset efficiency threshold is set, and the solar altitude angle intervals where the average efficiency is greater than or equal to the efficiency threshold are recorded as trigger angle intervals. The power generation rate of the solar panel under different light intensities is obtained, and a preset rate threshold is set. The light intensity intervals where the power generation rate is greater than or equal to the rate threshold are recorded as trigger intensity intervals. The trigger angle intervals and trigger intensity intervals are combined to form a trigger light condition set.

5. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 4, characterized in that, The data collection and comparison process is as follows: The illumination intensity and solar altitude angle within the outline of the current analysis block are obtained and recorded as real-time illumination intensity and real-time solar angle. When both real-time illumination intensity and real-time solar angle are within the corresponding interval of the trigger illumination condition set, the output current, DC voltage and backplane temperature range of each generator string within the analysis block are obtained. A temperature difference threshold is preset. Generator strings with a backplane temperature difference greater than or equal to the temperature difference threshold are selected as test strings. The average output current difference and average DC voltage difference between the test string and other generator strings are calculated and recorded as current difference value and voltage difference value, respectively. There are preset current difference thresholds and voltage difference thresholds. When either the current difference value or the voltage difference value of the string under test is greater than or equal to the corresponding threshold, the string under test is recorded as a problem string.

6. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 1, characterized in that, The image analysis process is as follows: The problem image is recorded as the real-time image of the problem sequence captured by the monitoring camera. The angle parameter of the problem image captured by the monitoring camera is recorded as the problem angle. A reference image library is preset, which contains multiple reference images corresponding to multiple angle parameters. The reference image corresponding to the problem angle is recorded as the comparison image. The problem image and the comparison image are compared to identify the abnormal power board.

7. The photovoltaic power plant monitoring system based on multi-source data analysis according to claim 6, characterized in that, The process for identifying abnormal power generation panels is as follows: The power generation string corresponding to the problem string in the comparison image is recorded as the comparison string. Based on image recognition technology, multiple solar power panels in the comparison image are identified, and multiple solar power panels corresponding to the comparison string are selected and recorded as the comparison power panels. In the comparison image, draw multiple closed contours corresponding to the comparison solar panel. The image within the closed contour area is recorded as the initial region. The comparison image is overlaid with the problem image. Based on the closed contour, the corresponding region image is extracted from the problem image and recorded as the problem region. The initial region and its corresponding problem region are combined to form a comparison combination. Each comparison combination corresponds to a solar panel. The similarity between two images in a comparison set is analyzed and calculated. A similarity threshold is preset. When the similarity between two images in any comparison set is less than the similarity threshold, the solar panel corresponding to that comparison set is marked as an abnormal solar panel.

8. A photovoltaic power plant monitoring method based on multi-source data analysis, applied to the photovoltaic power plant monitoring system based on multi-source data analysis as described in any one of claims 1-7, characterized in that, Includes the following steps: Step 1: Collect historical dust storm data of the power plant area and filter out extreme simulation data. Combine the digital elevation model of the power plant to calculate the theoretical dust accumulation index of each power generation string under simulated extreme dust conditions. Based on the difference of the theoretical dust accumulation index and the dust accumulation impact threshold, divide all strings into multiple power generation blocks. Step 2: A set of trigger lighting conditions is set for each block. When the lighting conditions meet the set of trigger lighting conditions, the system automatically collects the operating data of all power generation strings in the block and performs a horizontal comparison to identify the problematic string. Step 3: When a problematic sequence of images is identified, the nearest surveillance camera to the problematic sequence is designated as the target camera. The target camera is then used to capture an image of the problematic sequence. The captured real-time image is then overlaid and compared with the corresponding baseline image in the baseline image library. Specifically: The baseline image library was established in the early stages of power plant construction. The baseline image library includes the angle parameters and corresponding images of each string of surrounding objects captured by the target camera. Step 4: Construct a corresponding comparison group for each solar panel in the problematic string, analyze the similarity between the two images corresponding to the comparison group, filter out abnormal solar panels based on the similarity threshold, assign an identification number to each comparison group, and ensure that the identification number of the comparison group is consistent with the installation number of the corresponding solar panel. Obtain the identification number of the abnormal solar panel and the corresponding solar panel string number and send them to the handheld mobile terminal of the maintenance personnel.

Images