A control method and system of a photovoltaic cleaning detection assembly based on infrared perception

Abstract

The present application relates to the technical field of photovoltaic equipment, and particularly relates to a control method and system of a photovoltaic cleaning detection assembly based on infrared sensing, comprising: starting a walking mechanism, controlling a front infrared sensing unit, a visible light collecting unit and an environment sensing unit to obtain infrared thermal image data, surface texture data and environment parameter data of a photovoltaic panel to be cleaned; calibrating the infrared thermal image data according to the environment parameter data, identifying a pollution area and grading in combination with the surface texture data, and generating a pollution distribution map; configuring lifting parameters and motion parameters for a plurality of independent cleaning execution units respectively according to the pollution distribution map, and controlling the operation of the units; controlling a rear infrared sensing unit to obtain infrared thermal image data after cleaning, and verifying the cleaning effect and identifying abnormal hot spots by differentiating the data before and after cleaning; matching the abnormal hot spots with a fault feature library, completing fault diagnosis, and transmitting relevant data to a cloud platform. The present application takes into account the cleaning pertinence and fault detection function, and can improve the operation and maintenance efficiency and reliability of photovoltaic equipment.

Description

Technical Field

[0001] This invention relates to the technical field of photovoltaic equipment, and in particular to a control method and system for a photovoltaic cleaning detection component based on infrared sensing. Background Technology

[0002] With the acceleration of the global energy transition, photovoltaic power generation has achieved sustained and rapid growth in installed capacity due to its clean and renewable advantages. During use, dust, bird droppings, pollen, and other pollutants deposited on the surface of photovoltaic panels can significantly reduce their photoelectric conversion efficiency. Therefore, regular cleaning of the surface is necessary to ensure the power generation benefits of photovoltaic power plants.

[0003] In photovoltaic power plants, the distribution of pollutants on photovoltaic panels varies across different areas due to environmental factors. For example, areas near roads tend to accumulate dust, areas with surrounding vegetation are prone to pollen and fallen leaves, and high-mounted panels are susceptible to bird droppings. While existing photovoltaic cleaning devices attempt to integrate sensors, their functions are limited to basic obstacle avoidance or simple pollution identification, failing to achieve quantitative classification of pollution levels. This results in a significantly inefficient cleaning strategy. In lightly polluted areas, over-cleaning causes excessive wear and tear on the photovoltaic panels and brushes, increasing energy consumption and material replacement costs. In heavily polluted areas, insufficient cleaning affects power generation efficiency. Furthermore, fault diagnosis of photovoltaic panels still relies on independent infrared detection equipment, which not only increases equipment procurement and maintenance costs but may also lead to the detection results failing to reflect the true performance status of the photovoltaic panels after cleaning due to the time interval between two cleaning operations. Summary of the Invention

[0004] This invention provides a control method and system for a photovoltaic cleaning detection component based on infrared sensing, which integrates differentiated cleaning of photovoltaic panels, verification of cleaning effect, and fault diagnosis, and can effectively solve the problems in the background art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a control method for a photovoltaic cleaning detection module based on infrared sensing, comprising:

[0006] The walking mechanism is activated, and the front infrared sensing unit, visible light acquisition unit and environmental sensing unit are controlled to work synchronously to acquire infrared thermal image data, surface texture data and environmental parameter data of the photovoltaic panel to be cleaned.

[0007] Radiation calibration is performed on infrared thermal image data based on environmental parameter data. Contaminated areas are identified and contamination levels are calculated by combining surface texture data, generating a contamination distribution map of the photovoltaic panels to be cleaned. The contamination level is determined by a weighted average of the infrared radiation temperature difference and the area proportion of the contaminated area, calculated using the following formula:

[0008] ;

[0009] in, Pollution level, This is the temperature difference weighting coefficient. This is the area proportion weighting coefficient, and ; The infrared radiation temperature difference between the contaminated area and the clean area. To be the minimum identifiable temperature difference threshold, The maximum temperature difference threshold; The area of ​​the contaminated zone. This represents the total area of ​​the region detected in a single test.

[0010] Based on the pollution levels of each area in the pollution distribution map, lifting and motion parameters are configured for multiple independent cleaning execution units; the lifting parameter is the descent height of the cleaning execution unit, and the motion parameter is the rotation speed of the cleaning execution unit.

[0011] Multiple independent cleaning execution units are controlled to operate according to their respective lifting and motion parameters to complete the differentiated cleaning of the photovoltaic panels to be cleaned.

[0012] The control unit acquires infrared thermal image data of the photovoltaic panel after cleaning, calibrates it, and performs differential calculation with the infrared thermal image data before cleaning. The cleaning effect is verified and abnormal hot spots are identified based on the differential result.

[0013] Features are extracted from abnormal hot spots, and the extracted features are matched with a preset fault feature library to generate fault diagnosis results.

[0014] The pollution distribution map, elevation parameters, motion parameters, cleaning effect data, and fault diagnosis results are uploaded to the cloud platform.

[0015] In conjunction with the first aspect, in one possible design, the pollution level is positively correlated with the descent height and negatively correlated with the rotation speed.

[0016] In conjunction with the first aspect, in one possible design, the difference calculation specifically involves:

[0017] The infrared thermal image data of the same location before and after cleaning are used to calculate the temperature difference pixel by pixel to generate a differential thermal map. When the temperature difference value of a certain area in the differential thermal map is greater than the preset cleaning threshold, the area is determined to be incompletely cleaned and marked as a stubborn stain area.

[0018] In conjunction with the first aspect, in one possible design, abnormal hotspot identification specifically involves:

[0019] The persistent hot spot area outside the stubborn stain area is extracted from the differential thermal map. The shape feature, area feature and temperature gradient feature of the area are obtained. The features are matched with the diode failure, EVA delamination and cell crack features in the preset fault feature library. When the matching degree is greater than the preset threshold, the corresponding area is determined to have photovoltaic panel fault.

[0020] In conjunction with the first aspect, in one possible design, environmental parameter data includes ambient temperature, humidity, and irradiance. The infrared thermal image data is then calibrated based on this environmental parameter data using the following calibration formula:

[0021] ;

[0022] in, The actual temperature after calibration. To measure temperature, For ambient temperature, For ambient humidity, Irradiance, , , This is the environmental calibration coefficient.

[0023] In conjunction with the first aspect, in one possible design, multiple independent cleaning execution units are arranged in two staggered rows, with the projection positions of the front and rear cleaning execution units not overlapping and covering the entire cleaning area.

[0024] In conjunction with the first aspect, in one possible design, the cloud platform generates an operation and maintenance report based on the uploaded data. The operation and maintenance report includes statistics on pollution distribution, cleaning effect assessment, and statistics on fault location and type. At the same time, it predicts the performance degradation trend of photovoltaic panels based on historical data.

[0025] Secondly, the present invention also provides a control system for a photovoltaic cleaning detection module based on infrared sensing, comprising:

[0026] The multi-source sensing module includes a front infrared sensing unit, a visible light acquisition unit, an environmental sensing unit, and a rear infrared sensing unit. It is used to collect infrared thermal image data, surface texture data, environmental parameter data, and infrared thermal image data of the photovoltaic panel to be cleaned, and transmit the collected raw data to the data processing module.

[0027] The data processing module receives raw data, performs radiation calibration on infrared thermal image data based on environmental parameter data, identifies contaminated areas and calculates the contamination level by combining surface texture data, and generates a contamination distribution map of the photovoltaic panels to be cleaned. The contamination level is determined by calculating the weighted value of the infrared radiation temperature difference and area ratio of the contaminated area. The module is also used to perform differential calculation and abnormal hot spot matching on the infrared thermal image data before and after cleaning, and transmits the processing results to the control decision module.

[0028] The control decision module is used to generate differentiated cleaning strategies based on the pollution distribution map output by the data processing module, send corresponding control commands to the cleaning execution module, and generate cleaning effect verification commands and fault diagnosis commands based on the differential results and hot spot matching results output by the data processing module.

[0029] The cleaning execution module is used to receive control commands from the control decision module and convert them into drive signals to control multiple sets of independent external cleaning execution units to complete the corresponding cleaning actions.

[0030] The data transmission module is used to upload the data generated by the data processing module and the control decision module to the cloud platform.

[0031] In conjunction with the second aspect, in one possible design, each set of independent cleaning execution units includes a lifting mechanism, a drive mechanism, and a cleaning brush. The lifting mechanism is used to drive the drive mechanism and the cleaning brush to adjust the height relative to the photovoltaic panel surface, and the drive mechanism is used to drive the cleaning brush to rotate.

[0032] The technical solution of this invention can achieve the following technical effects:

[0033] By combining a front-end infrared sensing unit and a visible light acquisition unit with environmental parameter calibration, polluted areas can be accurately identified, and the pollution level can be calculated by weighting temperature difference and area ratio to generate a pollution distribution map. By configuring differentiated lifting and motion parameters for multiple independent cleaning execution units based on this map, the problems of over-cleaning and under-cleaning in traditional extensive cleaning can be avoided, reducing the cost of consumable replacement and energy consumption in photovoltaic panel maintenance. The rear-end infrared sensing unit acquires thermal image data after cleaning, and the difference between the data and the data before cleaning can directly determine whether the cleaning is thorough and mark stubborn stains. At the same time, by identifying abnormal hot spots in the differential thermal map and matching them with a preset fault feature library, rapid fault diagnosis can be achieved. The cleaning, effect verification, and fault diagnosis functions are integrated into one unit, without relying on separate infrared detection equipment. This saves on equipment procurement and maintenance costs and avoids the detection lag caused by the interval between two operations, thereby improving the continuity and efficiency of photovoltaic power plant operation and maintenance. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1This is a flowchart illustrating the control method for the photovoltaic cleaning detection component based on infrared sensing in this invention.

[0036] Figure 2 This is a schematic diagram of the control system of the photovoltaic cleaning detection component based on infrared sensing in this invention.

[0037] Figure 3 This is a schematic diagram showing the distribution of the infrared sensing unit, visible light acquisition unit, and environmental sensing unit in the control system of the photovoltaic clean detection component based on infrared sensing in this invention.

[0038] Figure 4 This is a schematic diagram of the independent cleaning execution unit structure of the photovoltaic cleaning and testing module.

[0039] Reference numerals: 1. Frame; 2. Front-end rod; 3. Front infrared sensing unit; 4. Visible light acquisition unit; 5. Environmental sensing unit; 6. Rear-end rod; 7. Rear infrared sensing unit; 8. Lifting mechanism; 9. Drive mechanism; 10. Cleaning brush. Detailed Implementation

[0040] This application will now be described with reference to the accompanying drawings.

[0041] like Figure 1 As shown, the control method for the photovoltaic cleaning detection module based on infrared sensing of the present invention specifically includes the following steps:

[0042] Step S1: Start the walking mechanism and control the front infrared sensing unit, visible light acquisition unit and environmental sensing unit to work synchronously to acquire infrared thermal image data, surface texture data and environmental parameter data of the photovoltaic panel to be cleaned.

[0043] Step S2: Perform radiation calibration on the infrared thermal image data based on environmental parameter data, identify contaminated areas and calculate the contamination level by combining surface texture data, and generate a contamination distribution map of the photovoltaic panel to be cleaned.

[0044] Step S3: Based on the pollution level of each area in the pollution distribution map, configure lifting and motion parameters for multiple independent cleaning execution units respectively;

[0045] Step S4: Control multiple independent cleaning execution units to operate according to their respective lifting and motion parameters to complete the differentiated cleaning of the photovoltaic panels to be cleaned;

[0046] Step S5: Control the rear infrared sensing unit to acquire infrared thermal image data of the photovoltaic panel after cleaning, calibrate it and perform differential calculation with the infrared thermal image data before cleaning, verify the cleaning effect and identify abnormal hot spots based on the differential result.

[0047] Step S6: Extract features from abnormal hot spots, match the extracted features with a preset fault feature library, and generate fault diagnosis results;

[0048] Step S7: Upload the pollution distribution map, elevation parameters, motion parameters, cleaning effect data, and fault diagnosis results to the cloud platform.

[0049] In this embodiment, the pollution level data before cleaning directly determines the cleaning parameters, and the differential results after cleaning can inversely correct the pollution level calculation model. For example, if a certain area still fails to meet the standard after cleaning with moderate pollution parameters, the pollution level judgment threshold for that area can be adjusted. Simultaneously, fault diagnosis results can mark the decrease in power generation efficiency caused by hardware failure, avoiding ineffective cleaning of faulty panels. Dynamic adaptive capability is achieved through continuous optimization of sensing accuracy and cleaning strategies via data self-feedback without manual intervention. The use of front-end and rear-end infrared sensing units enables pollution identification, cleaning execution, effect verification, and fault diagnosis to be completed within the same workflow and time window, avoiding the problem of distorted detection results caused by dust accumulation on photovoltaic panels or environmental changes during the interval between two cleaning operations, ensuring the accuracy of fault diagnosis. Furthermore, the absence of an additional independent infrared detection device walking mechanism reduces equipment complexity and energy consumption, while also minimizing repeated disturbance to the photovoltaic panels during two operations, reducing the risk of panel damage. Pollution distribution maps can reflect regional environmental differences, cleaning parameter data reflects the optimal cleaning solutions for different types of pollution, and fault diagnosis results reflect the spatial distribution patterns of panel aging or damage. The fusion of these three types of data can form a panoramic picture of power plant operation and maintenance. Based on this panoramic picture, regional predictive operation and maintenance strategies can be formulated, and design defects or power plant layout can be deduced from the fault distribution patterns, thus achieving an upgraded management from single panel operation and maintenance to overall power plant optimization.

[0050] In step S1, the purpose of activating the walking mechanism is to drive the entire photovoltaic cleaning and inspection component to move smoothly along the photovoltaic panel array, thereby realizing continuous inspection and cleaning of large-area photovoltaic panels. If it is only fixed in a certain position, it cannot cover the entire surface of the photovoltaic panel. However, the walking mechanism can move at a constant speed along a preset path through a guide rail or wheel drive structure, ensuring that no inspection or cleaning is missed. The system utilizes a front-end infrared sensing unit, a visible light acquisition unit, and an environmental sensing unit to acquire multi-dimensional data about the photovoltaic panel to be cleaned. The front-end infrared sensing unit, employing an infrared thermal imager, generates a thermal image by capturing the distribution of infrared radiation intensity on the photovoltaic panel surface. The principle is that contaminants hinder the photovoltaic panel's heat absorption and dissipation, creating a temperature difference between the contaminated and clean areas. The more severe the contamination, the more significant the temperature difference. Therefore, infrared thermal image data can intuitively reflect the location of contamination. The visible light acquisition unit, using a high-definition camera, acquires surface texture data by capturing visible light images of the photovoltaic panel surface. This clearly presents details such as the morphology and distribution density of contaminants, aiding in the determination of contamination type and providing a basis for accurate identification of contaminated areas. The environmental sensing unit includes temperature and humidity sensors, irradiance sensors, etc., used to collect environmental parameter data, such as ambient temperature, humidity, and irradiance. Since the measurement accuracy of the infrared thermal imager is affected by environmental interference—for example, changes in ambient temperature affect heat exchange on the photovoltaic panel surface, humidity affects atmospheric absorption of infrared radiation, and irradiance affects the photovoltaic panel's own heating—the environmental parameter data collected by the environmental sensing unit is used to calibrate the infrared thermal image data, ensuring the accuracy of temperature measurements.

[0051] In some embodiments of the present invention, in order to eliminate the interference of environmental factors and obtain more accurate temperature data, the calibration formula is as follows:

[0052] ;

[0053] in, The actual temperature after calibration. To measure temperature, For ambient temperature, For ambient humidity, Irradiance, , , The environmental calibration coefficient is calculated by measuring a standard sample with a known temperature under different environmental conditions, comparing the measured value with the true value, and using methods such as regression analysis, to ensure accurate calibration under various combinations of environmental parameters.

[0054] After calibration, the infrared thermal image data obtained has eliminated environmental interference and can accurately reflect the temperature distribution differences on the photovoltaic panel surface. However, relying solely on temperature information is insufficient to completely distinguish the type of contamination and may mistakenly identify material differences within the photovoltaic panel itself as contamination. Therefore, it is necessary to combine the surface texture data collected by the visible light acquisition unit with image recognition algorithms, such as deep learning-based image segmentation algorithms, to analyze the photovoltaic panel surface and achieve accurate identification of contaminated areas. The specific process of the deep learning-based image segmentation algorithm is as follows:

[0055] Step S2a: Align the calibrated infrared thermal image data and visible light texture data according to pixel position to form a 4-channel input tensor; to enhance the generalization ability of the model, perform data augmentation operations such as rotation, scaling, and contrast adjustment on the data;

[0056] Step S2b: The encoder extracts temperature gradient features from infrared data and texture features from visible light data through convolutional layers, and fuses shallow detail features with deep semantic features through skip connections. The formula is as follows:

[0057] ;

[0058] in As a feature of fusion, Infrared characteristics, It has visible light characteristics; The fusion weights were determined experimentally.

[0059] In step S2c, the decoder maps the fused features back to the original image size using a deconvolution operation, and uses the Softmax function to output the probability that each pixel belongs to the "contaminated" or "clean" category:

[0060] ;

[0061] in, For the model's output logit value for the "pollution" category, when When this happens, the pixel is determined to belong to a contaminated area;

[0062] Step S2d: Perform morphological operations on the segmentation results to remove isolated noise points, and determine the boundaries of the contaminated region through a contour extraction algorithm, finally outputting a complete contaminated region mask.

[0063] Through the above steps, deep fusion and accurate analysis of infrared thermal imaging data and visible light texture data can be achieved, overcoming the limitations of single sensors in complex scenarios and improving the accuracy of polluted area identification. For example, visible light data can effectively supplement the identification of bird droppings with slight temperature differences but obvious texture features; infrared data can ensure that no dust accumulation is missed, even if it is obscured by shadows but has significant temperature differences.

[0064] In some embodiments of the present invention, the calculation of the pollution level needs to comprehensively consider the infrared radiation temperature difference between the polluted area and the clean area, as well as the area ratio of the polluted area, and the formula is as follows:

[0065] ;

[0066] in, Pollution level; This is the temperature difference weighting coefficient. This is the area proportion weighting coefficient, and , and The value of is determined based on the degree of influence of temperature difference and area ratio on photovoltaic power generation efficiency in actual applications. The evaluation effect under different ratios is tested through experiments, and the ratio that can most accurately reflect the impact of pollution on power generation efficiency is selected. The infrared radiation temperature difference between the contaminated area and the clean area; The minimum identifiable temperature difference threshold is set according to the resolution of the infrared thermal imager and the actual detection requirements. The maximum temperature difference threshold is determined based on the maximum temperature difference that the photovoltaic panel may experience under severe pollution conditions. The area of ​​the contaminated zone. The total area of ​​the area to be tested in a single instance is given. Based on the calculated pollution level, a pollution distribution map of the photovoltaic panels to be cleaned is generated. This map clearly marks the location, extent, and corresponding pollution level of each polluted area.

[0067] In step S3, the pollution level can be set as light, moderate, and heavy, corresponding to the range of L values ​​in the pollution level calculation formula. For example, L≤0.3 is light, 0.3<L≤0.6 is moderate, and L>0.6 is heavy. To ensure the pollution removal effect and avoid unnecessary damage to photovoltaic panels and cleaning components, areas with different pollution levels require different cleaning intensities and methods. That is, multiple independent cleaning execution units are configured with the same or different lifting and motion parameters to achieve targeted cleaning.

[0068] Specifically, the lifting parameter refers to the descent height of the cleaning execution unit. Its configuration is based on the required contact depth of the cleaning components in the contaminated area. Specifically, when the contamination level is high, the adhesion between the pollutants and the photovoltaic panel surface is strong, requiring the cleaning execution unit to shorten the distance between itself and the photovoltaic panel surface, allowing for closer contact and enhanced force transmission. When the contamination level is low, a larger distance between the cleaning execution unit and the photovoltaic panel surface is sufficient, allowing for light contact and reducing pressure on the panel surface. The motion parameter refers to the rotation speed of the cleaning execution unit. Its configuration matches the energy input requirements of the contaminated area. Specifically, for high-contamination areas, the motion rate of the cleaning components needs to be reduced to increase friction and extend the action time, using continuous and concentrated cleaning energy to break down stubborn pollutants. For low-contamination areas, the motion rate of the cleaning components can be increased to reduce friction loss and shorten the action time, quickly removing surface pollutants while avoiding over-cleaning. By configuring parameters of multiple independent cleaning execution units independently and collaboratively, each unit only responds to the pollution level signal of its corresponding area, ensuring that areas with different pollution levels can obtain a cleaning intensity that matches its level. This avoids leaving pollutants in high-pollution areas due to insufficient cleaning, and also prevents damage to the photovoltaic panel surface or wear and tear on cleaning components in low-pollution areas due to over-cleaning, ultimately achieving efficient and precise cleaning of the entire area.

[0069] In step S5, when the rear infrared sensing unit acquires infrared thermal image data of the cleaned photovoltaic panel, it must ensure that its acquisition angle, resolution, and sampling frequency are consistent with those of the front infrared sensing unit, and that the acquisition position corresponds one-to-one with the detection area before cleaning, thereby eliminating data deviations caused by equipment differences or spatial misalignment. The infrared thermal image data after cleaning also needs to undergo the same radiation calibration process as before cleaning to ensure that the two sets of data are comparable on a temperature reference. The difference calculation is performed between the infrared thermal image data after cleaning and the radiation-calibrated infrared thermal image data before cleaning, specifically as follows:

[0070] Perform point-by-point subtraction on the temperature values ​​of pixels at the same spatial coordinates (x, y) before and after cleaning to generate a temperature difference matrix. The calculation formula is:

[0071] ;

[0072] in, The calibration temperature after cleaning. This is the calibration temperature before cleaning; once contaminants in a certain area are effectively removed, its thermal conductivity recovers, and the temperature difference in that area will show a significant decrease, typically manifested as... Furthermore, the absolute value is relatively large; if cleaning is not thorough, residual contaminants will cause the temperature difference to decrease only slightly or even close to zero, typically manifesting as... The absolute value is relatively small;

[0073] The differential heat map generated based on the temperature difference matrix can be used to determine the cleaning effect based on a preset threshold. To verify the cleaning effect, this threshold is determined based on the temperature fluctuation range of the photovoltaic panel under standard cleaning conditions; when When the corresponding area is deemed unsatisfactory, it is marked as a stubborn stain area, and its location and extent must be recorded for subsequent adjustments to the secondary cleaning strategy; when At that time, the area was deemed clean and qualified.

[0074] Hot spots caused by contamination will disappear or weaken as the temperature difference decreases after cleaning, while hot spots caused by photovoltaic panel malfunctions are stable and have specific morphologies; persistent high-temperature areas still existing within the cleaned and qualified areas in the differential thermal map, i.e. It is still significantly higher than the temperature in the surrounding normal area, but This indicates that the high temperature was not caused by pollutants, but by a malfunction in the photovoltaic panel itself.

[0075] In some embodiments of the present invention, a preset fault feature library is generated based on feature data from a large number of known fault samples. It includes feature templates for typical fault types, such as diode failure, EVA delamination, and cell cracks. Each template consists of a feature vector and a fault type label. The process of extracting features from abnormal hot spots and matching them with the preset fault feature library to generate fault diagnosis results is achieved through multi-dimensional feature quantization and intelligent matching algorithms to accurately determine the fault type of the photovoltaic panel. The specific process is as follows:

[0076] Step S61: Extract multi-dimensional features from the identified abnormal hotspot regions to form a structured feature vector, including:

[0077] A) Shape characteristics: The aspect ratio of the hot spot is calculated by the minimum bounding rectangle to distinguish between regular and irregular shapes, such as the approximately rectangular hot spot of diode failure and the long strip-shaped hot spot of battery cell crack; the contour complexity describes the degree of edge irregularity. This value varies significantly depending on the type of failure. For example, the edge of the EVA delamination hot spot is relatively smooth, with a ratio close to 1, while the edge of the crack hot spot is rough, with a ratio significantly greater than 1.

[0078] B) Temperature characteristics: Calculate the difference between the average temperature of the hot spot area and the average temperature of the surrounding normal area to reflect the heating intensity of the hot spot; calculate the temperature decay rate from the center to the edge of the hot spot, i.e., the temperature gradient. Hot spots caused by faults usually have characteristic gradient values; for example, the gradient of a diode failure hot spot is steep, while the gradient of an EVA delamination hot spot is gentle.

[0079] C) Spatial distribution characteristics: Statistical analysis of the location distribution of hot spots in the photovoltaic cell array, as well as the relative position of hot spots to the photovoltaic circuit connection path, provides a spatial basis for fault location.

[0080] Step S62: Calculate the matching degree between the feature vector of the hot spot to be identified and each template in the library. The calculation formula is as follows:

[0081] ;

[0082] in, For feature dimension, For the hot spot to be identified 3D eigenvalues The first template 3D eigenvalues; For the first The weights of the features are set according to the distinguishability of the features to fault types.

[0083] Step S63: When the matching degree of a certain fault template is greater than the preset threshold, it is determined that the hot spot to be identified corresponds to the fault type; if multiple templates have matching degrees exceeding the threshold, the template with the highest matching degree is selected as the diagnosis result; if all templates have matching degrees below the threshold, they are marked as "unknown fault" and the manual review process is triggered.

[0084] Step S64: The generated fault diagnosis results may include the precise coordinates of the fault area, fault type, feature matching degree, confidence level divided according to the matching degree value range, and suggested maintenance plan, providing clear maintenance guidance for operation and maintenance personnel.

[0085] In step S7, the pollution distribution map, elevation parameters, motion parameters, cleaning effect data, and fault diagnosis results are uploaded to the cloud platform. The cloud platform can store, analyze, and process this data to generate a detailed operation and maintenance report. The operation and maintenance report may include pollution distribution statistics, cleaning effect assessment, and fault location and type statistics, providing comprehensive operation and maintenance information for photovoltaic power plant managers. Managers can use this information to formulate more reasonable cleaning plans and maintenance schemes, address existing problems in a timely manner, and improve the power generation efficiency and operational reliability of the photovoltaic power plant. At the same time, the cloud platform can also analyze historical data to predict the performance degradation trend of photovoltaic panels, achieving forward-looking operation and maintenance management.

[0086] like Figure 2 As shown, the present invention also provides a control system for a photovoltaic cleaning detection component based on infrared sensing, specifically including:

[0087] The multi-source sensing module includes a front infrared sensing unit, a visible light acquisition unit, an environmental sensing unit, and a rear infrared sensing unit. It is used to collect infrared thermal image data, surface texture data, environmental parameter data, and infrared thermal image data of the photovoltaic panel to be cleaned, and transmit the collected raw data to the data processing module.

[0088] The data processing module receives raw data, performs radiation calibration on infrared thermal image data based on environmental parameter data, identifies contaminated areas and calculates the contamination level by combining surface texture data, and generates a contamination distribution map of the photovoltaic panels to be cleaned. The contamination level is determined by calculating the weighted value of the infrared radiation temperature difference and area ratio of the contaminated area. The module is also used to perform differential calculation and abnormal hot spot matching on the infrared thermal image data before and after cleaning, and transmits the processing results to the control decision module.

[0089] The control decision module is used to generate differentiated cleaning strategies based on the pollution distribution map output by the data processing module, send corresponding control commands to the cleaning execution module, and generate cleaning effect verification commands and fault diagnosis commands based on the differential results and hot spot matching results output by the data processing module.

[0090] The cleaning execution module is used to receive control commands from the control decision module and convert them into drive signals to control multiple sets of independent external cleaning execution units to complete the corresponding cleaning actions.

[0091] The data transmission module is used to upload the data generated by the data processing module and the control decision module to the cloud platform.

[0092] In this embodiment, the infrared sensing unit, visible light acquisition unit, and environmental sensing unit in the multi-source sensing module are distributed as follows: Figure 3 As shown, the multi-source sensing module can comprehensively collect multi-dimensional data before and after photovoltaic panel cleaning; the data processing module transforms the raw data into accurate pollution distribution maps and abnormal hot spot analysis results through core processing such as radiation calibration, pollution identification, and differential calculation; the control decision module generates targeted cleaning strategies and diagnostic instructions based on the analysis results; the cleaning execution module controls multiple independent units to perform differentiated cleaning, taking into account both cleaning effectiveness and equipment protection; and the data transmission module constructs a data closed loop between the terminal and the cloud, supporting operation and maintenance optimization. Each module has a clear division of labor and close collaboration, which can effectively improve the automation and accuracy of photovoltaic cleaning detection, while reducing operation and maintenance costs, providing reliable technical support for the efficient operation of photovoltaic power plants.

[0093] In a specific implementation, as one example, such as Figure 4As shown, each independent cleaning execution unit includes a lifting mechanism, a drive mechanism, and a cleaning brush. The lifting mechanism is used to precisely adjust the distance between the cleaning brush and the photovoltaic panel surface according to control commands to match the cleaning contact intensity requirements of different pollution levels. It typically uses a screw-nut drive structure or an electric push rod structure, converting the rotational motion of the motor into linear motion to achieve height adjustment. To ensure the accuracy of height adjustment, the lifting mechanism can integrate a position feedback element, which can collect the actual height position of the brush in real time and send the data back to the control module to form a closed-loop control, avoiding height adjustment deviations caused by mechanical errors. The drive mechanism is used to output rotational power at different speeds according to control commands to match the requirements of contaminant removal for action time and friction intensity. It typically uses a servo motor or stepper motor with a gearbox. Each independent cleaning execution unit is equipped with a dedicated control interface, which can receive height adjustment commands and speed commands from the cleaning execution module independently, and their actions do not interfere with each other. For example, when two adjacent units correspond to heavily polluted areas and lightly polluted areas respectively, the former can be lowered to a close position through the lifting mechanism and the drive mechanism can run at low speed and high torque, while the latter can be raised to a far position and run at high speed and low torque. When the two units work synchronously, they can apply matching cleaning intensity to different polluted areas at the same time dimension, ultimately achieving differentiated and precise cleaning of the entire photovoltaic panel area.

[0094] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A control method for a photovoltaic cleaning detection module based on infrared sensing, characterized in that, include: The walking mechanism is activated, and the front infrared sensing unit, visible light acquisition unit and environmental sensing unit are controlled to work synchronously to acquire infrared thermal image data, surface texture data and environmental parameter data of the photovoltaic panel to be cleaned. The infrared thermal image data is calibrated based on the environmental parameter data, and contaminated areas are identified and contamination levels are calculated by combining surface texture data to generate a contamination distribution map of the photovoltaic panels to be cleaned. The contamination level is determined by calculating a weighted value of the infrared radiation temperature difference and area ratio of the contaminated area, using the following formula: ； in, Pollution level, This is the temperature difference weighting coefficient. This is the area proportion weighting coefficient, and ; The infrared radiation temperature difference between the contaminated area and the clean area. To be the minimum identifiable temperature difference threshold, The maximum temperature difference threshold; The area of ​​the contaminated zone. This represents the total area of ​​the region detected in a single test. Based on the pollution level of each area in the pollution distribution map, lifting and motion parameters are configured for multiple independent cleaning execution units; the lifting parameter is the descent height of the cleaning execution unit, and the motion parameter is the rotation speed of the cleaning execution unit. The multiple independent cleaning execution units are controlled to operate according to their respective lifting and motion parameters to complete the differentiated cleaning of the photovoltaic panels to be cleaned. The control unit acquires infrared thermal image data of the photovoltaic panel after cleaning, calibrates it, and performs differential calculation with the infrared thermal image data before cleaning. The cleaning effect is verified and abnormal hot spots are identified based on the differential result. Features are extracted from abnormal hot spots, and the extracted features are matched with a preset fault feature library to generate fault diagnosis results. The pollution distribution map, elevation parameters, motion parameters, cleaning effect data, and fault diagnosis results are uploaded to the cloud platform.

2. The control method for a photovoltaic cleaning detection module based on infrared sensing according to claim 1, characterized in that, The pollution level is positively correlated with the descent height and negatively correlated with the rotation speed.

3. The control method for a photovoltaic cleaning detection module based on infrared sensing according to claim 1, characterized in that, The difference calculation is specifically as follows: The infrared thermal image data of the same location before and after cleaning are used to calculate the temperature difference pixel by pixel to generate a differential thermal map. When the temperature difference value of a certain area in the differential thermal map is greater than the preset cleaning threshold, the area is determined to be incompletely cleaned and marked as a stubborn stain area.

4. The control method for the photovoltaic cleaning detection module based on infrared sensing according to claim 3, characterized in that, The identification of abnormal hot spots specifically involves: The persistent hot spot area outside the stubborn stain area is extracted from the differential thermal map, and the shape feature, area feature and temperature gradient feature of the area are obtained. The features are matched with the diode failure, EVA delamination and cell crack features in the preset fault feature library. When the matching degree is greater than the preset threshold, it is determined that there is a photovoltaic panel fault in the corresponding area.

5. The control method for a photovoltaic cleaning detection module based on infrared sensing according to claim 1, characterized in that, The environmental parameter data includes ambient temperature, humidity, and irradiance. The infrared thermal image data is then calibrated based on this environmental parameter data using the following calibration formula: ； in, The actual temperature after calibration. To measure temperature, For ambient temperature, For ambient humidity, Irradiance, , , This is the environmental calibration coefficient.

6. The control method for a photovoltaic cleaning detection module based on infrared sensing according to claim 1, characterized in that, The multiple sets of independent cleaning execution units are arranged in two staggered rows, with no overlap between the projection positions of the front and rear cleaning execution units and covering the entire cleaning area.

7. The control method for a photovoltaic cleaning detection module based on infrared sensing according to claim 1, characterized in that, The cloud platform generates an operation and maintenance report based on the uploaded data. The operation and maintenance report includes pollution distribution statistics, cleaning effect assessment, fault location and type statistics, and predicts the performance degradation trend of photovoltaic panels based on historical data.

8. A control system for a photovoltaic cleaning detection module based on infrared sensing, characterized in that, include: The multi-source sensing module includes a front infrared sensing unit, a visible light acquisition unit, an environmental sensing unit, and a rear infrared sensing unit. It is used to collect infrared thermal image data, surface texture data, environmental parameter data, and infrared thermal image data of the photovoltaic panel to be cleaned, and transmit the collected raw data to the data processing module. The data processing module is used to receive raw data, perform radiation calibration on the infrared thermal image data based on the environmental parameter data, identify contaminated areas and calculate the contamination level by combining surface texture data, and generate a contamination distribution map of the photovoltaic panel to be cleaned; wherein, the contamination level is determined by calculating the weighted value of the infrared radiation temperature difference and the area ratio of the contaminated area; it is also used to perform differential calculation and abnormal hot spot matching on the infrared thermal image data before and after cleaning, and transmit the processing results to the control decision module. The control decision module is used to generate differentiated cleaning strategies based on the pollution distribution map output by the data processing module, send corresponding control commands to the cleaning execution module, and generate cleaning effect verification commands and fault diagnosis commands based on the differential results and hot spot matching results output by the data processing module. The cleaning execution module is used to receive control commands from the control decision module and convert them into drive signals to control multiple sets of independent external cleaning execution units to complete the corresponding cleaning actions. The data transmission module is used to upload the data generated by the data processing module and the control decision module to the cloud platform.

9. The control system for the photovoltaic cleaning detection module based on infrared sensing according to claim 8, characterized in that, Each of the independent cleaning execution units includes a lifting mechanism, a drive mechanism, and a cleaning brush. The lifting mechanism is used to drive the drive mechanism and the cleaning brush to adjust the height relative to the surface of the photovoltaic panel, and the drive mechanism is used to drive the cleaning brush to rotate.

Images