Photovoltaic Multi-Source Sensing Intelligent Cleaning and Cooling Method and System

Abstract

The present invention discloses a photovoltaic multi-source perception intelligent cleaning and cooling method and system, including: S1, acquiring weather data, photovoltaic power generation data, the temperature of photovoltaic equipment, the pollution information of photovoltaic panels, and the pollution information of the cleaning sponge of a photovoltaic cleaner; S2, preprocessing the acquired data; S3, analyzing the preprocessed weather data, photovoltaic power generation data, and the adhesion situation of surface pollutants on photovoltaic panels to generate a cleaning instruction for the photovoltaic cleaner; S4, analyzing the temperature of the photovoltaic equipment in S1 to generate a cooling instruction for the photovoltaic cleaner; S5, the photovoltaic cleaner executes the cleaning or cooling instruction; S6, according to the pollution degree of the cleaning sponge during the cleaning process and the water consumption of the cleaning or cooling operation, automatically completes the replacement and cleaning of the cleaning sponge of the photovoltaic cleaner and the water replenishment operation of the water tank. The present invention realizes the unmanned and intelligent deployment of a photovoltaic power station, improves the power generation efficiency, reduces the operation and maintenance cost of the photovoltaic power station, and improves the economic benefit of the photovoltaic power station.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic pollution detection, cleaning and cooling, and particularly relates to a photovoltaic multi-source perception intelligent cleaning and cooling method and system. Background Art

[0002] In recent years, as an important development direction in the field of clean energy, photovoltaic power generation has gradually become the core pillar of the global energy structure transformation. With the continuous progress of photovoltaic technology and the continuous expansion of photovoltaic installation scale, its proportion in the renewable energy market has increased steadily. However, the photovoltaic power generation efficiency is affected by various factors. Among them, the accumulation of pollutants on the surface of photovoltaic panels and the excessive temperature of photovoltaic equipment are two key problems, which seriously restrict the actual power generation performance and economic benefits of photovoltaic systems.

[0003] Since photovoltaic panels are exposed to the outdoors for a long time, their surfaces are extremely prone to accumulating dust, bird droppings, industrial pollutants (such as paint, chemical deposits, etc.). These pollutants not only block the incidence of sunlight, weaken the light absorption ability of photovoltaic cells, resulting in a significant decrease in the photoelectric conversion efficiency, but also may exacerbate heat accumulation, further reducing the power generation performance, and greatly affecting the economic benefits of photovoltaic power stations.

[0004] In addition, the accumulation of pollutants may also trigger the hot spot effect. When some photovoltaic panels cannot generate electricity normally due to pollutant blockage, this area may become a local "load", resulting in abnormal temperature rise inside the battery. Long-term temperature accumulation will not only accelerate the deterioration of photovoltaic materials, but also may cause thermal damage, shorten the service life, and increase the operation and maintenance costs. Therefore, regularly cleaning the photovoltaic panels and keeping their surfaces clean are key measures to ensure the long-term stable operation of photovoltaic systems and maintain high power generation efficiency.

[0005] In addition to pollutant accumulation, photovoltaic power generation also faces the problem of decreased power generation performance in high-temperature environments. Since the working efficiency of photovoltaic cells is closely related to temperature, when the ambient temperature is too high, the operating temperature of the photovoltaic panels may rise significantly, resulting in a decrease in its output voltage, thereby reducing the overall power generation efficiency. Therefore, in high-temperature environments, intelligent cooling technology has become an important means to improve the stability and power generation efficiency of photovoltaic systems.

[0006] Currently, the existing cleaning and equipment cooling solutions for photovoltaic pollutants still have problems such as high cost, low efficiency, and insufficient automation. For example, although the traditional manual method is simple to operate, it is costly in large-scale photovoltaic power stations and it is difficult to ensure the cleaning frequency and coverage. Although using mechanical equipment can improve the cleaning and cooling efficiency, it often lacks intelligent scheduling and cannot accurately judge the cleaning and cooling timing according to the pollutant accumulation situation and ambient temperature, affecting the overall economy of photovoltaic power stations.

[0007] In view of the above problems, this patent proposes a photovoltaic adaptive cleaning and cooling method and system based on multi-source information perception and intelligent decision-making. By analyzing weather data, power generation data, photovoltaic device temperature, and pollutant monitoring results based on computer vision in real time, the cleaning and cooling strategies for photovoltaic are formulated to meet the conditional requirements of high-quality photovoltaic power generation, thereby improving the power generation efficiency and economic benefits of photovoltaic power stations. Summary of the Invention

[0008] The first object of the present invention is to overcome the deficiencies and drawbacks in the prior art and provide a photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making. This method formulates photovoltaic cleaning and cooling strategies by analyzing weather data, power generation data, photovoltaic device temperature, and pollutant monitoring results based on computer vision in real time, effectively solving the problems of high cost, low efficiency, and insufficient automation and intelligence in existing photovoltaic cleaning and cooling solutions, and playing a key role in aspects such as intelligent energy management, intelligent cleaner scheduling, and adaptive operation and maintenance of photovoltaic systems in subsequent photovoltaic power stations.

[0009] The second object of the present invention is to provide a photovoltaic adaptive cleaning and cooling system based on multi-source information perception and intelligent decision-making.

[0010] The third object of the present invention is to provide a medium.

[0011] The fourth object of the present invention is to provide a computing device.

[0012] The first object of the present invention is achieved through the following technical solutions: A photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making, comprising the following steps:

[0013] S1. Obtain monitoring data, namely weather data, photovoltaic power generation data, photovoltaic device temperature, pollution information of the photovoltaic panel, and pollution information of the cleaning sponge of the photovoltaic cleaner;

[0014] S2. Perform data preprocessing, extract key data from the weather data and photovoltaic power generation data in S1, and optimize the pollution information of the photovoltaic panel and the pollution information of the cleaning sponge of the photovoltaic cleaner;

[0015] S3. Jointly analyze the key data of the weather data and photovoltaic power generation data in S2 and the attachment situation of pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel photo, and generate a cleaning execution instruction for the photovoltaic cleaner;

[0016] S4. Analyze the photovoltaic device temperature in S1 and generate a cooling execution instruction for the photovoltaic cleaner;

[0017] S5. According to the cleaning execution instruction of the photovoltaic cleaner in S3 and the cooling execution instruction of the photovoltaic cleaner in S4, the photovoltaic cleaner performs adaptive cleaning and cooling operations on the photovoltaic panel;

[0018] S6. According to the water consumption of the photovoltaic cleaner and the pollution degree of the cleaning sponge in S5, automatically complete the replacement and cleaning of the cleaning sponge and the water replenishment operation of the water tank of the photovoltaic cleaner.

[0019] Furthermore, in S1, the weather data is real-time weather data obtained by accessing the China Meteorological Data Network interface, which is a text data describing the weather conditions; the photovoltaic power generation data is the photovoltaic power generation recorded in the photovoltaic power station system; the photovoltaic device temperature is provided by the temperature sensor of the photovoltaic panel; the pollution information of the photovoltaic panel is the monitored photovoltaic panel video stream file; the pollution information of the cleaning sponge of the photovoltaic cleaner is the photo of the cleaning sponge taken by the camera under the photovoltaic panel;

[0020] Among them, every half an hour, the weather data update, the acquisition of photovoltaic power generation data, and the operation of intercepting photos from the photovoltaic panel video stream file are automatically synchronized. The cleaning sponge photos are taken once every half minute when the photovoltaic cleaner executes the cleaning instruction.

[0021] Furthermore, in S2, the specific situations of extracting the key data of the weather data and photovoltaic power generation data in S1 and optimizing the pollution information of the photovoltaic panel and the pollution information of the cleaning sponge of the photovoltaic cleaner are as follows:

[0022] A. Extract the air quality index, temperature (°C), wind speed (m / s), precipitation (mm), air humidity (%), and solar irradiance (W / m 2 ) in the weather data;

[0023] B. Extract the actual power generation of each photovoltaic panel from the photovoltaic power generation data, and obtain the theoretical power generation of each photovoltaic panel under ideal conditions from the equipment manual;

[0024] C. The specific steps for optimizing the intercepted photovoltaic panel photos and the cleaning sponge photos of the photovoltaic cleaner are as follows:

[0025] C.1. The specific steps for denoising, sharpening, and super-resolution optimization of the multiple monitored photovoltaic panel photos and the cleaning sponge photos of the photovoltaic cleaner by introducing a new photovoltaic pollutant image enhancement method based on adaptive frequency domain fusion and super-resolution reconstruction are as follows:

[0026] a. Decompose the image by wavelet transform WT, and decompose the image into a low-frequency part L and a high-frequency part H:

[0027] ;

[0028] Among them, the low-frequency L contains illumination and shadow information, and the high-frequency H contains pollutant edge and high-frequency noise information;

[0029] b. Calculate the noise standard deviation using the BayesShrink method :

[0030] ;

[0031] Among them, is the median function, and then obtain the adaptive threshold , and filter the image noise:

[0032] ;

[0033] Among them, is the signal standard deviation, is the sign function, is the high-frequency component of the image after filtering the noise;

[0034] c. After combining the high-frequency component after noise filtering and the low-frequency part L, perform inverse wavelet transform reconstruction to obtain the image after filtering the noise:

[0035]

[0036] C.2. The specific steps for image sharpening based on structure tensor gradient enhancement are as follows:

[0037] a. Calculate the Sobel direction gradient:

[0038] ;

[0039] Among them, represents the gradient along the horizontal direction of direction, represents the gradient along the horizontal direction of direction, and then calculate the structure tensor to identify the pollutant boundary:

[0040] ;

[0041] Among them, and represent the main direction change rate of the structure.

[0042] b. Calculate the sharpening weight of each pixel:

[0043] ;

[0044] c. Perform adaptive sharpening on the image:

[0045] ;

[0046] Among them, is the second-order Laplacian derivative for enhancing edges, is the adaptive sharpening coefficient;

[0047] C.3. The specific steps to optimize the image sharpness using adversarial learning-based super-resolution are as follows:

[0048] a. Use ESRGAN for super-resolution reconstruction to obtain the super-resolution reconstructed image:

[0049] ;

[0050] Among them, is the ESRGAN generator, which calculates the feature difference between the SR image and the original high-definition image:

[0051] ;

[0052] Among them, is the high-level feature of the VGG network extraction layer, is the actual real image, and the adversarial loss is used to naturalize the super-resolution image:

[0053] ;

[0054] Among them; is the discriminator network;

[0055] b. Train the final goal to optimize the super-resolution image:

[0056] ;

[0057] Among them; is the L1 loss, and are the hyperparameter weights for controlling different losses.

[0058] Furthermore, in S3, the specific situation of jointly analyzing the key data of the weather data and the photovoltaic power generation data in S2 and the attachment of pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel photo to generate the cleaning execution instruction of the photovoltaic cleaner is as follows:

[0059] A. According to the influence of weather factors at different levels on the degree of pollutant adhesion and the interaction between various weather factors, corresponding pollutant adhesion weights are assigned to different weather factors. The pollutant adhesion weight can indicate whether the weather factor is conducive to pollutant settlement. A positive weight indicates that the weather factor is conducive to pollutant settlement, a weight of zero indicates that the weather factor has no obvious influence on pollutant settlement, and a negative weight indicates that the weather factor is not conducive to pollutant settlement. By comprehensively analyzing the comprehensive influence of various weather factors on pollutant adhesion, the pollutant adhesion weights of the following six key weather factors are determined:

[0060] First, the pollutant adhesion weight corresponding to the air quality index: If the air quality index within the selected period is less than 50, it is considered that the pollutant accumulation is less, and the pollutant adhesion weight corresponding to the air quality index is assigned -0.2; if the air quality index within the selected period is greater than 50 and less than 150, it is considered that the pollutant accumulation rate is relatively fast, and the pollutant adhesion weight corresponding to the air quality index is assigned 0.1; if the air quality index within the selected period is greater than 150, it is considered that the pollutant accumulation rate is extremely fast, and the pollutant adhesion weight corresponding to the air quality index is assigned 0.3;

[0061] Second, the pollutant adhesion weight corresponding to the precipitation: If the precipitation within the selected time period is less than 2 mm, it cannot only clean the dust on the photovoltaic surface but may even adsorb the dust in the air, so the pollutant adhesion weight corresponding to the precipitation is assigned 0.2; if the precipitation within the selected period is greater than 2 mm and less than 10 mm, it may only clean some of the pollutants on the photovoltaic surface, so the pollutant adhesion weight corresponding to the precipitation is assigned -0.1; if the precipitation within the selected period is greater than 10 mm, it may clean most of the pollutants on the photovoltaic surface, so the pollutant adhesion weight corresponding to the precipitation is assigned -0.3;

[0062] Third, the pollutant adhesion weight corresponding to the air humidity: If the air humidity within the selected time period is less than 30%, it indicates that the air is dry and it is easy to form electrostatic adsorption pollution, so the pollutant adhesion weight corresponding to the air humidity is assigned 0.2; if the air humidity within the selected time period is greater than 30% and less than 80%, it has no obvious influence on pollutant adhesion, so the pollutant adhesion weight corresponding to the air humidity is assigned 0.0; if the air humidity within the selected time period is greater than 80%, it may cause muddy dust pollution on the photovoltaic surface, so the pollutant adhesion weight corresponding to the air humidity is assigned 0.15;

[0063] Fourth, the pollutant adhesion weight corresponding to the solar irradiance: If the solar irradiance within the selected time period is less than 200 W / m 2 ², the temperature on the photovoltaic surface is low and the pollutants are not easy to solidify, so the pollutant adhesion weight corresponding to the solar irradiance is assigned -0.1; if the solar irradiance within the selected time period is greater than 200 W / m2 and less than 600 W / m 2 , then there is no obvious impact on the attachment of pollutants. Therefore, the pollutant attachment weight corresponding to the solar irradiance is 0.0; when the solar irradiance within the selected time period is greater than 600 W / m 2 , it may cause high temperature on the photovoltaic surface, resulting in the solidification of pollutants and difficult cleaning. Therefore, the pollutant attachment weight corresponding to the solar irradiance is 0.2;

[0064] Fifth, the pollutant attachment weight corresponding to wind speed: when the wind speed within the selected time period is less than 1 m / s, it indicates that the air is relatively static and the pollutant sedimentation is serious. Therefore, the pollutant attachment weight corresponding to the wind speed is 0.3; when the wind speed within the selected time period is greater than 1 m / s and less than 5 m / s, the wind speed is moderate, which helps to reduce pollutant sedimentation. Therefore, the pollutant attachment weight corresponding to the wind speed is -0.1; when the wind speed within the selected time period is greater than 5 m / s, it may bring new dust. Therefore, the pollutant attachment weight corresponding to the wind speed is 0.1;

[0065] Sixth, the pollutant attachment weight corresponding to temperature: when the temperature within the selected time period is less than 20 °C, the environment is in a low temperature, and the pollutants on the photovoltaic surface are not easily solidified. Therefore, the pollutant attachment weight corresponding to the temperature is -0.2; when the temperature within the selected time period is greater than 20 °C and less than 30 °C, the environmental temperature is moderate and the pollutants are moderately solidified. Then the pollutant attachment weight corresponding to the temperature is -0.1; when the temperature within the selected time period is greater than 30 °C, the environment is in a high temperature, and the pollutants on the photovoltaic surface are significantly solidified and difficult to clean. Therefore, the pollutant attachment weight corresponding to the temperature is 0.3;

[0066] Calculate the necessity index of photovoltaic cleaning according to the above different weather factors , and the calculation formula is:

[0067] ;

[0068] where, is the air quality index, is the precipitation, is the air humidity, is the solar irradiance, is the wind speed, is the temperature, is the pollutant attachment weight corresponding to the air quality index, is the pollutant attachment weight corresponding to the precipitation, is the pollutant attachment weight corresponding to the air humidity, is the pollutant attachment weight corresponding to the solar irradiance, is the pollutant attachment weight corresponding to the wind speed and is the pollutant adhesion weight corresponding to the temperature;

[0069] B. According to the influence of weather factors at different levels on the photovoltaic power generation, and the interaction between various weather factors, corresponding weights of the influence on the photovoltaic power generation are assigned to different weather factors. The weight of the influence on the photovoltaic power generation can indicate whether the weather factor is favorable for the photovoltaic power generation. The closer the influence weight of the weather factor is to 1, the more favorable the weather factor condition is for the photovoltaic power generation. By comprehensively analyzing the comprehensive influence of various weather factors on the photovoltaic power generation, the weights of the influence on the photovoltaic power generation of the following seven key weather factors are determined:

[0070] First, the weight of the influence on the photovoltaic power generation of solar irradiance: Calculate the ratio of the actual solar irradiance to the solar irradiance in the theoretical test scenario as the weight of the influence on the power generation of solar irradiance:

[0071] ;

[0072] Among them, is the actual solar irradiance, is the solar irradiance in the theoretical test scenario;

[0073] Second, the weight of the influence on the photovoltaic power generation of cloud cover: Use the exponential decay model to calculate the weight of the influence on the photovoltaic power generation of cloud cover:

[0074] ;

[0075] Among them, is the cloud cover, is the empirical coefficient that can be adjusted according to the measured data, is the weight of the influence on the photovoltaic power generation of cloud cover;

[0076] Third, the weight of the influence on the photovoltaic power generation of temperature: The temperature of the photovoltaic module and the ambient temperature around the photovoltaic comprehensively affect the efficiency of the photovoltaic power generation:

[0077] ;

[0078] Among them, is the temperature of the photovoltaic device measured by the temperature sensor on the photovoltaic, is the ambient temperature, is the temperature coefficient, is the standard temperature of 25 °C, is the weight of the influence on the photovoltaic power generation of temperature;

[0079] Fourth, the weight of the influence on the photovoltaic power generation of wind speed: The wind speed affects the heat dissipation of the photovoltaic and thus affects the power generation efficiency:

[0080] ;

[0081] Among them, is the actual wind speed, is the standard reference wind speed; is the wind speed influence coefficient, which is the influence weight of wind speed on photovoltaic power generation;

[0082] Fifth, the influence weight of air humidity on photovoltaic power generation: Different air humidities have different effects on the attachment of pollutants on the photovoltaic surface, further affecting the photovoltaic power generation efficiency. When the air humidity is greater than 80%, the influence weight of air humidity on photovoltaic power generation is assigned as 0.95. When the air humidity is greater than 30% and less than 80%, the influence weight of air humidity on photovoltaic power generation is assigned as 1. When the air humidity is less than 30%, the influence weight of air humidity on photovoltaic power generation is assigned as 0.98;

[0083] Sixth, the influence weight of precipitation on photovoltaic power generation: Rainwater can refract sunlight in the air, affecting the direct reception of light sources by the photovoltaic. There is a relationship between the precipitation of rainwater and the refraction of sunlight. When the precipitation is greater than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.2. When the precipitation is greater than 2 mm and less than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.05. When the precipitation is less than 2 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 0.9;

[0084] Seventh, the influence weight of air quality on photovoltaic power generation: Impurities such as dust in the air will reduce the intensity of sunlight radiation. When the air quality index is greater than 150, the influence weight of air quality on photovoltaic power generation is assigned as 0.8. When the air quality index is greater than 50 and less than 150, the influence weight of air quality on photovoltaic power generation is assigned as 0.9. When the air quality index is less than 50, the influence weight of air quality on photovoltaic power generation is assigned as 1;

[0085] Calculate the theoretical power generation of the photovoltaic according to the above different weather factors. The calculation formula is:

[0086] ;

[0087] Among them, is the power generation of the photovoltaic under ideal environmental conditions, is the theoretical power generation of the photovoltaic in the actual environmental conditions. Then, compare the actual power generation of the photovoltaic with the actual theoretical power generation. If the actual power generation of the photovoltaic can reach 90% of the actual theoretical power generation, it indicates that there is no obvious abnormality in the actual power generation of the photovoltaic. Let the photovoltaic power generation abnormality index , otherwise, it is considered that the power generation of the photovoltaic is abnormal. Let the photovoltaic power generation abnormality index ;

[0088] C. Introduce a new pollutant recognition method based on adaptive feature fusion to identify the specific situation of small-range pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photos in S2 as follows:

[0089] C.1. Use YOLOv8 for pollutant target detection and output the pollutant bounding box:

[0090] ;

[0091] Among them, represents the leftmost position of the pollutant area, represents the bottommost position of the pollutant area, represents the rightmost position of the pollutant area, represents the topmost position of the pollutant area. Crop the target area based on the bounding box:

[0092] ;

[0093] C.2. Use Swin Transformer to extract pollutant area features through the sliding window attention mechanism:

[0094] ;

[0095] Calculate the spatial consistency of the pollutant area and merge pollutants with adjacent boundaries:

[0096] ;

[0097] C.3. Combine CNN and Swin Transformer to extract features:

[0098] ;

[0099] Calculate the final fused features:

[0100] ;

[0101] Among them, and are adaptive weights;

[0102] C.4. Introduce a multi-frame information fusion method to calculate the consistency of pollutants in multiple photos:

[0103] ;

[0104] Among them, when is greater than the retention threshold, retain ;

[0105] C.5. Use U-Net to segment the retained pollutant areas to generate a pollutant mask:

[0106] ;

[0107] D. The specific steps to identify large-scale pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photo in S2 using wide-area pollutant identification based on spatial color analysis are as follows:

[0108] D.1. Calculate the color histogram difference between the clean photovoltaic panel and the panel in the actual image:

[0109] ;

[0110] Among them, is the color space, respectively represent the red, green, and blue channels in the RGB space, represents the color channel in the th color interval frequency, represents the color histogram of the clean photovoltaic panel, represents the color histogram of the actual image;

[0111] b. Calculate the mean value of the color channels of the clean photovoltaic panel

[0112] ;

[0113] Among them, is the color value of the pixel point in the clean photovoltaic panel, calculate the mean value of the color channels of the panel in the actual image:

[0114] ;

[0115] Among them, is the color mean value of the pixel point in the photovoltaic panel in the actual image, calculate the difference in the mean values of the color channels between the clean photovoltaic panel and the panel in the actual image:

[0116] ;

[0117] c. Generate a mask for the pollutant region based on the comparison result of the difference in the color histogram in a and the color histogram difference threshold and the comparison result of the difference in the mean value of the color channels in b and the difference threshold of the mean value of the color channels :

[0118] ;

[0119] E. Calculate the small - range pollutant area and the large - range pollutant area by the number of pixels:

[0120] ;

[0121] where the area of each pixel is 1 cm 2 , represents the number of pixels of the small - range pollutant, i.e., the pollutant area, represents the number of pixels of the large - range pollutant, i.e., the pollutant area, and pollution region represents the pollutant region;

[0122] F. Comprehensively analyze the necessity index of photovoltaic cleaning in A , the photovoltaic power generation anomaly index in B and the pollutant area in C and to determine the specific steps for whether the photovoltaic needs to be cleaned as follows:

[0123] F.1. According to the influence of the photovoltaic cleaning necessity index, the photovoltaic power generation anomaly index, and the pollutant area on the photovoltaic cleaning demand, assign different importance weights to them: ;

[0124] F.2. Calculate the photovoltaic comprehensive cleaning index for pollutants in different ranges according to the importance weights in F.1:

[0125] ;

[0126] When or , the photovoltaic cleaner obtains the execution instruction to clean the pollutants, where is the comprehensive cleaning threshold interval for small - range pollutants, is the comprehensive cleaning threshold interval for large - range pollutants.

[0127] Furthermore, in S4, analyze the temperature of the photovoltaic device in S1 to generate a cooling execution instruction for the photovoltaic cleaner. When the temperature of the photovoltaic device is greater than the high - temperature resistance threshold for normal operation of the photovoltaic, the photovoltaic cleaner obtains the cooling execution instruction;

[0128] Further, in S5, according to the cleaning execution instruction of the photovoltaic cleaner in S3 and the temperature reduction execution instruction of the photovoltaic cleaner in S4, the specific situation of the photovoltaic cleaner performing adaptive cleaning and temperature reduction operations on the photovoltaic panel is as follows:

[0129] A. The specific steps for the photovoltaic cleaner to execute the cleaning pollutant instruction are as follows:

[0130] A.1. According to the information of the pollutants in S3, locate the number of the photovoltaic panel where the pollutants are located;

[0131] A.2. According to the number of the photovoltaic panel in A.1, the specific steps for the photovoltaic cleaner corresponding to the number to execute the cleaning pollutant instruction are as follows:

[0132] a. Use YOLOv8 to perform real-time target recognition of the photovoltaic cleaner on the monitored video stream. Taking the coordinate system of the monitoring screen as a reference, the photovoltaic cleaner starts running to the right from the starting point while the photovoltaic cleaner is lifted, and the cleaning sponge in the photovoltaic cleaner does not contact the photovoltaic panel, and the water tank of the photovoltaic cleaner does not release water;

[0133] b. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutant, the photovoltaic cleaner is lowered and the right water pipe of the water tank of the photovoltaic cleaner releases water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutants. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutant, the photovoltaic cleaner is lifted and the right water pipe of the water tank of the photovoltaic cleaner stops releasing water. The first cleaning of the target pollutant ends, record the water release time of this process, and repeat this operation for the pollutants that have not been cleaned for the first time;

[0134] c. After the first cleaning of all pollutants ends, the photovoltaic cleaner starts to return. During this process, the pollutants cleaned in step b are cleaned again. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutant, the photovoltaic cleaner is lowered and the left water pipe of the water tank of the photovoltaic cleaner releases water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutants. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutant, the photovoltaic cleaner is lifted and the left water pipe of the water tank of the photovoltaic cleaner stops releasing water. The second cleaning of the target pollutant ends, record the water release time of this process, and repeat this operation for the pollutants that have not been cleaned for the second time;

[0135] d. The photovoltaic cleaner returns to the starting point, and statistically calculates the total water release time of the water tank for the two round trips of the photovoltaic cleaner;

[0136] B. The specific steps for the photovoltaic cleaner to execute the temperature reduction instruction are as follows:

[0137] a. The photovoltaic cleaner runs back and forth twice from the starting point to the rightmost end of the photovoltaic panel. During the process, the photovoltaic cleaner is lifted, and the water pipes on both sides of the water tank release water simultaneously.

[0138] b. The photovoltaic cleaner returns to the starting point, the water pipes on both sides of the water tank stop releasing water, and the water release time for the photovoltaic cleaner to run back and forth twice is statistically calculated.

[0139] Furthermore, in S6, according to the pollution degree of the cleaning sponge during the cleaning process in S5 and the water consumption of the photovoltaic cleaner when performing cleaning or cooling tasks, the specific situations of automatically replacing the cleaning sponge of the photovoltaic cleaner, cleaning, and replenishing water in the water tank are as follows:

[0140] a. There are the following two operation methods for automatically cleaning the cleaning sponge of the photovoltaic cleaner:

[0141] The first is to replace the cleaning sponge: Use the wide-area pollutant identification method based on spatial color analysis in S3 to identify the photo of the cleaning sponge of the optimized photovoltaic cleaner in S2, judge the pollution situation of the cleaning sponge. During the operation of the photovoltaic cleaner, when the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the first time, the photovoltaic cleaner controls the cleaning sponge to turn over and change the surface. When the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the second time, the photovoltaic cleaner controls to replace a new cleaning sponge;

[0142] The second is to clean the cleaning sponge: When the photovoltaic cleaner returns to the starting point after completing the cleaning task, the flushing nozzle at the starting point is automatically turned on to flush the used cleaning sponge in the photovoltaic cleaner, and at the same time, the photovoltaic cleaner controls the cleaning sponge to rotate.

[0143] b. After the photovoltaic cleaner completes the cleaning or cooling task and returns to the starting point, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishing device are turned on. The water filling port of the photovoltaic water replenishing device and the water replenishing port of the water tank of the photovoltaic cleaner are magnetically docked, and at the same time, the water replenishing pipe valve is opened to replenish water to the water tank. The water replenishing volume is calculated according to the total water release time of 300Ml / s. After the water replenishing is completed, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishing device are turned off and the water replenishing pipe valve is closed at the same time;

[0144] The second object of the present invention is achieved through the following technical solutions: A photovoltaic adaptive cleaning and cooling system based on multi-source information perception and intelligent decision-making is used to implement the above-mentioned photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making, and it includes:

[0145] A data acquisition module for acquiring weather data, photovoltaic power generation data, photovoltaic device temperature, photovoltaic panel monitoring video, and photos of the cleaning sponge of the photovoltaic cleaner.

[0146] A data preprocessing module, which is used to extract key weather data, power generation data of each photovoltaic panel, the temperature of photovoltaic equipment, and optimize the intercepted photos of photovoltaic panels and the cleaning sponge photos of photovoltaic cleaners;

[0147] A photovoltaic pollution condition analysis module, which is used to analyze the cleaning requirements of photovoltaic panels through weather conditions, power generation conditions of photovoltaic power generation, and the actual attachment of pollutants on photovoltaic panels, and give cleaning execution instructions to photovoltaic cleaners;

[0148] A photovoltaic equipment temperature analysis module, which is used to analyze the cooling requirements of photovoltaic equipment according to the temperature of photovoltaic equipment, and give cooling execution instructions to photovoltaic cleaners;

[0149] A photovoltaic cleaner scheduling module, which is used to execute photovoltaic cleaning and cooling instructions according to the cleaning and cooling requirements of photovoltaic panels;

[0150] A photovoltaic cleaner water replenishment module, which is used to automatically replenish water according to the water consumption of photovoltaic cleaners during cleaning and cooling operations;

[0151] A photovoltaic cleaner cleaning module, which is used to replace or clean the cleaning sponge in the photovoltaic cleaner according to the pollution condition of the cleaning sponge;

[0152] The third object of the present invention is achieved through the following technical solution. A storage medium stores a program, and when the program is executed by a processor, the above-mentioned photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making is realized.

[0153] The fourth object of the present invention is achieved through the following technical solution: A computing device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making is realized.

[0154] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0155] 1. For the first time, the present invention calculates the pollutant attachment weight and the cleaning necessity index based on weather data, photovoltaic power generation data, and image recognition technology, makes an intelligent decision on when to clean, avoids unnecessary cleaning, optimizes the cleaning schedule, and improves the cleaning efficiency.

[0156] 2. For the first time, the present invention adopts an image enhancement method based on adaptive frequency domain fusion and super-resolution reconstruction, uses wavelet transform for noise reduction, structure tensor gradient for enhancement and sharpening, and combines ESRGAN to achieve super-resolution optimization, making pollutant recognition more accurate.

[0157] 3. The present invention constructs a photovoltaic power generation impact weight model for the first time. By combining factors such as solar irradiance, cloud cover, temperature, wind speed, and humidity, it calculates the theoretical power generation and compares it with the actual power generation to accurately determine whether the power generation efficiency decreases due to pollutants.

[0158] 4. The present invention combines the efficient object detection ability of YOLOv8 and the global feature extraction ability of Swin Transformer for the first time to improve the ability to detect pollutants and enhance the accuracy of pollutant identification.

[0159] 5. The present invention intelligently analyzes data for the first time, automatically provides execution instructions for the photovoltaic cleaning robot, realizes autonomous scheduling and adaptive cleaning, reduces manual intervention, and improves the degree of automation.

[0160] 6. The present invention intelligently sets the trigger conditions for photovoltaic cleaning based on the impact of pollutant areas in different ranges on the power generation performance of photovoltaic devices for the first time, thereby effectively reducing the occurrence probability of the hot spot effect, reducing unnecessary cleaning times, and lowering the maintenance cost.

[0161] 7. The present invention has a wide range of application spaces in the identification of photovoltaic pollution and the improvement of cleaning technology, and has broad prospects for improving the power generation efficiency and economic benefits of photovoltaic power stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0162] Figure 1 It is a schematic logical flow diagram of the method of the present invention.

[0163] Figure 2 It is an architecture diagram of the system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0164] The present invention will be further described below in conjunction with specific embodiments.

[0165] Such as Figure 1As shown in the figure, this embodiment discloses a photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making. First, weather data, photovoltaic power generation data, pollution information of the photovoltaic panel, sensor temperature for analyzing the temperature of the photovoltaic equipment, and pollution information of the cleaning sponge of the photovoltaic cleaner for analyzing the pollution situation of the cleaning sponge are obtained. Then, the acquired data is preprocessed to extract key information and optimize the intercepted photovoltaic panel photos and the taken cleaning sponge photos. After that, based on the preprocessed weather data, photovoltaic power generation data, photovoltaic panel photos, and photovoltaic equipment temperature, an analysis of the cleaning or cooling requirements of the photovoltaic panel is carried out to generate an execution instruction for the photovoltaic cleaner. Then, the photovoltaic cleaner performs panel cleaning or cooling operations according to the execution instruction. During this period, the photovoltaic cleaner performs intelligent replacement according to the pollution situation of the cleaning sponge. Finally, the used cleaning sponge during the cleaning process is washed and replenished according to the water consumption of the cleaning or cooling operation. It includes the following steps:

[0166] S1. Obtain monitoring data, namely weather data, photovoltaic power generation data, photovoltaic equipment temperature, pollution information of the photovoltaic panel, and pollution information of the cleaning sponge of the photovoltaic cleaner; the weather data is real-time weather data obtained by accessing the China Meteorological Data Network interface, which is a text data describing the weather conditions; the photovoltaic power generation data is the photovoltaic power generation recorded in the photovoltaic power station system; the photovoltaic equipment temperature is provided by the temperature sensor of the photovoltaic panel; the pollution information of the photovoltaic panel is the monitored photovoltaic panel video stream file; the pollution information of the cleaning sponge of the photovoltaic cleaner is the cleaning sponge photo taken by the camera under the photovoltaic panel;

[0167] Among them, every half hour, the weather data update, photovoltaic power generation data acquisition, and the operation of intercepting photos from the photovoltaic panel video stream file are automatically synchronized. The cleaning sponge photos are taken every half minute when the photovoltaic cleaner executes the cleaning instruction.

[0168] By adopting the above steps, the original weather data 、the original photovoltaic power generation data 、the photovoltaic equipment temperature 、the photovoltaic panel video ,the intercepted photovoltaic panel photos and the original cleaning sponge photos are obtained. Among them, represents the th photovoltaic panel photo intercepted from the photovoltaic panel video.

[0169] S2. The specific situations of extracting the key data of the weather data and photovoltaic power generation data in S1 and optimizing the pollution information of the photovoltaic panel and the pollution information of the cleaning sponge of the photovoltaic cleaner are as follows:

[0170] A. Extract the air quality index, temperature (°C), wind speed (m / s), precipitation (mm), air humidity (%), and solar irradiance (W / m 2 ) from the weather data;

[0171] B. Extract the actual power generation of each photovoltaic panel from the photovoltaic power generation data, and obtain the theoretical power generation of each photovoltaic panel under ideal conditions from the equipment manual;

[0172] C. The specific steps for optimizing the intercepted photovoltaic panel photos and the cleaning sponge photos of the photovoltaic cleaner are as follows:

[0173] C.1. The specific steps for denoising, sharpening, and super-resolution optimization of multiple monitored photovoltaic panel photos and cleaning sponge photos of the photovoltaic cleaner by introducing a new photovoltaic pollutant image enhancement method based on adaptive frequency-domain fusion and super-resolution reconstruction are as follows:

[0174] a. Decompose the image by wavelet transform WT, and decompose the image into a low-frequency part L and a high-frequency part H:

[0175] [[ID=ID=18]] ;

[0176] Among them, the low-frequency L contains illumination and shadow information, and the high-frequency H contains pollutant edges and high-frequency noise information;

[0177] b. Use the BayesShrink method to calculate the noise standard deviation :

[0178] ;

[0179] Among them, is the median function, and then obtain the adaptive threshold , and filter the image noise:

[0180] ;

[0181] Among them, is the signal standard deviation, is the sign function, is the high-frequency component of the image after filtering the noise;

[0182] c. After combining the high-frequency component filtered by noise and the low-frequency part L, perform wavelet inverse transform reconstruction to obtain the image after filtering the noise:

[0183]

[0184] C.2. The specific steps for enhancing image sharpness based on structural tensor gradient are as follows:

[0185] a. Calculate the Sobel directional gradient:

[0186] ;

[0187] Among them, represents the gradient horizontal along the direction, represents the gradient horizontal along the direction, and then calculate the structure tensor Identify the pollutant boundary:

[0188] ;

[0189] Among them, and represent the main direction change rates of the structure.

[0190] b. Calculate the sharpening weight of each pixel:

[0191] ;

[0192] c. Perform adaptive sharpening on the image:

[0193] ;

[0194] Among them, is the second-order derivative of the Laplacian used to enhance the edge, is the adaptive sharpening coefficient;

[0195] C.3. The specific steps to optimize the image sharpness using adversarial learning-based super-resolution are as follows:

[0196] a. Use ESRGAN for super-resolution reconstruction to obtain the super-resolution reconstructed image:

[0197] ;

[0198] Among them, is the ESRGAN generator, calculate the feature difference between the SR image and the original high-definition image:

[0199] ;

[0200] Among them, is the high-level feature of the VGG network extraction layer, is the actual real image, and use the adversarial loss to naturalize the super-resolution image:

[0201] ;

[0202] Among them; is the discriminator network;

[0203] b. The final training goal is to optimize the super-resolution image:

[0204] ;

[0205] where; is the L1 loss, and are the hyperparameter weights for controlling different losses.

[0206] By adopting the above steps, the air quality index , precipitation , air humidity , solar irradiance , wind speed , and temperature are extracted from the original weather data, and the power generation of each photovoltaic panel is extracted from the original photovoltaic power generation data. , where is the power generation of the th photovoltaic panel. The intercepted photovoltaic panel photos and the original clean sponge photos are optimized to obtain high-definition photovoltaic panel photos and clean sponge photos , where represents the photo after optimizing the th intercepted photovoltaic panel photo.

[0207] S3. The specific situation of jointly analyzing the key data of the weather data and the photovoltaic power generation data in S2 and the attachment of pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel photos to generate the cleaning execution instructions for the photovoltaic cleaner is as follows:

[0208] A. According to the influence of weather factors at different levels on the degree of pollutant attachment and the interaction between various weather factors, corresponding pollutant attachment weights are assigned to different weather factors. The pollutant attachment weight can indicate whether the weather factor is conducive to pollutant sedimentation. A positive weight indicates that the weather factor is conducive to pollutant sedimentation, a weight of zero indicates that the weather factor has no obvious influence on pollutant sedimentation, and a negative weight indicates that the weather factor is not conducive to pollutant sedimentation. By comprehensively analyzing the comprehensive influence of various weather factors on pollutant attachment, the pollutant attachment weights of the following six key weather factors are determined:

[0209] First, the pollutant attachment weight corresponding to the air quality index: If the air quality index within the selected period is less than 50, it is considered that the pollutant accumulation is less, and the pollutant attachment weight corresponding to the air quality index is given as -0.2; if the air quality index within the selected period is greater than 50 and less than 150, it is considered that the pollutant accumulation rate is relatively fast, and the pollutant attachment weight corresponding to the air quality index is given as 0.1; if the air quality index within the selected period is greater than 150, it is considered that the pollutant accumulation rate is extremely fast, and the pollutant attachment weight corresponding to the air quality index is given as 0.3;

[0210] Second, the pollutant attachment weight corresponding to the precipitation: If the precipitation within the selected time period is less than 2 mm, it cannot only clean the dust on the photovoltaic surface but may even adsorb the dust in the air. Therefore, the pollutant attachment weight corresponding to the precipitation is given as 0.2; if the precipitation within the selected period is greater than 2 mm and less than 10 mm, it may only clean some of the pollutants on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the precipitation is given as -0.1; if the precipitation within the selected period is greater than 10 mm, it may clean most of the pollutants on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the precipitation is given as -0.3;

[0211] Third, the pollutant attachment weight corresponding to the air humidity: If the air humidity within the selected time period is less than 30%, it indicates that the air is dry and it is easy to form electrostatic adsorption pollution. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.2; if the air humidity within the selected time period is greater than 30% and less than 80%, it has no obvious effect on the attachment of pollutants. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.0; if the air humidity within the selected time period is greater than 80%, it may cause mud-like dust pollution on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.15;

[0212] Fourth, the pollutant attachment weight corresponding to the solar irradiance: If the solar irradiance within the selected time period is less than 200 W / m 2 and the temperature of the photovoltaic surface is low, the pollutants are not easily solidified. Therefore, the pollutant attachment weight corresponding to the solar irradiance is given as -0.1; if the solar irradiance within the selected time period is greater than 200 W / m 2 and less than 600 W / m 2 it has no obvious effect on the attachment of pollutants. Therefore, the pollutant attachment weight corresponding to the solar irradiance is given as 0.0; if the solar irradiance within the selected time period is greater than 600 W / m 2 it may cause the temperature of the photovoltaic surface to be high, resulting in the solidification of pollutants and difficult cleaning. Therefore, the pollutant attachment weight corresponding to the solar irradiance is given as 0.2;

[0213] The fifth type is the pollutant attachment weight corresponding to wind speed: if the wind speed in the selected time period is less than 1 m / s, it indicates that the air is relatively still and pollutant deposition is serious, so the pollutant attachment weight corresponding to wind speed is 0.3; if the wind speed in the selected time period is greater than 1 m / s and less than 5 m / s, the wind speed is moderate, which helps to reduce pollutant deposition, so the pollutant attachment weight corresponding to wind speed is -0.1; if the wind speed in the selected time period is greater than 5 m / s, it may bring new dust, so the pollutant attachment weight corresponding to wind speed is 0.1;

[0214] The sixth type is the pollutant attachment weight corresponding to temperature: if the temperature during the selected time period is less than 20°C, the environment is at a low temperature, and the pollutants on the photovoltaic surface are not easy to solidify, the pollutant attachment weight corresponding to temperature is -0.2; if the temperature during the selected time period is greater than 20°C and less than 30°C, the ambient temperature is moderate, and the pollutants are moderately solidified, the pollutant attachment weight corresponding to temperature is -0.1; if the temperature during the selected time period is greater than 30°C, the environment is at a high temperature, and the pollutants on the photovoltaic surface are obviously solidified, making cleaning difficult, the pollutant attachment weight corresponding to temperature is 0.3;

[0215] Calculate the necessity index of photovoltaic cleaning based on the above different weather factors , the calculation formula is:

[0216] ;

[0217] in, The air quality index, is the precipitation, Is the air humidity, is the solar irradiance, is the wind speed, It is temperature, is the pollutant attachment weight corresponding to the air quality index, is the pollutant attachment weight corresponding to precipitation, is the pollutant attachment weight corresponding to air humidity, is the pollutant attachment weight corresponding to solar irradiance, is the pollutant attachment weight corresponding to wind speed and is the pollutant attachment weight corresponding to temperature;

[0218] B. According to the influence of weather factors at different levels on photovoltaic power generation and the interaction between various weather factors, corresponding weights of influence on photovoltaic power generation are assigned to different weather factors. The weight of influence on photovoltaic power generation can indicate whether the weather factor is conducive to photovoltaic power generation. The closer the influence weight of the weather factor is to 1, the more conducive the weather factor condition is to photovoltaic power generation. By comprehensively analyzing the comprehensive influence of various weather factors on photovoltaic power generation, the weights of influence on photovoltaic power generation of the following seven key weather factors are determined:

[0219] First, the weight of influence on photovoltaic power generation of solar irradiance: Calculate the ratio of the actual solar irradiance to the solar irradiance in the theoretical test scenario as the weight of influence on the power generation of solar irradiance:

[0220] ;

[0221] Among them, is the actual solar irradiance, is the solar irradiance in the theoretical test scenario;

[0222] Second, the weight of influence on photovoltaic power generation of cloud cover: Use the exponential decay model to calculate the weight of influence on photovoltaic power generation of cloud cover:

[0223] ;

[0224] Among them, is the cloud cover, is the empirical coefficient that can be adjusted according to measured data, is the weight of influence on photovoltaic power generation of cloud cover;

[0225] Third, the weight of influence on photovoltaic power generation of temperature: The temperature of the photovoltaic module and the ambient temperature around the photovoltaic comprehensively affect the efficiency of photovoltaic power generation:

[0226] ;

[0227] Among them, is the temperature of the photovoltaic device measured by the temperature sensor on the photovoltaic, is the ambient temperature, is the temperature coefficient, is the standard temperature of 25°C, is the weight of influence on photovoltaic power generation of temperature;

[0228] Fourth, the weight of influence on photovoltaic power generation of wind speed: Wind speed affects the heat dissipation of the photovoltaic and thus affects the power generation efficiency:

[0229] ;

[0230] Among them, is the actual wind speed, is the standard reference wind speed; is the wind speed influence coefficient, which is the influence weight of wind speed on photovoltaic power generation;

[0231] Fifth, the influence weight of air humidity on photovoltaic power generation: Different air humidities have different effects on the attachment of pollutants on the photovoltaic surface, further affecting the photovoltaic power generation efficiency. When the air humidity is greater than 80%, the influence weight of air humidity on photovoltaic power generation is assigned as 0.95. When the air humidity is greater than 30% and less than 80%, the influence weight of air humidity on photovoltaic power generation is assigned as 1. When the air humidity is less than 30%, the influence weight of air humidity on photovoltaic power generation is assigned as 0.98;

[0232] Sixth, the influence weight of precipitation on photovoltaic power generation: Rainwater can refract sunlight in the air, affecting the direct reception of light sources by the photovoltaic. There is a relationship between the precipitation of rainwater and the refraction of sunlight. When the precipitation is greater than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.2. When the precipitation is greater than 2 mm and less than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.05. When the precipitation is less than 2 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 0.9;

[0233] Seventh, the influence weight of air quality on photovoltaic power generation: Impurities such as dust in the air will reduce the intensity of sunlight radiation. When the air quality index is greater than 150, the influence weight of air quality on photovoltaic power generation is assigned as 0.8. When the air quality index is greater than 50 and less than 150, the influence weight of air quality on photovoltaic power generation is assigned as 0.9. When the air quality index is less than 50, the influence weight of air quality on photovoltaic power generation is assigned as 1;

[0234] Calculate the theoretical power generation of the photovoltaic according to the above different weather factors. The calculation formula is:

[0235] ;

[0236] Among them, is the power generation of the photovoltaic under ideal environmental conditions, is the theoretical power generation of the photovoltaic under actual environmental conditions. Then compare the actual power generation of the photovoltaic with the actual theoretical power generation. If the actual power generation of the photovoltaic can reach 90% of the actual theoretical power generation, it indicates that there is no obvious abnormality in the actual power generation of the photovoltaic. Set the photovoltaic power generation abnormality index , otherwise, it is considered that the photovoltaic power generation is abnormal. Let the photovoltaic power generation anomaly index be ;

[0237] C. Introduce a new pollutant recognition method based on adaptive feature fusion to identify the specific situation of small-scale pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photos in S2 as follows:

[0238] C.1. Use YOLOv8 for pollutant target detection and output the pollutant bounding box:

[0239] ;

[0240] Among them, represents the leftmost position of the pollutant area, represents the bottommost position of the pollutant area, represents the rightmost position of the pollutant area, represents the uppermost position of the pollutant area. Crop the target area based on the bounding box:

[0241] ;

[0242] C.2. Use Swin Transformer to extract pollutant area features through the sliding window attention mechanism:

[0243] ;

[0244] Calculate the spatial consistency of the pollutant area and merge pollutants with adjacent boundaries:

[0245] ;

[0246] C.3. Combine CNN and Swin Transformer to extract features:

[0247] ;

[0248] Calculate the final fused features:

[0249] ;

[0250] Among them, and are adaptive weights;

[0251] C.4. Introduce a multi-frame information fusion method to calculate the consistency of pollutants in multiple photos:

[0252] ;

[0253] Among them, when is greater than the retention threshold, retain ;

[0254] C.5. Use U-Net to segment the remaining pollutant areas to generate a pollutant mask:

[0255] ;

[0256] D. The specific steps for identifying large-scale pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photo in S2 using wide-area pollutant identification based on spatial color analysis are as follows:

[0257] D.1. Calculate the difference in color histograms between the clean photovoltaic panel and the panel in the actual image:

[0258] ;

[0259] where is the color space, respectively represent the red, green, and blue channels in the RGB space, represents the color channel in the th color interval frequency, represents the color histogram of the clean photovoltaic panel, represents the color histogram of the actual image;

[0260] b. Calculate the mean value of the color channels of the clean photovoltaic panel

[0261] ;

[0262] where is the color value of the pixel point in the clean photovoltaic panel, and calculate the mean value of the color channels of the panel in the actual image:

[0263] ;

[0264] where is the color mean value of the pixel point in the photovoltaic panel in the actual image, and calculate the difference in the mean values of the color channels between the clean photovoltaic panel and the panel in the actual image:

[0265] ;

[0266] c. Generate a mask for the pollutant area based on the comparison result of the difference in color histograms in a and the color histogram difference threshold and the comparison result of the difference in mean values of color channels in b and the difference threshold of mean values of color channels :

[0267] ;

[0268] E. Calculate the small - scale pollutant area and large - scale pollutant area through the number of pixels:

[0269] ;

[0270] Among them, the area of each pixel is 1 cm 2 , The number of pixels representing the small - scale pollutants is the pollutant area, The number of pixels representing the large - scale pollutants is the pollutant area, and pollution region represents the pollutant area;

[0271] F. Comprehensively analyze the necessity index of PV cleaning in A , the abnormal power generation index in B and the pollutant area in C and The specific steps to determine whether PV needs to be cleaned are as follows:

[0272] F.1. According to the influence of the PV cleaning necessity index, abnormal power generation index, and pollutant area on the PV cleaning demand, assign different importance weights to them: ;

[0273] F.2. Calculate the PV comprehensive cleaning index for pollutants in different ranges according to the importance weights in F.1:

[0274] ;

[0275] When or , the PV cleaner obtains the execution instruction to clean the pollutants, where is the comprehensive cleaning threshold interval for small - scale pollutants, is the comprehensive cleaning threshold interval for large - scale pollutants;

[0276] Adopt the above steps, according to the key weather data extracted in S2: air quality index precipitation , air humidity H, solar irradiance , wind speed and temperature and their corresponding weights to calculate the necessity index of PV cleaning , and then according to solar irradiance , cloud cover , temperature , wind speed , air humidity , precipitation and air quality index Calculate the theoretical power generation of a single photovoltaic panel under the actual weather conditions , compare it with the actual power generation of each photovoltaic panel to generate a photovoltaic power generation anomaly index , then use the pollutant recognition method based on adaptive feature fusion and the wide-area pollutant recognition method based on spatial color analysis to identify the optimized photovoltaic panel photos in S2 for small-range pollutants and large-range pollutant areas, generate a pollutant area mask, and calculate the small-range pollutant area by calculating the number of pixels and the large-range pollutant area , and then calculate the photovoltaic comprehensive cleaning index corresponding to the small-range pollutants and the photovoltaic comprehensive cleaning index corresponding to the large-range pollutants respectively according to the photovoltaic cleaning necessity index, the photovoltaic power generation anomaly index, the pollutant area and the assigned influence weights and the photovoltaic comprehensive cleaning index corresponding to the large-range pollutants , and then compare them with the small-range pollutant comprehensive cleaning threshold interval and the large-range pollutant comprehensive cleaning threshold interval respectively to generate an execution instruction for the photovoltaic cleaner to clean pollutants

[0277] S4. Analyze the temperature of the photovoltaic equipment in S1 to generate a cooling execution instruction for the photovoltaic cleaner. When the temperature of the photovoltaic equipment is greater than the high-temperature resistance threshold for normal operation of the photovoltaic, the photovoltaic cleaner obtains a cooling execution instruction

[0278] Adopt the above steps, compare the actual temperature of the photovoltaic equipment with the high-temperature resistance threshold for normal operation of the photovoltaic equipment to generate a cooling execution instruction for the photovoltaic cleaner

[0279] S5. According to the cleaning execution instruction of the photovoltaic cleaner in S3 and the cooling execution instruction of the photovoltaic cleaner in S4, the specific situation of the photovoltaic cleaner performing adaptive cleaning and cooling operations on the photovoltaic panel is as follows

[0280] A. The specific steps for the photovoltaic cleaner to execute the cleaning pollutant instruction are as follows

[0281] A.1. According to the pollutant information in S3, locate the number of the photovoltaic panel where the pollutant is located

[0282] A.2. According to the number of the photovoltaic panel in A.1, the specific steps for the photovoltaic cleaner corresponding to the number to execute the cleaning pollutant instruction are as follows

[0283] a. Use YOLOv8 to perform real-time target recognition of the photovoltaic cleaner on the monitored video stream. Taking the coordinate system of the monitoring screen as a reference, the photovoltaic cleaner starts running to the right from the starting point while the photovoltaic cleaner is lifted, the cleaning sponge in the photovoltaic cleaner does not contact the photovoltaic panel, and the water tank of the photovoltaic cleaner does not release water

[0284] b. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutant, the photovoltaic cleaner descends while the water pipe on the right side of the water tank of the photovoltaic cleaner discharges water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutant. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutant, the photovoltaic cleaner rises while the water pipe on the right side of the water tank of the photovoltaic cleaner stops discharging water, and the first cleaning of the target pollutant ends. Record the water discharge time of this process, and repeat this operation for the pollutants that have not been cleaned for the first time;

[0285] c. After the first cleaning of all pollutants is completed, the photovoltaic cleaner starts to return. During this process, the pollutants cleaned in step b are cleaned again. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutant, the photovoltaic cleaner descends while the water pipe on the left side of the water tank of the photovoltaic cleaner discharges water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutant. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutant, the photovoltaic cleaner rises while the water pipe on the left side of the water tank of the photovoltaic cleaner stops discharging water, and the second cleaning of the target pollutant ends. Record the water discharge time of this process, and repeat this operation for the pollutants that have not been cleaned for the second time;

[0286] d. The photovoltaic cleaner returns to the starting point, and statistically calculates the total water discharge time of the water tank for the two round trips of the photovoltaic cleaner;

[0287] B. The specific steps for the photovoltaic cleaner to execute the cooling instruction are as follows:

[0288] a. The photovoltaic cleaner runs back and forth two times from the starting point to the rightmost end of the photovoltaic panel. During the process, the photovoltaic cleaner rises, and the water pipes on both the left and right sides of the water tank discharge water simultaneously;

[0289] b. The photovoltaic cleaner returns to the starting point, and the water pipes on both the left and right sides of the water tank stop discharging water. Statistically calculate the water discharge time for the two round trips of the photovoltaic cleaner;

[0290] By adopting the above steps, according to the cleaning execution instruction of the photovoltaic cleaner generated in S3 and the cooling execution instruction of the photovoltaic cleaner generated in S4, clean the specified pollutant area and sprinkle water to cool down the photovoltaic panel.

[0291] S6. According to the pollution degree of the cleaning sponge during the cleaning process in S5 and the water consumption of the photovoltaic cleaner when performing the cleaning or cooling task, the specific situations of automatically completing the replacement, cleaning of the cleaning sponge of the photovoltaic cleaner and the water replenishment operation of the water tank are as follows:

[0292] a. There are the following two operation methods for the automatic cleaning of the cleaning sponge of the photovoltaic cleaner:

[0293] First, replace the cleaning sponge: Use the wide - area pollutant identification method based on spatial color analysis in S3 to identify the photo of the cleaning sponge of the optimized photovoltaic cleaner in S2, and judge the pollution condition of the cleaning sponge. During the operation of the photovoltaic cleaner, when the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the first time, the photovoltaic cleaner controls the cleaning sponge to turn over and change sides. When the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the second time, the photovoltaic cleaner controls to replace a new cleaning sponge;

[0294] Second, clean the cleaning sponge: When the photovoltaic cleaner returns to the starting point after completing the cleaning task, the flushing nozzle at the starting point is automatically turned on to flush the used cleaning sponge in the photovoltaic cleaner, and at the same time, the photovoltaic cleaner controls the cleaning sponge to rotate.

[0295] b. After the photovoltaic cleaner completes the cleaning or cooling task and returns to the starting point, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishment device are turned on. The water filling port of the photovoltaic water replenishment device and the water replenishment port of the water tank of the photovoltaic cleaner are magnetically docked. At the same time, the water filling pipe valve is opened to replenish water to the water tank. The water replenishment volume is calculated according to the total water discharge time of 300Ml / s. After the water replenishment is completed, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishment device are turned off and the water filling pipe valve is closed at the same time;

[0296] By adopting the above steps, during the cleaning process, it is automatically replaced according to the pollution degree of the cleaning sponge. After completing the cleaning task and the sprinkler cooling task, the water tank of the photovoltaic cleaner is automatically replenished with water, and the cleaning sponge is automatically sprayed and cleaned.

[0297] Embodiment 2

[0298] This embodiment discloses a photovoltaic adaptive cleaning and cooling system based on multi - source information perception and intelligent decision - making, which is used to implement the photovoltaic adaptive cleaning and cooling method based on multi - source information perception and intelligent decision - making described in Embodiment 1, as Figure 2 shown. The system includes the following functional modules:

[0299] The data acquisition module is used to acquire weather data, photovoltaic power generation data, photovoltaic device temperature, photovoltaic panel monitoring video, and photos of the cleaning sponge of the photovoltaic cleaner.

[0300] The data pre - processing module is used to extract key weather data, power generation data of each photovoltaic panel, photovoltaic device temperature, and optimize the intercepted photovoltaic panel photos and photos of the cleaning sponge of the photovoltaic cleaner;

[0301] The photovoltaic pollution condition analysis module is used to analyze the cleaning requirements of the photovoltaic panel based on the weather condition, photovoltaic power generation condition, and actual attachment condition of pollutants on the photovoltaic panel, and give the cleaning execution instruction to the photovoltaic cleaner;

[0302] A photovoltaic device temperature analysis module, which is used to analyze the cooling requirement of the photovoltaic device according to the temperature of the photovoltaic device and give a cooling execution instruction to the photovoltaic cleaner;

[0303] A photovoltaic cleaner scheduling module, which is used to execute the photovoltaic cleaning and cooling instructions according to the cleaning and cooling requirements of the photovoltaic panel;

[0304] A photovoltaic cleaner water replenishing module, which is used to automatically replenish water according to the water consumption of the photovoltaic cleaner during the cleaning and cooling operations;

[0305] A photovoltaic cleaner cleaning module, which is used to replace or clean the cleaning sponge in the photovoltaic cleaner according to its pollution condition.

[0306] Embodiment 3

[0307] This embodiment discloses a storage medium storing a program, which when executed by a processor, implements the photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making described in Embodiment 1.

[0308] The storage medium in this embodiment may be a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), a USB flash drive, a mobile hard disk, or other media.

[0309] Embodiment 4

[0310] This embodiment discloses a computing device including a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, it implements the photovoltaic adaptive cleaning and cooling method based on multi-source information perception and intelligent decision-making described in Embodiment 1.

[0311] The computing device described in this embodiment may be a desktop computer, a laptop computer, a smart phone, a PDA handheld terminal, a tablet computer, a programmable logic controller (PLC), or other terminal devices with processor functions.

[0312] In summary, the present invention proposes a photovoltaic adaptive cleaning and cooling method and system based on multi-source information perception and intelligent decision-making. By using weather information, photovoltaic power generation data, and computer vision-based pollutant recognition technology, it accurately determines the cleaning timing of different types of pollutants and intelligently formulates cleaning strategies. At the same time, according to the equipment temperature, it dynamically schedules the photovoltaic cleaner to perform sprinkler cooling operations to ensure the efficient operation of the photovoltaic panel. In addition, during the cleaning process, the system can real-time monitor the pollution degree of the cleaning sponge, realize intelligent replacement and cleaning, and ensure the continuous stability of the cleaning effect. Finally, combined with the water consumption of the cleaning and cooling tasks, it realizes automated water replenishment management. The whole process realizes unmanned and intelligent autonomous operation, not only effectively improves the photovoltaic power generation efficiency, but also greatly reduces the operation and maintenance costs of the photovoltaic power station, has practical popularization value, and is worthy of promotion.

[0313] The above embodiments are only the preferred embodiments of the present invention, and do not limit the scope of implementation of the present invention. Therefore, all changes made according to the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A photovoltaic multi-source perception intelligent cleaning and cooling method, characterized in that, Including the following steps: S1. Obtain monitoring data, namely weather data, photovoltaic power generation data, photovoltaic equipment temperature, pollution information of photovoltaic panels, and pollution information of the cleaning sponges of photovoltaic cleaners; S2. Data preprocessing: Extract the key data of the weather data and photovoltaic power generation data in S1 and optimize the pollution information of the photovoltaic panels and the pollution information of the cleaning sponges of the photovoltaic cleaners. The specific situations of extracting the key data of the weather data and photovoltaic power generation data in S1 and optimizing the pollution information of the photovoltaic panels and the pollution information of the cleaning sponges of the photovoltaic cleaners are as follows: A. Extract the air quality index, temperature, wind speed, precipitation, air humidity, and solar irradiance in the weather data; B. Extract the actual power generation of each photovoltaic panel from the photovoltaic power generation data, and obtain the theoretical power generation of each photovoltaic panel under ideal conditions from the equipment instruction manual; C. The specific steps for optimizing the intercepted photos of photovoltaic panels and the cleaning sponges of photovoltaic cleaners are as follows: C.

1. The specific steps for denoising, sharpening, and super-resolution optimizing the multiple monitored photos of photovoltaic panels and the cleaning sponges of photovoltaic cleaners by introducing a new photovoltaic pollutant image enhancement method based on adaptive frequency domain fusion and super-resolution reconstruction are as follows: a. Decompose the image by wavelet transform WT, and decompose the image into a low-frequency part L and a high-frequency part H: ； Among them, the low-frequency L contains illumination and shadow information, and the high-frequency H contains pollutant edges and high-frequency noise information; b. Calculate the noise standard deviation using the BayesShrink method : ； Among them, is the median value function, and then the adaptive threshold is obtained , filtering image noise: ； Among them, is the signal standard deviation, is the sign function, is the high-frequency component after image filtering of noise; c. The high-frequency components after noise filtering and the low-frequency part L are combined and then subjected to inverse wavelet transform for reconstruction to obtain the image after noise filtering: C.

2. The specific steps for enhancing image sharpness based on structural tensor gradient are as follows: a. Calculate the Sobel direction gradient; ； Among them, represents the horizontal gradient along direction, represents the horizontal gradient along direction, and then calculate the structure tensor Identify the pollutant boundary: ； Among them, and represent the main directional change rate of the structure; b. Calculate the sharpening weight of each pixel; ； c. Perform adaptive sharpening on the image; ； Among them, is the second-order Laplacian derivative for enhancing edges, is the adaptive sharpening coefficient; C.

3. The specific steps for optimizing the image sharpness using super-resolution based on adversarial learning are as follows: a. Use ESRGAN for super-resolution reconstruction to obtain the super-resolution reconstructed image; ； Among them, is the ESRGAN generator, which calculates the feature difference between the SR image and the original high-definition image: ； Among them, is the high-level feature of the VGG network extraction layer, is the actual real image, and the adversarial loss is used to naturalize the super-resolution image: ； Among them; is a discriminator network; b. Train the final target to optimize the super-resolution image; ； Among them; is the L1 loss, and are hyperparameter weights for controlling different losses; S3. Jointly analyze the key data of the weather data and photovoltaic power generation data in S2 and the pollutant attachment situation on the surface of the photovoltaic panels in the optimized photovoltaic panel photos, and generate a cleaning execution instruction for the photovoltaic cleaner. The specific situations of jointly analyzing the key data of the weather data and photovoltaic power generation data in S2 and the pollutant attachment situation on the surface of the photovoltaic panels in the optimized photovoltaic panel photos and generating a cleaning execution instruction for the photovoltaic cleaner are as follows: A. According to the influence of weather factors at different levels on the pollutant attachment degree and the interaction between various weather factors, assign corresponding pollutant attachment weights to different weather factors. The pollutant attachment weight can indicate whether the weather factor is conducive to pollutant sedimentation. A positive weight indicates that the weather factor is conducive to pollutant sedimentation, a weight of zero indicates that the weather factor has no obvious influence on pollutant sedimentation, and a negative weight indicates that the weather factor is not conducive to pollutant sedimentation. By comprehensively analyzing the comprehensive influence of various weather factors on pollutant attachment, determine the pollutant attachment weights of the following six key weather factors: First, the pollutant attachment weight corresponding to the air quality index: If the air quality index within the selected period is less than 50, it is considered that the pollutant accumulation is less, and the pollutant attachment weight corresponding to the air quality index is given as -0.2; if the air quality index within the selected period is greater than 50 and less than 150, it is considered that the pollutant accumulation rate is relatively fast, and the pollutant attachment weight corresponding to the air quality index is given as 0.1; if the air quality index within the selected period is greater than 150, it is considered that the pollutant accumulation rate is extremely fast, and the pollutant attachment weight corresponding to the air quality index is given as 0.3; Second, the pollutant attachment weight corresponding to the precipitation: If the precipitation within the selected time period is less than 2 mm, it not only cannot clean the dust on the photovoltaic surface but may even adsorb the dust in the air. Therefore, the pollutant attachment weight corresponding to the precipitation is given as 0.2; if the precipitation within the selected period is greater than 2 mm and less than 10 mm, it may only clean some of the pollutants on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the precipitation is given as -0.1; if the precipitation within the selected period is greater than 10 mm, it may clean most of the pollutants on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the precipitation is given as -0.3; Third, the pollutant attachment weight corresponding to the air humidity: If the air humidity within the selected time period is less than 30%, it indicates that the air is dry and it is easy to form electrostatic adsorption pollution. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.2; if the air humidity within the selected time period is greater than 30% and less than 80%, it has no obvious effect on the attachment of pollutants. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.0; if the air humidity within the selected time period is greater than 80%, it may cause mud-like dust pollution on the photovoltaic surface. Therefore, the pollutant attachment weight corresponding to the air humidity is given as 0.15; Fourth, the pollutant adhesion weight corresponding to solar irradiance: When the solar irradiance within the selected time period is less than 200 W / m 2 , the temperature of the photovoltaic surface is low, and pollutants are not easily solidified. Therefore, the pollutant adhesion weight corresponding to solar irradiance is given as -0.1; When the solar irradiance within the selected time period is greater than 200 W / m 2 and less than 600 W / m 2 , there is no obvious effect on the adhesion of pollutants. Therefore, the pollutant adhesion weight corresponding to solar irradiance is given as 0.0; When the solar irradiance within the selected time period is greater than 600 W / m 2 , it may cause the temperature of the photovoltaic surface to be high, resulting in the solidification of pollutants and difficult cleaning. Therefore, the pollutant adhesion weight corresponding to solar irradiance is given as 0.2; Fifth, the pollutant attachment weight corresponding to the wind speed: If the wind speed within the selected time period is less than 1 m / s, it indicates that the air is relatively static and the pollutant sedimentation is serious. Therefore, the pollutant attachment weight corresponding to the wind speed is given as 0.3; if the wind speed within the selected time period is greater than 1 m / s and less than 5 m / s, the wind speed is moderate and helps to reduce the pollutant sedimentation. Therefore, the pollutant attachment weight corresponding to the wind speed is given as -0.1; if the wind speed within the selected time period is greater than 5 m / s, it may bring new dust. Therefore, the pollutant attachment weight corresponding to the wind speed is given as 0.1; Sixth, the pollutant attachment weight corresponding to the temperature: If the temperature within the selected time period is less than 20 °C, the environment is in a low temperature and the pollutants on the photovoltaic surface are not easy to solidify. Therefore, the pollutant attachment weight corresponding to the temperature is given as -0.2; if the temperature within the selected time period is greater than 20 °C and less than 30 °C, the environmental temperature is moderate and the pollutants are moderately solidified. Then the pollutant attachment weight corresponding to the temperature is given as -0.1; if the temperature within the selected time period is greater than 30 °C, the environment is in a high temperature and the pollutants on the photovoltaic surface are significantly solidified and difficult to clean. Therefore, the pollutant attachment weight corresponding to the temperature is given as 0.3; Calculate the necessity index of PV cleaning based on the above different weather factors , and the calculation formula is as follows: ； Among them, is the air quality index, is the precipitation, is the air humidity, is the solar irradiance, is the wind speed, is the temperature, is the pollutant attachment weight corresponding to the air quality index, is the pollutant attachment weight corresponding to the precipitation, is the pollutant attachment weight corresponding to the air humidity, is the pollutant attachment weight corresponding to the solar irradiance, is the pollutant attachment weight corresponding to the wind speed and is the pollutant attachment weight corresponding to the temperature; B. According to the influence of weather factors at different levels on photovoltaic power generation and the interaction between various weather factors, corresponding weights of influence on photovoltaic power generation are assigned to different weather factors. The weight of influence on photovoltaic power generation can indicate whether the weather factor is favorable for photovoltaic power generation. The closer the influence weight of the weather factor is to 1, the more favorable the weather factor condition is for photovoltaic power generation. By comprehensively analyzing the comprehensive influence of various weather factors on photovoltaic power generation, the weights of influence on photovoltaic power generation of the following seven key weather factors are determined: First, the weight of influence on photovoltaic power generation of solar irradiance: Calculate the ratio of the actual solar irradiance to the solar irradiance in the theoretical test scenario as the weight of influence on power generation of solar irradiance: ； Among them, is the actual solar irradiance, is the solar irradiance under the theoretical test scenario; Second, the weight of influence on photovoltaic power generation of cloud cover: Use the exponential decay model to calculate the weight of influence on photovoltaic power generation of cloud cover: ； Among them, is the cloud cover, is an empirical coefficient that can be adjusted according to measured data, which is the influence weight of cloud cover on photovoltaic power generation; Third, the weight of influence on photovoltaic power generation of temperature: The temperature of the photovoltaic module and the ambient temperature around the photovoltaic comprehensively affect the efficiency of photovoltaic power generation: ； Among them, is the temperature of the photovoltaic device measured by the temperature sensor on the photovoltaic panel, is the temperature coefficient, is the standard temperature of 25 °C, which is the influence weight of temperature on photovoltaic power generation; Fourth, the weight of influence on photovoltaic power generation of wind speed: The wind speed affects the heat dissipation of the photovoltaic and thus affects the power generation efficiency: ； Among them, is the actual wind speed, is the standard reference wind speed; is the wind speed influence coefficient, which is the influence weight of wind speed on photovoltaic power generation; Fifth, the influencing weight of air humidity on photovoltaic power generation: Different air humidities have different effects on the attachment of pollutants on the photovoltaic surface, further affecting the photovoltaic power generation efficiency. When the air humidity is greater than 80%, the influencing weight of air humidity on photovoltaic power generation is assigned as 0.

95. When the air humidity is greater than 30% and less than 80%, the influencing weight of air humidity on photovoltaic power generation is assigned as 1. When the air humidity is less than 30%, the influencing weight of air humidity on photovoltaic power generation is assigned as 0.98; Sixth, the influence weight of precipitation on photovoltaic power generation: Rainwater in the air can cause refraction of sunlight, affecting the direct reception of light sources by photovoltaics. There is a relationship between the precipitation of rainwater and the refraction of sunlight. When the precipitation is greater than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.

2. When the precipitation is greater than 2 mm and less than 10 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 1.

05. When the precipitation is less than 2 mm, the influence weight of precipitation on photovoltaic power generation is assigned as 0.9; The seventh one, the influence weight of photovoltaic power generation on air quality: Impurities such as dust in the air will reduce the intensity of sunlight radiation. When the air quality index is greater than 150, the influence weight of photovoltaic power generation on air quality is assigned as 0.

8. When the air quality index is greater than 50 and less than 150, the influence weight of photovoltaic power generation on air quality is assigned as 0.

9. When the air quality index is less than 50, the influence weight of photovoltaic power generation on air quality is assigned as 1; Calculate the theoretical power generation of the photovoltaic according to the above different weather factors. The calculation formula is: ； Among them, is the power generation of the photovoltaic in ideal environmental conditions, is the theoretical power generation of the photovoltaic in actual environmental conditions. Then, compare the actual power generation of the photovoltaic with the actual theoretical power generation. If the actual power generation of the photovoltaic can reach 90% of the actual theoretical power generation, it indicates that there is no obvious abnormality in the actual power generation of the photovoltaic. Let the photovoltaic power generation anomaly index , otherwise, it is considered that the photovoltaic power generation is abnormal. Let the photovoltaic power generation anomaly index ; C. Introduce a new pollutant recognition method based on adaptive feature fusion to identify the specific situation of small-scale pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photo in step 2) as follows: C.

1. Use YOLOv8 for pollutant target detection and output the pollutant bounding box: ； Among them, represents the leftmost position of the pollutant area, represents the bottommost position of the pollutant area, represents the rightmost position of the pollutant area, represents the uppermost position of the pollutant area, and crop the target area based on the bounding box: ； C.

2. Use Swin Transformer to extract the pollutant area features through the sliding window attention mechanism: ； Calculate the spatial consistency of the pollutant area and merge the pollutants with adjacent boundaries: ； C.

3. Combine CNN and Swin Transformer to extract features: ； Calculate the final fusion feature: ； Among them, and are adaptive weights; C.

4. Introduce a multi-frame information fusion method to calculate the consistency of pollutants in multiple photos: ； Among them, when is greater than the retention threshold, retain ; C.

5. Use U-Net to segment the remaining pollutant areas to generate a pollutant mask: ； D. Use the wide-area pollutant recognition based on spatial color analysis to identify the specific steps of large-scale pollutants on the surface of the photovoltaic panel in the optimized photovoltaic panel monitoring photo in step 2) as follows: D.

1. Calculate the difference in color histograms between the clean photovoltaic panel and the panel in the actual image: ； Among them, is a color space, respectively represent the red, green, and blue channels in the RGB space, represents the color channel in the frequency of the represents the color histogram of the cleaned photovoltaic panel, represents the color histogram of the actual image; D.

2. Calculate the mean value of the color channels of the clean photovoltaic panel ； Among them, is the color value of the pixel points in the cleaning photovoltaic panel to calculate the mean value of the color channels of the panel in the actual image: ； Among them, is the color mean value of the pixel points in the photovoltaic panel in the actual image Calculate the difference in the color channel means of the clean photovoltaic panel and the photovoltaic panel in the actual image: ； D.

3. Generate a mask for the pollutant area based on the comparison result between the difference in the color histogram in a and the color histogram difference threshold and the comparison result between the difference in the color channel means in b and the color channel mean difference threshold : ； E. Calculate the area of small-scale pollutants and large-scale pollutants through the number of pixels: ； Among them, the area of each pixel is 1 cm 2 , The number of pixels representing small-scale pollutants is the pollutant area, The number of pixels representing large-scale pollutants is the pollutant area, and pollution region represents the pollutant area; F. Comprehensive analysis of the necessity index of photovoltaic cleaning in A , the abnormal photovoltaic power generation index in B , and the pollutant area in C and The specific steps to determine whether the photovoltaic needs to be cleaned are as follows: F.

1. According to the influence of the photovoltaic cleaning necessity index, the photovoltaic power generation anomaly index, and the pollutant area on the photovoltaic cleaning demand, different importance weights are assigned to them respectively: ; F.

2. Calculate the comprehensive photovoltaic cleaning index of pollutants in different ranges according to the importance weights in F.1: ； When or the photovoltaic cleaner obtains an execution instruction to clean pollutants, where is the comprehensive cleaning threshold interval for small-range pollutants, is the comprehensive cleaning threshold interval for large-range pollutants; S4. Analyze the temperature of the photovoltaic equipment in S1 and generate a cooling execution instruction for the photovoltaic cleaner; S5. According to the cleaning execution instruction of the photovoltaic cleaner in S3 and the cooling execution instruction of the photovoltaic cleaner in S4, the photovoltaic cleaner performs adaptive cleaning and cooling operations on the photovoltaic panel; S6. According to the water consumption of the photovoltaic cleaner and the pollution degree of the cleaning sponge in S5, automatically complete the replacement and cleaning of the cleaning sponge and the water replenishment operation of the water tank of the photovoltaic cleaner.

2. The photovoltaic multi-source perception intelligent cleaning and cooling method according to claim 1, characterized in that: In S1, the weather data is real-time weather data obtained by accessing the China Meteorological Data Network interface, which is text data describing the weather conditions; the photovoltaic power generation data is the photovoltaic power generation recorded in the photovoltaic power station system; the photovoltaic device temperature is provided by the temperature sensor of the photovoltaic panel; the pollution information of the photovoltaic panel is the video stream file of the photovoltaic panel captured by monitoring; the pollution information of the cleaning sponge of the photovoltaic cleaner is the photo of the cleaning sponge taken by the camera under the photovoltaic panel. Among them, every half an hour, the weather data is automatically updated synchronously, the photovoltaic power generation data is obtained, and the operation of intercepting photos from the photovoltaic panel video stream file is performed. The cleaning sponge photo is taken every half a minute when the photovoltaic cleaner executes the cleaning instruction.

3. The photovoltaic multi-source perception intelligent cleaning and cooling method according to claim 1, characterized in that: In S4, the temperature of the photovoltaic device in S1 is analyzed to generate a cooling execution instruction for the photovoltaic cleaner. When the temperature of the photovoltaic device is greater than the high-temperature resistance threshold for normal operation of the photovoltaic, the photovoltaic cleaner obtains the cooling execution instruction.

4. The photovoltaic multi-source perception intelligent cleaning and cooling method according to claim 1, characterized in that: In S5, according to the cleaning execution instruction of the photovoltaic cleaner in S3 and the cooling execution instruction of the photovoltaic cleaner in S4, the specific situation of the photovoltaic cleaner performing adaptive cleaning and cooling operations on the photovoltaic panel is as follows: A. The specific steps for the photovoltaic cleaner to execute the cleaning pollutant instruction are as follows: A.

1. According to the information of the pollutant in S3, locate the number of the photovoltaic panel where the pollutant is located. A.

2. According to the number of the photovoltaic panel in A.1, the specific steps for the photovoltaic cleaner corresponding to the number to execute the cleaning pollutant instruction are as follows: a. Use YOLOv8 to perform real-time target recognition of the photovoltaic cleaner on the monitored video stream. Taking the coordinate system of the monitoring screen as a reference, the photovoltaic cleaner starts running to the right from the starting point while the photovoltaic cleaner lifts up, and the cleaning sponge in the photovoltaic cleaner does not contact the photovoltaic panel, and the water tank of the photovoltaic cleaner does not release water. b. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutant, the photovoltaic cleaner lowers while the right water pipe of the water tank of the photovoltaic cleaner releases water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutant. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutant, the photovoltaic cleaner lifts up while the right water pipe of the water tank of the photovoltaic cleaner stops releasing water, and the first cleaning of the target pollutant ends. Record the water release time of this process, and repeat this operation for the pollutants that have not been cleaned for the first time. c. After the first cleaning of all pollutants is completed, the photovoltaic cleaner starts to return. During this process, the pollutants cleaned in step b are cleaned again. When the left coordinate of the photovoltaic cleaner is 10 cm away from the rightmost position of the pollutants, the photovoltaic cleaner descends and the left water pipe of the water tank of the photovoltaic cleaner discharges water, so that the cleaning sponge in the photovoltaic cleaner contacts the photovoltaic panel to clean the pollutants. When the right coordinate of the photovoltaic cleaner is 10 cm away from the leftmost position of the pollutants, the photovoltaic cleaner lifts up and the left water pipe of the water tank of the photovoltaic cleaner stops discharging water. The second cleaning of the target pollutants is completed, and the water discharge time of this process is recorded. Repeat this operation for the pollutants that have not been cleaned for the second time; d. The photovoltaic cleaner returns to the starting point, and the total water discharge time of the water tank for two round trips of the photovoltaic cleaner is statistically calculated; B. The specific steps for the photovoltaic cleaner to execute the cooling instruction are as follows: a. The photovoltaic cleaner makes two round trips from the starting point to the rightmost end of the photovoltaic panel. During the process, the photovoltaic cleaner lifts up, and the water pipes on both the left and right sides of the water tank discharge water simultaneously; b. The photovoltaic cleaner returns to the starting point, the water pipes on both the left and right sides of the water tank stop discharging water, and the water discharge time for two round trips of the photovoltaic cleaner is statistically calculated.

5. The photovoltaic multi-source perception intelligent cleaning and cooling method according to claim 1, wherein: In S6, according to the pollution degree of the cleaning sponge during the cleaning process in S5 and the water consumption of the photovoltaic cleaner when performing the cleaning or cooling task, the specific conditions for automatically replacing the cleaning sponge of the photovoltaic cleaner, cleaning, and replenishing water to the water tank are as follows: a. There are the following two operation methods for the automatic cleaning of the cleaning sponge of the photovoltaic cleaner: The first is to replace the cleaning sponge: Use the wide-area pollutant recognition method based on spatial color analysis in S3 to identify the photo of the cleaning sponge of the optimized photovoltaic cleaner in S2, and judge the pollution situation of the cleaning sponge. During the operation of the photovoltaic cleaner, when the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the first time, the photovoltaic cleaner controls the cleaning sponge to turn over. When the pollution degree of each cleaning sponge is greater than the pollution threshold of the cleaning sponge for the second time, the photovoltaic cleaner controls to replace a new cleaning sponge; The second is to clean the cleaning sponge: When the photovoltaic cleaner returns to the starting point after completing the cleaning task, the flushing nozzle at the starting point is automatically turned on to flush the used cleaning sponge in the photovoltaic cleaner, and at the same time, the photovoltaic cleaner controls the cleaning sponge to rotate; b. After the photovoltaic cleaner completes the cleaning or cooling task and returns to the starting point, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishing device are turned on. The water filling port of the photovoltaic water replenishing device and the water replenishing port of the water tank of the photovoltaic cleaner are magnetically docked, and at the same time, the water replenishing water pipe valve is opened to replenish water to the water tank. The water replenishing amount is calculated at 300 Ml / s according to the total water discharge time. After the water replenishing is completed, the battery on the photovoltaic cleaner and the power supply of the photovoltaic water replenishing device are turned off and the water replenishing water pipe valve is closed.

6. A photovoltaic multi-source perception intelligent cleaning and cooling system, characterized in that For implementing the photovoltaic multi-source perception intelligent cleaning and cooling method according to any one of claims 1 to 5, it includes: A data acquisition module, configured to acquire weather data, photovoltaic power generation data, the temperature of photovoltaic equipment, the pollution information of photovoltaic panels, and the pollution information of the cleaning sponges of photovoltaic cleaners; A data preprocessing module, configured to extract key weather data, the power generation data of each photovoltaic panel, the temperature of photovoltaic equipment, and optimize the captured photos of photovoltaic panels and the photos of the cleaning sponges of photovoltaic cleaners; A photovoltaic pollution condition analysis module, configured to analyze the cleaning requirements of photovoltaic panels based on weather conditions, photovoltaic power generation conditions, and the actual attachment of pollutants on photovoltaic panels, and give cleaning execution instructions to photovoltaic cleaners; A photovoltaic equipment temperature analysis module, configured to analyze the cooling requirements of photovoltaic equipment based on the temperature of photovoltaic equipment, and give cooling execution instructions to photovoltaic cleaners; A photovoltaic cleaner scheduling module, configured to execute photovoltaic cleaning and cooling instructions according to the cleaning and cooling requirements of photovoltaic panels; A photovoltaic cleaner water replenishment module, configured to automatically replenish water according to the water consumption of photovoltaic cleaners during cleaning and cooling operations; A photovoltaic cleaner cleaning module, configured to replace or clean the cleaning sponges in photovoltaic cleaners according to the pollution conditions of the cleaning sponges; 7. A storage medium stores a program, characterized in that, When the program is executed by a processor, it implements the photovoltaic multi-source perception intelligent cleaning and cooling method according to any one of claims 1 to 5.

8. A computing device, comprising a processor and a memory for storing processor-executable programs, characterized in that, When the processor executes the program stored in the memory, it implements the photovoltaic multi-source perception intelligent cleaning and cooling method according to any one of claims 1 to 5.

Images