Intelligent system for cleaning photovoltaic cell based on ultrasonic waves

Abstract

The invention discloses an intelligent system for cleaning a photovoltaic cell based on ultrasonic waves. The intelligent system comprises a cleaning agent, a first intelligent control module and a second intelligent control module. The first intelligent control module regulates and controls the output power and frequency of the ultrasonic generator according to the stain type of the photovoltaic cell, and adjusts the amplitude of an amplitude-change pole of the transducer; the cleaning solvent preparation unit is subjected to solvent proportioning and blending; the second intelligent control module performs spectrum analysis according to the returned cleaning waveform of the cleaned photovoltaic cell to judge the cleaning effect; the transportation mechanism dispatches unmanned automatic control to convey the intelligent agent to the photovoltaic cell for cleaning stains at different positions; the advanced ultrasonic cleaning technology is deeply fused with artificial intelligence, automatic control and robot technologies, so that the core pain points of a traditional photovoltaic cell cleaning mode in the aspects of efficiency, quality, cost, applicability and the like are solved, and the spanning from extensive manual cleaning to precise intelligent cleaning is achieved; the method has remarkable technical progress and wide industrial application prospects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power station operation and maintenance, and particularly relates to an intelligent system based on ultrasonic cleaning of photovoltaic cells. BACKGROUND

[0002] With the global energy structure accelerating the transformation to green and low carbon, photovoltaic power generation, as the main force of clean energy, is realizing large-scale development at an unprecedented speed. The installed capacity of large ground power stations, distributed agricultural / fishing complementary projects and industrial and commercial rooftop photovoltaics continues to rise, and the single size and geographical distribution complexity of photovoltaic power stations have significantly increased. However, in the operation and maintenance of the whole life cycle of photovoltaic power stations, the pollution problem of the surface of photovoltaic components has become one of the key factors leading to power generation loss and increasing the levelized cost of electricity (LCOE).

[0003] Photovoltaic panels are exposed to the natural environment all year round, and their surfaces are prone to accumulate various types of pollutants. Common dust and sand can uniformly reduce the light transmittance; bird droppings, oil stains (such as from nearby industries or traffic), pollen, tree sap, etc. Organic matter can form stubborn stains, not only blocking light, but also possibly causing "hot spot effect", accelerating component aging and even causing permanent damage; Photovoltaic arrays near construction sites or roads may also be attached to cement dust, mineral dust and other hard pollutants. Studies have shown that in heavily polluted areas or seasons, the annual average power generation efficiency loss caused by stains can reach 10%-30%, and in specific cases (such as in dry and rainy areas with sticky dirt), the loss is even higher, causing huge economic losses to power station investors.

[0004] In the face of this serious challenge, the existing mainstream cleaning technology has significant limitations and cannot meet the operation and maintenance needs of large-scale and intelligent power stations: Manual cleaning: relies on manual operation, has high labor intensity, low efficiency, safety hazards and other problems. The cleaning quality is uneven, and the use of rough tools or improper operation can easily scratch the anti-reflection coating of the component surface glass, affecting the light transmittance and possibly inducing corrosion. In large power stations, it is difficult to organize and coordinate manual cleaning, and the single operation period is long, making it difficult to respond quickly in the best window period (such as after a sandstorm, dry season), and as the labor cost continues to rise, its economy is deteriorating.

[0005] Traditional high-pressure water / vehicle-mounted cleaning: Although it improves efficiency to some extent, it still belongs to extensive cleaning. Its core defect is the "one-size-fits-all" cleaning mode: regardless of the type and degree of adhesion of the stain, similar water pressure and flow rate are used for flushing. For stubborn stains such as dried bird droppings and viscous oil stains, the cleaning effect is limited, and repeated flushing or manual wiping is required. More importantly, this process lacks a pretreatment step to soften and decompose stains, and lacks real-time detection and judgment of cleaning effect, leading to excessive water use to achieve visual cleaning, which is particularly problematic in water-scarce areas, and cannot guarantee that stains are completely removed to restore the best performance of the component.

[0006] Early automation / ultrasonic cleaning equipment: Some fixed automatic cleaning devices (such as fixed spraying, track equipment with brush rollers) or experimental fixed ultrasonic cleaning tanks, although they reduce reliance on manual labor, have serious adaptability problems. They are usually designed for specific arrays and cannot be flexibly applied to different inclination angles, different layouts, and complex power station sites with terrain undulations. Fixed ultrasonic cleaning cannot handle photovoltaic panels already installed on brackets. In addition, most of these devices run on pre-set fixed programs, lack intelligent sensing and feedback mechanisms based on stain status, and cannot dynamically optimize cleaning parameters (such as mechanical force, water temperature, cleaning agent concentration, ultrasonic power / frequency, etc.) according to the degree of contamination and the type of stain, resulting in waste of energy and materials or incomplete cleaning.

[0007] Therefore, there is an urgent need for a technological revolution in the field of photovoltaic power station operation and maintenance to develop a new generation of cleaning system that can deeply integrate efficient physical / chemical cleaning mechanisms, intelligent environmental sensing and decision-making, and flexible automatic execution capabilities. The ideal system should be able to accurately identify and pretreat different contaminants, use multiple cleaning methods in coordination (such as ultrasonic pretreatment of loose stains combined with interface, followed by high-pressure water for efficient stripping), and integrate real-time effect detection and closed-loop control, while using flexible platforms such as drones and mobile robots to achieve no-dead-zone, self-adaptive coverage operations on large-scale, complex terrain power stations. Such a system not only greatly improves cleaning efficiency and quality, ensuring power generation income and reducing operation and maintenance costs, but also is an inevitable trend towards intelligent, unmanned, and lean operation and maintenance of photovoltaic power stations. SUMMARY

[0008] To solve the technical problems existing in the prior art, the present application provides an intelligent agent for ultrasonic cleaning of photovoltaic cells, which cleans the stains on the surface of the photovoltaic cell through the return waveform spectrum of ultrasonic cleaning, avoids damage to the photovoltaic surface during the cleaning process, and prolongs the service life of the photovoltaic cell. At the same time, the present application can automatically match the cleaning solvent according to the type of stains on the surface of the photovoltaic cell, and optimize the ultrasonic cleaning effect.

[0009] The present application adopts the following technical solutions: An intelligent system for cleaning photovoltaic cells based on ultrasonic waves, the intelligent system comprising a cleaning intelligent agent with a moving mechanism, a first intelligent control module, a second intelligent control module, a third intelligent control module and a drone; wherein: the cleaning intelligent agent is provided with a water outlet on one side of the bottom and a water inlet on the other side; a water tank is arranged between the water inlet and the water outlet; the water outlet is connected to a cleaning solvent configuration unit at one end and connected to the ultrasonic cleaning unit through a spray head at the other end; and the bottom of the water tank is connected to the water return unit through a water return hole; wherein: The cleaning intelligent agent uses ultrasonic waves to clean the surface stains of photovoltaic cells, and the cleaning intelligent agent comprises an image acquisition and processing unit, a cleaning solvent configuration unit, an ultrasonic cleaning unit and a water return unit; The first intelligent control module regulates the output power and frequency of the ultrasonic generator, adjusts the amplitude of the amplitude rod of the transducer, and adjusts the solvent ratio of the cleaning solvent configuration unit according to the stain type of the photovoltaic cell; wherein: The second intelligent control module performs frequency spectrum analysis on the backwash cleaning photovoltaic cell cleaning waveform to determine the cleaning effect; The third intelligent control module links the drone and the moving mechanism to realize intelligent agent path planning, positioning and obstacle avoidance.

[0010] The process of regulating the output power and frequency of the ultrasonic generator by the first intelligent control module according to the stain type of the photovoltaic cell, comprising: The image acquisition and processing unit obtains stain data information by scanning the surface image of the photovoltaic cell through an infrared light sensor; The first intelligent control module regulates the output matching cleaning frequency of the ultrasonic cleaning unit after classifying the stain data information; the ultrasonic cleaning unit comprises an ultrasonic generator and a transducer; wherein: The stain data information is divided into stain categories by extracting feature vectors through a convolutional neural network; The output power of the ultrasonic generator is regulated according to the stain category according to the following formula; that is: ; Wherein: sound intensity, ; Cleaning liquid density, ; The speed of sound in liquid ; Vibration angular frequency ; Vibration displacement of medium particles, ; The output frequency of the ultrasonic generator is regulated according to the stain category according to the following formula; that is:

[0011] in: cavitation bubble radius ; Fluid density, kg / m³; , hydrostatic pressure of liquid, Pa.

[0012] The process of extracting feature vectors from stain data using a convolutional neural network to classify stains into categories includes: The convolutional feature map of the stain data is obtained by calculating the feature map of the stain data using the following formula. ; ; Where: is the convolution kernel ( × ), This is a stain feature diagram. To accurately locate the x-coordinate, To accurately locate the vertical coordinate, This represents the horizontal coordinate offset. This represents the offset of the ordinate. For bias terms; Max pooling is applied to the convolutional feature map of the stain to obtain the stain feature vector:

[0013] in: The pooling kernel size, To accurately locate the x-coordinate, To accurately locate the vertical coordinate, This represents the horizontal coordinate offset. The vertical coordinate is shifted to retain the maximum value in the local area, thus enhancing the robustness of the feature.

[0014] The feature vector is mapped to the probability distribution of various stains using the following activation function: ; in: The first output of the fully connected layer The logit value is used to identify the stain category, where C is the number of stain categories. The category with the highest probability is taken as the identification result. The probability of each type of stain is calculated using the following cross-entropy loss to obtain the category of each stain: .

[0015] The process by which the first intelligent control module adjusts the amplitude of the transducer (202)'s amplitude rod according to the type of contamination in the photovoltaic cell includes: Adjust the output amplitude of the amplitude transformer in transducer (202) according to the following formula based on the type of stain; that is: ; in:

[0016] in: The force applied to the input end of the amplitude rod of the transducer (202) (the initial driving force provided by the transducer (202), which is positively correlated with the ultrasonic power) N; Input power to the transducer (202); The vibration velocity at the input end of the transducer (202); The force acting on the output end of the transducer (202) amplitude rod (acting directly on the cleaning fluid to generate impact kinetic energy), N; Vibration velocity at the output end of the amplitude transformer , , It is the angular frequency of vibration; Transmit matrix elements for the amplitude transformer; The cross-sectional area of ​​the front of the amplitude transformer is [area missing]. ; The cross-sectional area behind the amplitude transformer is... ; The characteristic coefficient of the amplitude transformer is dimensionless. , The length of the gear shift lever. The wavelength of the sound wave in the amplitude transformer; N is the stiffness coefficient of the amplitude transformer material. ; The gradient of the cross-sectional area at the output end of the amplitude transformer; The density of the cleaning fluid, ; The speed at which sound waves propagate in the cleaning fluid.

[0017] The transducer adjusts the amplitude of the amplitude transformer according to the solid stains, resulting in a stepped amplitude output. ; in: The characteristic impedance of the amplitude transformer characterizes the ability of the ultrasonic amplitude transformer material to transmit acoustic energy.

[0018] The transducer adjusts the amplitude of the output cone based on the soft stains, that is: ; in: The output end diameter of the conical amplitude transformer; The diameter of the input end of the conical amplitude transformer; The transducer adjusts the amplitude of the output cone-shaped variable rod according to the water quality pollution, that is: ; in: For the amplitude rod damping attenuation system (i.e., sound energy loss coefficient). This is the attenuation factor.

[0019] The transducer adjusts the amplitude of the catenary output from the amplitude transformer based on the oil stains, that is: ; in: The catenary amplitude transformer attenuation-focusing coefficient; This is the normalized catenary coefficient; The acoustic impedance of the cleaning fluid characterizes the resistance and propagation efficiency of the cleaning fluid to sound energy. The wavenumber of the ultrasonic wave in the cleaning fluid characterizes the propagation characteristics of ultrasonic vibration in the cleaning fluid and determines the intensity and distribution of the cavitation effect.

[0020] The cleaning solvent preparation unit includes an acidic solvent chamber, an alkaline solvent chamber, an organic solvent chamber, a first metering pump, a second metering pump, a third metering pump collector, a pressurizing pump, and a high-pressure nozzle. The input end of the high-pressure nozzle is connected to the first metering pump, the second metering pump, and the third metering pump in sequence through the pressurizing pump and the collector. The first metering pump is connected to the acidic solvent chamber; the second metering pump is connected to the alkaline solvent chamber; and the third metering pump is connected to the organic solvent chamber.

[0021] The water return unit includes a turbidity sensor, a stain concentration detector, and a filter device. The water return unit collects the turbidity and stain concentration data of the wastewater in the water tank in real time. If the turbidity of the return water is >5 NTU, the pressure of the water nozzle is increased until the return water index meets the standard.

[0022] Beneficial effects First, this invention utilizes image recognition technology to intelligently diagnose the type of stain (such as dust, oil, bird droppings, etc.) and dynamically adjusts ultrasonic parameters (power, frequency, amplitude) and cleaning agent formulation accordingly, achieving "tailored solutions based on stain type" and avoiding the blindness of traditional methods. Simultaneously, this invention employs a second intelligent control module to perform spectral analysis on the ultrasonic echo or vibration waveform during the cleaning process, judging the degree of stain removal in real time, forming a closed-loop control of "perception-execution-evaluation-optimization." This ensures thorough cleaning, prevents resource waste caused by over-cleaning and incomplete cleaning caused by under-cleaning, fundamentally guaranteeing cleaning quality and maximizing the restoration of photovoltaic panel light transmittance and power generation efficiency.

[0023] Secondly, this invention utilizes the physical effects of ultrasonic cavitation and micro-jet to remove stains, completely avoiding the risk of scratching the anti-reflective film and glass on the photovoltaic panel surface caused by physical contact methods such as brushing and scraping. Simultaneously, this invention can automatically adjust the amplitude and output power of the amplitude transformer according to the intensity of stain adhesion, ensuring cleaning power while effectively avoiding potential micro-damage to photovoltaic cell encapsulation materials (EVA film, backsheet, etc.) caused by excessive ultrasonic energy, thus helping to extend the overall service life of photovoltaic modules.

[0024] Third, this invention is integrated into an unmanned, autonomous dispatching and transportation mechanism, enabling the cleaning agent to autonomously plan its path and move and position itself flexibly within large-area, complexly arranged photovoltaic arrays. This completely solves the fundamental problem of limited coverage for fixed cleaning equipment, making it particularly suitable for large-scale ground-mounted power plants and distributed rooftop power plants. Furthermore, the cleaning process of this invention requires no human intervention, achieving fully automated operation from identification, movement, cleaning to effect verification, significantly reducing reliance on high-intensity, high-risk manual labor.

[0025] Fourth, the water return unit of this invention, through a closed-loop water circulation system of "water inlet-cleaning-recycling-filtration-reconfiguration" and precise spraying, greatly reduces the consumption of clean water for each cleaning, which is particularly advantageous in water-scarce areas. Combined with an intelligent solvent configuration unit, it can accurately mix cleaning solutions and use them as needed, reducing the overall amount and emissions of chemicals, making it more environmentally friendly.

[0026] Fifth, this invention can serve as an intelligent operation and maintenance node, enabling data linkage with the power plant's monitoring and power generation management systems. Cleaning records, stain distribution maps, and power generation improvement data can all provide data support for preventative maintenance, performance analysis, and optimized layout of the power plant, propelling photovoltaic power plant operation and maintenance into a new stage of digitalization and intelligence.

[0027] In summary, this invention deeply integrates advanced ultrasonic cleaning technology with artificial intelligence, automatic control, and robotics, creatively solving the core pain points of traditional photovoltaic panel cleaning methods in terms of efficiency, quality, cost, and applicability. It achieves a leap from "extensive manual cleaning" to "precision intelligent cleaning," demonstrating significant technological progress and broad prospects for industrial application. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the intelligent agent structure of an intelligent system for cleaning photovoltaic cells based on ultrasonic waves, according to the present invention.

[0029] Figure label: 101-Image acquisition and processing unit; 102-Cleaning solvent preparation unit; 103-Ultrasonic cleaning unit; 104-Return water unit; 105-Return water hole; 106-Moving mechanism; 107-Inlet; 108-Outlet; 109-Water tank; 110-Sprayer head; 201-Ultrasonic generator; 202-Transducer; Detailed Implementation

[0030] The following is in conjunction with the appendix Figure 1 The present invention will be described in detail as follows: This invention provides an intelligent system for ultrasonic cleaning of photovoltaic cells. The intelligent system includes a cleaning intelligent agent 100, a first intelligent control module, a second intelligent control module, a third intelligent control module, and a drone. The cleaning intelligent agent uses ultrasound to clean the surface stains of the photovoltaic cells. The cleaning intelligent agent 100 includes: an image acquisition and processing unit 101, a cleaning solvent preparation unit 102, an ultrasonic cleaning unit 103, a water return unit 104, and a moving mechanism 106. The ultrasonic cleaning unit 103 includes an ultrasonic generator 201 and a transducer 202. The cleaning intelligent agent 100 has a water outlet 108 on one side of its bottom and a water inlet on the other side. A water tank 109 is provided between the water inlet 107 and the water outlet 108; one end of the water outlet 108 is connected to the cleaning solvent preparation unit 102, and the other end is connected to the ultrasonic cleaning unit 103 via a nozzle 110; the bottom of the water tank 109 is connected to the return water unit 104 via a return water hole 105; the first intelligent control module adjusts the output power and frequency of the ultrasonic generator 201 and the amplitude of the transducer 202 according to the type of stains on the photovoltaic cell; and adjusts the solvent ratio of the cleaning solvent preparation unit 102; the second intelligent control module performs spectrum analysis based on the returned cleaning waveform of the photovoltaic cell to determine the cleaning effect. In this invention, when the intensity of the characteristic peak (frequency 2-5kHz) representing "stain adhesion" in the returned spectrum decreases to less than 10% of the initial value, the cleaning is considered complete.

[0031] The third intelligent control module remotely controls a drone to transport the intelligent agent to a designated location to complete the cleaning work. The water tank 109 has a U-shaped structure (suitable for common photovoltaic panel widths of 1.6m) and wraps around the edge of the photovoltaic cells to prevent cleaning fluid from splashing. The bottom of the intelligent agent is also connected to a cleaning brush and a moving mechanism 106; the cleaning brush is made of wear-resistant nylon bristles (0.5mm in diameter) and is installed at the rear end of the cleaning water tank 109, which can clean residual stains a second time after high-pressure water cleaning, and the brush speed is adjustable (50-100rpm); the moving mechanism 106 includes a tracked walking belt (walking speed 0.2-0.5m / s) and positioning wheels (deviation ≤1mm) to ensure that the intelligent agent moves smoothly along the surface of the photovoltaic panel and is adapted to different tilt angles (0°-30°) on the photovoltaic cells.

[0032] The hardware parameters involved in this invention are shown in the table below:

[0033] in: The image acquisition and processing unit 101 obtains stain data information by scanning the surface image of the photovoltaic cell with an infrared light sensor; The first intelligent control module classifies the stain data and then adjusts the ultrasonic cleaning unit 103 to output a matching cleaning frequency; the ultrasonic cleaning unit 103 includes an ultrasonic generator and a transducer 202; wherein: The process of extracting feature vectors from stain data using a convolutional neural network to classify stains into categories includes: The convolutional feature map of the stain data is obtained by calculating the feature map of the stain data using the following formula. ; ; Where: is the convolution kernel ( × ), This is a stain feature diagram. To accurately locate the x-coordinate, To accurately locate the vertical coordinate, This represents the horizontal coordinate offset. This represents the offset of the ordinate. For bias terms; Max pooling is applied to the convolutional feature map of the stain to obtain the stain feature vector:

[0034] in: The pooling kernel size, To accurately locate the x-coordinate, To accurately locate the vertical coordinate, This represents the horizontal coordinate offset. The vertical coordinate is shifted to retain the maximum value in the local area, thus enhancing the robustness of the feature.

[0035] The feature vector is mapped to the probability distribution of various stains using the following activation function: ; in: The first output of the fully connected layer The logit value is used to identify the stain category, where C is the number of stain categories. The category with the highest probability is taken as the identification result. The probability of each type of stain is calculated using the following cross-entropy loss to obtain the category of each stain: .

[0036] Adjust the output power of the ultrasonic generator according to the type of stain using the following formula: ; Among them: sound intensity, ; Cleaning fluid density, ; speed of sound waves in liquids ; Vibration angular frequency ; Vibrational displacement of medium particles ; Adjust the output frequency of the ultrasonic generator according to the type of stain using the following formula:

[0037] in: cavitation bubble radius ; Fluid density, kg / m³; The hydrostatic pressure of the liquid is Pa. The ultrasonic power and frequency are adjusted dynamically according to the type of stain. For stubborn stains such as bird droppings, a low frequency (20-30kHz) and high power (800-1000W) are used to enhance the cavitation impact force. For light stains such as oil stains, a high frequency (60-80kHz) and medium power (500-700W) are used to avoid damaging the photovoltaic panel. Adjust the output amplitude of the amplitude transformer in transducer 202 according to the type of stain using the following formula: ; in:

[0038] in: The force applied to the input end of the amplitude transformer of transducer 202 (the initial driving force provided by transducer 202, which is positively correlated with the ultrasonic power). N; Input power to transducer 202; The vibration velocity at the input terminal of transducer 202; The force applied to the output end of the amplitude transformer of transducer 202 (acting directly on the cleaning fluid to generate impact kinetic energy), N; The vibration velocity at the output end of the amplitude transformer (the core dynamic parameter of the output amplitude, vibration intensity). , , It is the angular frequency of vibration; Transmission matrix elements for the amplitude transformer (dimensionless parameters, reflecting the transmission coefficients of Li Yu's velocity). The cross-sectional area of ​​the amplitude transformer (output end cross-sectional area, which directly affects the impact pressure) is the front cross-sectional area of ​​the amplitude transformer. ; This is the cross-sectional area of ​​the amplitude transformer (the cross-sectional area of ​​the input end and the area of ​​the end connected to the transducer 202). ; The characteristic coefficient of the amplitude transformer (which determines whether the amplitude transformer resonates and affects energy transfer efficiency) is dimensionless. , The length of the gear shift lever. The wavelength of the sound wave in the amplitude transformer; N is the stiffness coefficient of the amplitude transformer material. ; The gradient of the cross-sectional area at the output end of the amplitude transformer; The density of the cleaning fluid, ; The speed at which sound waves propagate in the cleaning fluid.

[0039] The transducer 202 adjusts the amplitude of the output of the amplitude transformer according to the solid stains, that is: ; in: The frequency equation is: ; The transducer 202 adjusts the amplitude of the output cone-shaped variable amplitude rod according to the soft stain, that is: ; The transducer 202 adjusts the amplitude of the output cone-shaped variable rod according to the water quality stains, that is: ; The transducer 202 adjusts the amplitude of the catenary output from the amplitude transformer according to the oil stains, that is: .

[0040] The performance parameters of the transducer 202 of the present invention are shown in the table below:

[0041] The first intelligent control module of this invention regulates the output power and frequency of the ultrasonic generator 201 based on bird droppings on the photovoltaic cells: Bird droppings characteristics: contain uric acid crystals, solidify and clump together, with strong adhesion; Ultrasonic parameter regulation: Frequency: 20-30kHz (low frequency enhances cavitation impact force); Power: 800-1000W (high power improves peeling efficiency); Amplitude of the amplitude transformer: stepped; Solvent ratio: pH=4-5 acidic solution (95% deionized water + 3% organic acid + 2% nonionic surfactant); Cleaning process: Image recognition locates the stain → Intelligent agent moves to the target area → U-shaped water tank 109 wraps around the edge of the photovoltaic panel (to prevent splashing); Spray acidic cleaning solution and soak for 5 minutes → Start ultrasonic cleaning (lasts 3-5 minutes); Spectrum analysis confirms characteristic peak intensity ≤10% → High-pressure water rinsing → Return water filtration and recycling.

[0042] The first intelligent control module of this invention regulates the output power and frequency of the ultrasonic generator 201 based on the oil stains on the photovoltaic cells: Oil stain characteristics: mineral oil / vegetable oil, oily dispersion, easy to adhere; Ultrasonic parameter regulation: Frequency: 60-80kHz (high frequency for gentle cleaning, avoiding damage to the coating); Power: 500-700W (medium power for balancing dissolution and protection); Amplitude of the amplitude transformer: catenary solvent ratio: 30% organic solvent (67% deionized water + 30% mixed hydrocarbon organic solvent + 3% anionic surfactant); Cleaning process: immersion in low-pressure spray cleaning solution for 10 minutes → start ultrasonic cleaning (lasting 5-8 minutes); real-time monitoring of echo spectrum → after reaching the standard, repeated rinsing with deionized water → ensuring no solvent residue.

[0043] The first intelligent control module of the present invention controls the output power and frequency of the ultrasonic generator 201 according to the cement ash stains on the photovoltaic cell: Stain characteristics: hard thin layer composed of calcium carbonate and calcium silicate, with extremely strong adhesion; Ultrasonic parameter control: Frequency: 25-35kHz (medium and low frequency take into account impact force and coverage); (1) Power: 900-1000W (high power to break through hard surface); Amplitude of the variable rod: stepped; Solvent ratio: pH=9-10 alkaline solution (92% deionized water + 4% alkalinity adjuster + 2% chelating agent + 2% corrosion inhibitor); Cleaning process: Spray alkaline cleaning solution and soak for 15-20 minutes → Start ultrasonic cleaning (for 8-10 minutes); Use a plastic scraper to scrape off softened residue → High-pressure water rinsing → Second wipe with neutral cleaning agent → Light transmittance test (≥95% for new boards).

[0044] The first intelligent control module of this invention regulates the output power and frequency of the ultrasonic generator 201 based on the dust and dirt on the photovoltaic cells: characteristics of dust and dirt: a thin mud layer formed by the combination of dust and moisture, with weak adhesion; ultrasonic parameter regulation: frequency: 50-60kHz (medium frequency high-efficiency coverage); power: 400-500W (low power energy-saving cleaning); amplitude of the amplitude transformer: conical; solvent ratio: neutral cleaning solution (98% deionized water + 2% nonionic surfactant); high-pressure spray cleaning solution → ultrasonic cleaning for 3-5 minutes → recycled water after filtration → air dry or wipe dry with a soft cloth.

[0045] The first intelligent control module classifies stain data and then regulates the cleaning solvent preparation unit 102 to spray the corresponding cleaning agent. The cleaning solvent preparation unit 102 includes an acidic solvent chamber, an alkaline solvent chamber, an organic solvent chamber, a first metering pump, a second metering pump, a third metering pump collector, a pressurizing pump, and a high-pressure nozzle. The input end of the high-pressure nozzle is connected to the first metering pump, the second metering pump, and the third metering pump sequentially through the pressurizing pump and the collector. The first metering pump is connected to the acidic solvent chamber; the second metering pump is connected to the alkaline solvent chamber; and the third metering pump is connected to the organic solvent chamber. The cleaning solvent preparation unit 102 can adjust the pH value of the cleaning solution (range 3-11) and the ratio of mixed hydrocarbons and halogenated hydrocarbon organic solvents according to the stain type, making it suitable for different stains such as bird droppings (pH=4-5, acidic ratio), oil stains (30% organic solvent), and cement ash (pH=9-10, alkaline ratio). The return water unit 104 collects the sewage turbidity and stain concentration data of the water tank 109 in real time. If the return water turbidity is >5 NTU, the pressure of the water nozzle 110 is increased (increment by 0.2 MPa) until the return water index meets the standard.

[0046] Although the present invention has been described above, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many modifications under the guidance of the present invention without departing from the spirit of the present invention, and these modifications are all within the protection scope of the present invention.

Claims

1. An intelligent system for cleaning photovoltaic cells based on ultrasonic waves, characterized in that: The intelligent system includes a cleaning intelligent body with a moving mechanism (106), a first intelligent control module, a second intelligent control module, a third intelligent control module, and a drone; wherein: the cleaning intelligent body uses ultrasonic waves to clean the stains on the surface of the photovoltaic cell, the cleaning intelligent body includes an image acquisition and processing unit (101), a cleaning solvent preparation unit (102), an ultrasonic cleaning unit (103), and a water return unit (104); a water outlet (108) is provided on one side of the bottom of the cleaning intelligent body, and a water inlet (107) is provided on the other side; a water tank (109) is provided between the water inlet (107) and the water outlet (108); one end of the water outlet (108) is connected to the cleaning solvent preparation unit (102), and the other end is connected to the ultrasonic cleaning unit (103) through a nozzle (110); and the bottom of the water tank (109) is connected to the water return unit (104) through a water return hole (105); The first intelligent control module regulates the output power and frequency of the ultrasonic generator (201) and adjusts the amplitude of the transducer (202)'s amplitude rod according to the type of stains on the photovoltaic cells; it also adjusts the solvent ratio of the cleaning solvent preparation unit (102); wherein: The second intelligent control module performs spectrum analysis based on the returned photovoltaic cell cleaning waveform to determine the cleaning effect; The third intelligent control module links the drone and the mobile mechanism (106) to realize intelligent agent path planning, positioning and obstacle avoidance.

2. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 1, characterized in that: The first intelligent control module regulates the output power and frequency of the ultrasonic generator (201) according to the type of stains on the photovoltaic cells, including: The image acquisition and processing unit (101) obtains stain data information by scanning the surface image of the photovoltaic cell with an infrared light sensor; The first intelligent control module classifies the stain data and then adjusts the ultrasonic cleaning unit (103) to output a matching cleaning frequency; the ultrasonic cleaning unit (103) includes an ultrasonic generator and a transducer (202); wherein: The stain data information is categorized by extracting feature vectors using a convolutional neural network. Adjust the output power of the ultrasonic generator according to the type of stain using the following formula: ； in: Cleaning fluid density, ; The speed of sound waves in a liquid; Vibration angular frequency ; Vibrational displacement of medium particles ; Adjust the output frequency of the ultrasonic generator according to the type of stain using the following formula: ； in: cavitation bubble radius ; , hydrostatic pressure of liquid, Pa.

3. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 2, characterized in that: The process of extracting feature vectors from stain data using a convolutional neural network to classify stains into categories includes: The convolutional feature map of the stain data is obtained by calculating the feature map of the stain data using the following formula. ; ； Where: K is the convolution kernel. This is a stain feature diagram. To accurately locate the x-coordinate, The vertical coordinate is the position. This represents the horizontal coordinate offset. This represents the offset of the ordinate. For bias terms; Max pooling is applied to the convolutional feature map of the stain to obtain the stain feature vector: ； in: The pooling kernel size, To accurately locate the x-coordinate, The vertical coordinate is the position. This represents the horizontal coordinate offset. The vertical coordinate is shifted to retain the maximum value in the local area, thus enhancing the robustness of the feature. The feature vector is mapped to the probability distribution of various stains using the following activation function: ； in: The first output of the fully connected layer The logit value is used to identify the stain category, where C is the number of stain categories. The category with the highest probability is taken as the identification result. The probability of each type of stain is calculated using the following cross-entropy loss to obtain the category of each stain: 。 4. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 1, characterized in that: The process by which the first intelligent control module adjusts the amplitude of the transducer (202)'s amplitude rod according to the type of contamination in the photovoltaic cell includes: Adjust the output amplitude of the amplitude transformer in transducer (202) according to the following formula based on the type of stain; that is: ； in: ； in: Force applied to the input end of the amplitude transformer of transducer (202) N; Input power to the transducer (202); The vibration velocity at the input end of the transducer (202); The force applied at the output end of the amplitude transformer of transducer (202) is N; Vibration velocity at the output end of the amplitude transformer , , It is the angular frequency of vibration; Transmit matrix elements for the amplitude transformer; The cross-sectional area of ​​the front of the amplitude transformer is [area missing]. ; The cross-sectional area behind the amplitude transformer is... ; The characteristic coefficient of the amplitude transformer is dimensionless. , The length of the gear shift lever. The wavelength of the sound wave in the amplitude transformer; N is the stiffness coefficient of the amplitude transformer material. ; The gradient of the cross-sectional area at the output end of the amplitude transformer; The density of the cleaning solution, ; The speed at which sound waves propagate in the cleaning fluid.

5. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 3, characterized in that: The transducer (202) adjusts the amplitude of the output of the amplitude transformer according to the solid stains, that is: ； in: The characteristic impedance of the amplitude transformer is used to characterize the acoustic energy transmitted by the ultrasonic amplitude transformer material.

6. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 3, characterized in that: The transducer (202) adjusts the amplitude of the output cone-shaped variable rod according to the soft stain, that is: ； in: The output end diameter of the conical amplitude transformer; The input end diameter of the conical amplitude transformer.

7. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 3, characterized in that: The transducer (202) adjusts the amplitude of the output cone-shaped variable rod according to the water quality pollution, that is: ； in: For the amplitude rod damping attenuation system (i.e., sound energy loss coefficient). This is the attenuation factor.

8. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 3, characterized in that: The transducer (202) adjusts the amplitude of the catenary output by the amplitude transformer according to the oil stains, that is: ； in: The catenary amplitude transformer attenuation-focusing coefficient; This is the normalized catenary coefficient; The acoustic impedance of the cleaning fluid characterizes the resistance and propagation efficiency of the cleaning fluid to sound energy. The wavenumber of the ultrasonic wave in the cleaning fluid characterizes the propagation characteristics of ultrasonic vibration in the cleaning fluid and determines the intensity and distribution of the cavitation effect.

9. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 1, characterized in that: The cleaning solvent preparation unit (102) includes an acidic solvent chamber, an alkaline solvent chamber, an organic solvent chamber, a first metering pump, a second metering pump, a third metering pump collector, a pressurizing pump, and a high-pressure nozzle. The input end of the high-pressure nozzle is connected to the first metering pump, the second metering pump, and the third metering pump in sequence through the pressurizing pump and the collector. The first metering pump is connected to the acidic solvent chamber. The second metering pump is connected to the alkaline solvent chamber. The third metering pump is connected to the organic solvent chamber.

10. The intelligent system for ultrasonic cleaning of photovoltaic cells according to claim 1, characterized in that: The return water unit (104) includes a turbidity sensor, a stain concentration detector and a filter device; the return water unit (104) collects the sewage turbidity and stain concentration data of the water tank (109) in real time. If the return water turbidity is >5 NTU, the pressure of the water nozzle (110) is increased until the return water index meets the standard.

Images