A wireless automatic charging method and device for photovoltaic cleaning robots

Abstract

The present application provides a wireless automatic charging method and device for a photovoltaic cleaning robot, which belongs to the technical field of photovoltaic cleaning robots. The method includes determining a second cleaning pollution information set based on a first cleaning pollution information set, historical cleaning data, and a preset pollution prediction model; performing dynamic power consumption optimization based on the second cleaning pollution information set and a preset pollution cleaning intensity curve and a preset pollution power generation impact curve to update the cleaning strategy corresponding to the area to be cleaned based on the optimization result, and then controlling the photovoltaic cleaning robot to clean the photovoltaic panel until it reaches the cleaning end point of the corresponding photovoltaic panel for the first time, so as to determine whether to perform a second cleaning on the corresponding photovoltaic panel based on the remaining power. If it is determined that the photovoltaic panel is to be cleaned a second time, a cleaning return path is generated based on the residual pollution information set and the remaining power corresponding to the photovoltaic panel, so as to perform the second cleaning task along the cleaning return path and determine whether to automatically charge through the wireless charging device based on the remaining power.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic cleaning robots, and in particular to a wireless automatic charging method and device for photovoltaic cleaning robots. Background Art

[0002] Against the backdrop of today's booming photovoltaic industry, the cleaning and maintenance of photovoltaic panels plays a crucial role in ensuring power generation efficiency. Currently, photovoltaic panel cleaning methods primarily fall into two categories: manual cleaning and automated cleaning. Manual cleaning relies on manual labor, is inefficient, and labor-intensive, making it difficult to meet the cleaning needs of large-scale photovoltaic power plants.

[0003] In the automated cleaning sector, while the emergence of numerous photovoltaic cleaning robots has significantly improved cleaning speeds, they face a series of technical challenges that urgently need to be overcome. For one thing, most photovoltaic cleaning robots suffer from serious shortcomings in power management. They lack accurate and effective power monitoring and scientific management strategies. This directly prevents them from closely adapting to the actual needs of the cleaning task and dynamically and accurately planning cleaning steps based on power consumption. This often leads to chaotic and disorganized cleaning operations. This shortcoming is exacerbated by inclement weather and the continued impact of the rainy season, and the resulting negative impacts are even more severe. Furthermore, when dealing with the diverse types of contaminants on photovoltaic panels, such as bird droppings, dust, oil, and algae, robots lack a deep understanding of the correlation between different contaminant characteristics and power consumption, resulting in a single, rigid cleaning pattern. For example, using the same power and duration to treat sticky bird droppings and light dust wastes energy while failing to guarantee effective cleaning, further exacerbating the excessive energy consumption.

[0004] Meanwhile, focusing on charging technology, most existing charging solutions are overly simplistic and crude. Focusing solely on the simple act of recharging power, they completely ignore the complex real-world scenarios involved in cleaning tasks and the potential and significant impact of pollution on power consumption. For example, because the correlation between the level and type of pollution and power consumption is not taken into account, cleaning robots often encounter sudden power drops without warning during the cleaning process, or even stop working midway due to rapid power depletion. This not only affects the continuity of the cleaning task but also poses a potential threat to the normal operation of the photovoltaic panels. This series of problems, intertwined, creates numerous obstacles to the flexible, intelligent, and efficient wireless automatic charging and integrated cleaning operations of photovoltaic cleaning robots, becoming a key bottleneck restricting the further development of the industry.

[0005] Based on this, there is an urgent need for a more flexible, intelligent and efficient technical solution for wireless automatic charging of photovoltaic cleaning robots. Summary of the Invention

[0006] To solve the above problems, the embodiments of the present application provide a wireless automatic charging method and device for a photovoltaic cleaning robot.

[0007] In one aspect, an embodiment of the present application provides a wireless automatic charging method for a photovoltaic cleaning robot, the method comprising:

[0008] Determining a first set of cleaning contamination information based on an image recognition result of an image of an area to be cleaned within a preset range; wherein the first set of cleaning contamination information includes one or more contamination sequences; the contamination sequence includes at least contamination type, contamination frequency, and contamination degree;

[0009] Determining a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model; the second cleaning contamination information set corresponds to the remaining areas to be cleaned outside the preset range;

[0010] generating a dynamic power consumption curve for the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve;

[0011] The dynamic power consumption curve is input into a cleaning intensity optimization model. After updating a cleaning strategy corresponding to the area to be cleaned based on the optimization result, the photovoltaic cleaning robot is controlled to clean the photovoltaic panel until the cleaning end point of the corresponding photovoltaic panel is reached for the first time, and whether to perform a second cleaning of the corresponding photovoltaic panel is determined based on the remaining power. The same cleaning strategy corresponds to the same preset range.

[0012] If it is determined that the photovoltaic panel needs to be cleaned a second time, a cleaning return path is generated based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power, so that the photovoltaic cleaning robot returns to the initial position after performing the second cleaning task along the cleaning return path, and determines whether to automatically charge through the wireless charging device based on the remaining power.

[0013] In one implementation of the present application, the method is applied to a rail-mounted photovoltaic cleaning robot hung on the edge of a photovoltaic panel; before determining the first cleaning contamination information set based on an image recognition result of an image of an area to be cleaned within a preset range, the method further includes:

[0014] Determining whether the photovoltaic cleaning robot is at the initial position or the end edge of the previous preset range;

[0015] If yes, obtain an image of the photovoltaic panel surface in the driving direction captured by an image acquisition device; wherein the image acquisition device is set on the body of the photovoltaic cleaning robot;

[0016] Based on the pre-stored size information corresponding to the preset range, the photovoltaic panel surface image is cropped with the current position of the photovoltaic cleaning robot as the starting edge reference point to obtain the image of the area to be cleaned corresponding to the preset range; wherein the shape and size of the area of the image of the area to be cleaned are consistent with the preset range; and the area of the photovoltaic panel is composed of an integer multiple of the area corresponding to the preset range;

[0017] Inputting the image of the area to be cleaned into a pre-trained image recognition model to determine each contaminated grid area within the preset range based on the model output result; wherein the contaminated grid area is obtained by clustering contaminated grids of each contamination type; and one contaminated grid area includes multiple contaminated grids corresponding to the same contamination type;

[0018] The first cleaning contamination information set is determined based on each of the contaminated grid areas and the contamination feature extraction model.

[0019] In one implementation of the present application, determining the first cleaning pollution information set based on each of the pollution grid areas and the pollution feature extraction model specifically includes:

[0020] Determining pollution feature information corresponding to each of the pollution grids using the pollution feature extraction model; wherein the pollution feature information includes at least a color feature value, a texture feature value, a shape feature value, an area feature value, and a reflectivity feature value associated with the pollution degree;

[0021] Calculating the product sum of the pollution characteristic information of the pollution grid and the preset pollution characteristic weight to determine a corresponding pollution degree coefficient; the pollution degree coefficient is used to quantify the pollution degree;

[0022] Calculating an average value of the pollution degree coefficients corresponding to the same pollution grid area, and calculating a first product value of the average value and the pollution degree weight of the corresponding pollution type as a pollution degree assessment value;

[0023] Matching the pollution level assessment value with a plurality of gradient threshold intervals to determine that the pollution level corresponding to the corresponding matching gradient threshold interval is the pollution level of the pollution type, and adding the pollution level to the corresponding pollution sequence;

[0024] The number of occurrences of the contaminated grids of each contamination type is counted respectively, and the contamination occurrence frequency is determined according to the ratio of the number of occurrences to the total number of contaminated grids, and is added to the corresponding contamination sequence to obtain the first cleaning contamination information set.

[0025] In one implementation of the present application, determining a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model specifically includes:

[0026] Determining, based on the current position, whether the first cleaning contamination information set has a cleaned preset range;

[0027] If so, the corresponding cleaned photovoltaic panel surface image, the historical cleaning pollution information set, and the historical cleaning strategy are used as the historical cleaning data and sent to the cloud server to output a corresponding pollution degree identification deviation value based on the comparison result of the cleaned photovoltaic panel surface image with a number of corresponding preset cleaning residual pollution sample images; wherein the pollution degree identification deviation value is used to correct one or more preset parameters in the preset pollution prediction model;

[0028] After receiving the pollution degree identification deviation value from the cloud server, the preset pollution prediction model is adjusted, and the first cleaning pollution information set is input into the adjusted preset pollution prediction model to determine the second cleaning pollution information set based on the output result.

[0029] In one implementation of the present application, the method further includes:

[0030] If it is determined that the first cleaning contamination information set does not contain the cleaned preset range, sending the image of the area to be cleaned to the cloud server so that the cloud server matches it in a historical case database;

[0031] If the match is successful, the matched historical case is used as the historical cleansing data;

[0032] Otherwise, the preset default cleaning data is used as the historical cleaning data.

[0033] In one implementation of the present application, a dynamic power consumption curve of the photovoltaic cleaning robot is generated based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve, specifically including:

[0034] Determining, in chronological order, a contamination degree sequence corresponding to the second cleaning contamination information set within a preset cleaning cycle;

[0035] generating a second cleaning intensity curve corresponding to the second cleaning contamination information set based on the contamination degree sequence and the preset contamination cleaning intensity curve; wherein the abscissa of the preset contamination cleaning intensity curve represents the contamination degree and the ordinate represents the cleaning intensity value; the cleaning intensity value at least corresponds to the cleaning power;

[0036] generating a second power generation efficiency curve corresponding to the second cleaning pollution information set based on the second cleaning intensity curve and the preset pollution power generation impact curve; wherein the abscissa of the preset pollution power generation impact curve represents cleanliness, and the ordinate represents power generation efficiency; the cleanliness is obtained based on the cleaning intensity value, the pollution type, the pollution degree, and a preset cleaning cleanliness comparison table; and the cleanliness and the power generation efficiency are positively correlated;

[0037] The second cleaning intensity curve and the second power generation efficiency curve are input into the following formula to generate the dynamic power consumption curve:

[0038]

[0039] Among them, E represents the power consumption of the dynamic power consumption curve within the cleaning cycle t; I(t) represents the cleaning intensity value at time t; C1 is the power consumption per unit time of cleaning; G(t) represents the power generation efficiency value at time t; C2 is the power consumption per unit time in the standby state.

[0040] In one implementation of the present application, after inputting the dynamic power consumption curve into the cleaning intensity optimization model, the method further includes:

[0041] A multi-objective optimization algorithm is established with the objective function of minimizing energy consumption and maximizing cleaning efficiency corresponding to the dynamic power consumption curve and the constraint condition that the cleaning intensity is greater than the cleaning intensity threshold;

[0042] The formula for maximizing cleaning efficiency is:

[0043]

[0044] Where η represents the cleaning efficiency; P(t) represents the actual cleaning power at time t; P max Indicates the maximum cleaning power of the photovoltaic cleaning robot;

[0045] Based on a preset genetic algorithm, the multi-objective optimization algorithm is solved to obtain an optimized cleaning intensity curve, and the optimized cleaning intensity curve is used as the optimization result to update the cleaning strategy; the cleaning strategy at least includes the cleaning intensity for photovoltaic pollution of different pollution types and different pollution degrees.

[0046] In one implementation of the present application, a cleaning return path is generated based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power, specifically including:

[0047] determining the residual contamination information set based on the cleaned area image corresponding to the cleaned photovoltaic panel;

[0048] Generate a secondary cleaning priority sequence according to the pollution degree corresponding to each residual pollution in the residual pollution information set and in descending order of pollution degree;

[0049] Determining corresponding surplus power based on the remaining power and the planned power corresponding to the remaining cleaning task amount;

[0050] Determining corresponding secondary cleaning residual contamination and corresponding residual contamination locations according to the surplus power, the secondary cleaning priority sequence, and the thorough cleaning intensity corresponding to each residual contamination;

[0051] According to the distance between the remaining contamination location and the cleaning end point, the remaining contamination location is sequentially added to the return cleaning node of the cleaning return path in order from near to far.

[0052] In one implementation of the present application, determining whether to perform a secondary cleaning on the corresponding photovoltaic panel based on the remaining power specifically includes:

[0053] Calculating the difference between the remaining power and the planned power corresponding to the remaining cleaning task amount as the surplus power; wherein the planned power at least includes the power required for the photovoltaic cleaning robot to return to its initial position;

[0054] Determining whether the residual contamination information set contains residual contamination with a contamination degree greater than a preset secondary cleaning threshold;

[0055] If yes, and the surplus power meets the intensity of thorough cleaning of the corresponding residual contamination, it is determined that the corresponding photovoltaic panel should be cleaned twice; wherein the thorough cleaning intensity at least includes the cleaning power required to completely clean the residual contamination;

[0056] Otherwise, the photovoltaic cleaning robot is controlled to return to the initial position to clean the next photovoltaic panel with the remaining cleaning task amount or perform wireless automatic charging.

[0057] On the other hand, an embodiment of the present application further provides a wireless automatic charging device for a photovoltaic cleaning robot, the device comprising:

[0058] A first determination module is configured to determine a first set of cleaning contamination information based on an image recognition result of an image of an area to be cleaned within a preset range; wherein the first set of cleaning contamination information includes one or more contamination sequences; and the contamination sequence includes at least contamination type, contamination frequency, and contamination degree.

[0059] a second determining module, configured to determine a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model; wherein the second cleaning contamination information set corresponds to the remaining to-be-cleaned area outside the preset range;

[0060] A first generating module is configured to generate a dynamic power consumption curve of the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve;

[0061] an input module, configured to input the dynamic power consumption curve into a cleaning intensity optimization model, and after updating a cleaning strategy corresponding to the area to be cleaned based on the optimization result, control the photovoltaic cleaning robot to clean the photovoltaic panel until the cleaning endpoint of the corresponding photovoltaic panel is reached for the first time, and determine whether to perform a second cleaning of the corresponding photovoltaic panel based on the remaining power; wherein the same cleaning strategy corresponds to the same preset range;

[0062] The second generation module is used to generate a cleaning return path based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power if it is determined that the photovoltaic panel needs to be cleaned for the second time, so as to return the photovoltaic cleaning robot to its initial position after performing the second cleaning task along the cleaning return path, and determine whether to automatically charge through a wireless charging device based on the remaining power.

[0063] Compared with the prior art, this application has the following significant effects:

[0064] Through the above technical solution, this application uses image recognition to determine the first set of cleaning pollution information, combines historical data with prediction models, and clarifies the pollution of the remaining area to be cleaned, providing a basis for accurate cleaning. At the same time, a dynamic power consumption curve is generated based on the pollution information and related curves, and input into the optimization model to achieve dynamic matching of cleaning intensity and power consumption, thereby improving power utilization efficiency. After the first cleaning, it can also determine whether to perform a second cleaning based on the remaining power and residual pollution information, and reasonably plan the return path to avoid ineffective work. After completing the task, the photovoltaic cleaning robot automatically decides whether to charge wirelessly based on the remaining power, realizing an intelligent closed loop of cleaning and charging.

[0065] The above solution significantly improves the intelligence and automation level of cleaning and charging of photovoltaic cleaning robots. Through intelligent, automated and refined management methods, it effectively solves the current technical problems of insufficient power management level of photovoltaic panel cleaning robots, simple and extensive charging solutions, and difficulty in achieving flexible, intelligent and efficient wireless automatic charging. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0067] Figure 1 A schematic flow chart of a wireless automatic charging method for a photovoltaic cleaning robot according to an embodiment of the present application;

[0068] Figure 2 This is a structural diagram of a photovoltaic cleaning system corresponding to a wireless automatic charging method for a photovoltaic cleaning robot in an embodiment of the present application;

[0069] Figure 3 This is a structural schematic diagram of a wireless automatic charging device for a photovoltaic cleaning robot in an embodiment of the present application. DETAILED DESCRIPTION

[0070] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0071] The embodiments of the present application provide a wireless automatic charging method and device for a photovoltaic cleaning robot, which is used to solve the current technical problems of insufficient power management for photovoltaic panel cleaning robots, simple and extensive charging solutions, and difficulty in achieving flexible, intelligent and efficient wireless automatic charging.

[0072] The wireless automatic charging technology solution for photovoltaic cleaning robots provided in this application allows the robot to clean photovoltaic panels without being affected by weather. For example, in the rainy season, the photovoltaic cleaning robot can always maintain a full charge by relying on wireless automatic charging. It can perform cleaning tasks multiple times a day, increase the frequency of cleaning photovoltaic panels, and ensure that the photovoltaic panels have the shortest contamination period, thereby improving the efficiency of photovoltaic power generation.

[0073] The following describes in detail various embodiments of the present application with reference to the accompanying drawings.

[0074] The embodiment of the present application provides a wireless automatic charging method for a photovoltaic cleaning robot, such as Figure 1 As shown, the method may include steps S101-S105:

[0075] S101 : determining a first cleaning contamination information set based on an image recognition result of an image of an area to be cleaned within a preset range.

[0076] The first cleaning contamination information set includes one or more contamination sequences, which at least include contamination type, contamination frequency, and contamination degree.

[0077] The executor of the wireless automatic charging method for a photovoltaic cleaning robot can be the microcontroller unit (MCU) within the photovoltaic cleaning robot, or an electronic device connected to the photovoltaic cleaning robot's single-chip microcomputer, including a wired or wireless connection. Electronic devices include, but are not limited to, handheld devices, computing devices, and servers with wireless communication capabilities. This application describes embodiments using the MCU as the executor for illustrative purposes only, and the executor is not limited to the MCU.

[0078] In an embodiment of the present application, the above method is applied to a rail-mounted photovoltaic cleaning robot hung on the edge of a photovoltaic panel. The above method is applied to photovoltaic power generation scenarios, especially large-area photovoltaic power stations.

[0079] Before determining the first cleaning contamination information set based on the image recognition result of the image of the area to be cleaned within the preset range, the method further includes:

[0080] Determine whether the photovoltaic cleaning robot is at its initial position or at the end edge of the previous preset range. If the photovoltaic cleaning robot is determined to be at its initial position or at the end edge of the previous preset range, obtain an image of the photovoltaic panel surface in the direction of travel captured by an image acquisition device. The image acquisition device is mounted on the body of the photovoltaic cleaning robot. Based on pre-stored dimension information corresponding to the preset range, the photovoltaic panel surface image is cropped with the photovoltaic cleaning robot's current position as the starting edge reference point to obtain an image of the area to be cleaned corresponding to the preset range. The shape and size of the image of the area to be cleaned are consistent with the preset range. The area of the photovoltaic panel is composed of an integer multiple of the area corresponding to the preset range. The image of the area to be cleaned is input into a pre-trained image recognition model to determine each contaminated grid area within the preset range based on the model output. The contaminated grid area is obtained by clustering contaminated grids of each pollution type. A contaminated grid area includes multiple contaminated grids corresponding to the same pollution type. Based on each contaminated grid area and the pollution feature extraction model, a first cleaning pollution information set is determined.

[0081] In other words, the photovoltaic cleaning robot is equipped with a position sensing module that can determine the current driving position in real time. The initial position can be understood as the robot is still at the edge of the photovoltaic panel and has not entered the track that contacts the photovoltaic panel. The previous preset range can be understood as the preset range in which the photovoltaic cleaning robot has cleaned the same photovoltaic panel. The end edge of the previous preset range is the boundary line between the cleaned preset range and another uncleaned preset range. Figure 2A photovoltaic cleaning system is shown, in which the photovoltaic cleaning robot is 201, the first preset range is 202, the end edge of which is 203, the second preset range is 204, the end edge of the second preset range is 205, the initial position is 206, the driving track of the photovoltaic cleaning robot is 207, and the cleaning end point is 208.

[0082] If it is determined that the photovoltaic cleaning robot is at the initial position or the end edge of the previous preset range at this time, and is not at the end point of the photovoltaic panel cleaning, and there is surplus power to perform subsequent cleaning tasks, then the image acquisition device is controlled to capture the surface image of the photovoltaic panel in the direction of travel. The image acquisition device is arranged on the body of the photovoltaic cleaning robot. The image acquisition device can be provided with multiple cameras, such as two, which are respectively arranged in front and behind the body, and can capture the surface image of the photovoltaic panel in the direction of travel and away from the direction of travel. The MCU of the present application also pre-stores the size information of the preset range, and the preset range can evenly divide the area of the photovoltaic panel. Among them, the photovoltaic panel surface image contains at least one area corresponding to the preset range, that is, the actual photovoltaic panel surface size corresponding to the photovoltaic panel surface image captured by the image acquisition device is greater than or equal to the size corresponding to the preset range. The MCU will use the current position as the starting edge reference point, such as Figure 2 The middle terminating edge is the starting edge reference point of the cropping operation, and the image of the area to be cleaned corresponding to the size information of the preset range is cropped. Subsequently, the image recognition model obtained by training a number of polluted grid area sample images and pollution type labels is used to divide the image of the area to be cleaned into an image composed of polluted grids, and each polluted grid carries a label of the pollution type; the image recognition model can be a convolutional neural network model. The present application can further divide the image of the area to be cleaned into images containing one or more polluted grid areas, and the polluted grid areas can be obtained by clustering the polluted grids. Clustering can use a density-based spatial clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) to cluster the polluted grids of each pollution type, thereby obtaining the polluted grid area.

[0083] For example, when clustering, each contaminated grid is treated as a data point, and spatially adjacent contaminated grids corresponding to the same contamination type are clustered together. A contaminated grid area contains several contaminated grids corresponding to the same contamination type.

[0084] Through the above processing, the present application can obtain accurate and finely divided pollution grid areas, which can be used to accurately plan the cleaning strategy for pollution on the one hand, and on the other hand, cluster the refined pollution grids into pollution grid areas, and plan cleaning according to the area to improve the cleaning efficiency.

[0085] Furthermore, the MCU will further generate a first cleaning pollution information set through each pollution grid area and the pollution feature extraction model.

[0086] Specifically, based on each contaminated grid area and the contamination feature extraction model, a first cleaning contamination information set is determined, specifically including:

[0087] The pollution feature extraction model is used to determine the pollution feature information corresponding to each polluted grid. The pollution feature information includes at least color feature values, texture feature values, shape feature values, area feature values, and reflectivity feature values associated with the pollution level. The product and value of the pollution feature information of the polluted grid and the preset pollution feature weight are calculated to determine the corresponding pollution level coefficient. The pollution level coefficient is used to quantify the pollution level. The average value of each pollution level coefficient corresponding to the same polluted grid area is calculated, and the first product value of the average value and the pollution level weight of the corresponding pollution type is calculated as the pollution level assessment value. The pollution level assessment value is matched with multiple gradient threshold intervals to determine the pollution level corresponding to the corresponding matching gradient threshold interval as the pollution level of the pollution type, and added to the corresponding pollution sequence. The number of occurrences of polluted grids of each pollution type is counted separately to determine the pollution occurrence frequency based on the ratio of the number of occurrences to the total number of polluted grids, and added to the corresponding pollution sequence to obtain the first cleaned pollution information set.

[0088] Among them, the pollution feature extraction model can be a pre-trained neural network model, which is used to extract the following feature dimensions corresponding to each pollution grid: color features, texture features, shape features, area features and reflectivity features. Specifically, color features, such as the obvious color difference between bird droppings and dust, can be used to intuitively reflect the characteristics of pollution; the texture of pollution can reflect its physical properties. The texture of oil pollution may appear smooth and flowing, while the texture of algae pollution is rough and granular. The more complex and special the texture, the more likely it is to indicate serious pollution; the shape of pollution can also reflect its formation mechanism and the degree of harm. Pollution with regular shapes and clear edges may be a special pollutant with a small local area. Compared with pollution with irregular shapes and large-scale diffusion, the severity is different; the size of the pollution area is directly related to the scope of pollution impact. Large-scale pollution covers photovoltaic panels, reducing the area receiving light and having a great impact on power generation efficiency; the normal reflectivity of photovoltaic panels is relatively stable, pollution will change its reflectivity characteristics, and coverage by pollutants will reduce the reflectivity. The more serious the pollution, the greater the change in reflectivity.

[0089] The above-mentioned characteristic values corresponding to different characteristic dimensions can be pre-scored by experts, and different characteristic values can correspond to different levels of correlation with pollution levels. The pollution degree coefficient formula is as follows:

[0090]

[0091] Among them, Pll(i,j) represents the pollution degree coefficient of the pollution grid corresponding to the coordinate (i,j) in the preset range, w k is the pollution feature weight of the kth feature dimension, N is the number of feature dimensions, for example, N is equal to 5, f k (i, j) represents the eigenvalue of the kth feature dimension of the pollution grid.

[0092] For example, the pollution feature information includes the following eigenvalues [a1, a2, a3, a4, a5], where a1 is the color eigenvalue, a2 is the texture eigenvalue, a3 is the shape eigenvalue, a4 is the area eigenvalue, and a5 is the reflectivity eigenvalue. Taking the pollution type of bird droppings as an example, the preset pollution feature weights corresponding to the bird droppings pollution type are [w1, w2, w3, w4, w5], and the pollution degree coefficient is w 1* a1+w 2* a2+w 3* a3+w 4* a4+w 5* This application integrates multiple parameter features to quantify the pollution level in order to regulate the robot's output power and effectively plan the robot's power consumption.

[0093] Subsequently, the present application also calculates the average value of each pollution degree coefficient in the same pollution grid area, and at the same time calculates the first product value of the calculated average value and the pollution degree weight of the corresponding pollution type to obtain the pollution degree assessment value of the pollution grid area. The pollution degree weight is used to unify the dimensions of the pollution degrees of various pollution types. Different pollution types can correspond to different pollution degree weights, which can be set by the user. This application does not make specific restrictions on this. Furthermore, the MCU can match the pollution degree assessment value with multiple gradient threshold intervals corresponding to different pollution degrees. The pollution degree classification standards such as mild, moderate, and severe are only illustrative and are not specifically limited here. According to the gradient threshold interval in which each pollution degree assessment value is located, that is, matching the gradient threshold interval, the pollution degree corresponding to each pollution grid area can be obtained and then added to the pollution sequence. The specific value of the gradient threshold interval can be set by the user during actual use, and this application does not make specific restrictions on this.

[0094] The MCU then counts the number of contaminated grids of each contamination type within a preset range and calculates the ratio of the number of contaminated grids to the total number of contaminated grids within the preset range, thereby obtaining the frequency of contamination of different types within the preset range. This, together with the aforementioned contamination levels, forms the first set of cleaning contamination information.

[0095] In one embodiment of the present application, if the preset range is always a fixed size, it may also cause unreasonable cleaning. For example, one side of the preset range is heavily contaminated, while the other side is less contaminated. If a unified cleaning strategy is provided according to the preset range, it is obviously not reasonable. Therefore, the present application also provides the following embodiments, including:

[0096] Calculate the second product value of each pollution degree coefficient corresponding to the same pollution type and the pollution degree weight of the corresponding pollution type. Compare each second product value with the preset range division threshold to determine whether the polluted grids corresponding to the second product values greater than the preset range division threshold are concentrated on one side of the preset range. If they are concentrated on one side of the preset range, the proportion of polluted grids corresponding to the second product values greater than the preset range division threshold is greater than a predetermined value. One side of the preset range is the side after the preset range is divided into two parts perpendicular to the travel direction of the photovoltaic cleaning robot. If so, divide the original preset range into two parts according to the aforementioned one-half ratio.

[0097] That is to say, according to the pollution type, each pollution degree coefficient is multiplied by the pollution degree weight of the corresponding pollution type to obtain the second product values corresponding to each pollution grid. Subsequently, the second product value is compared with the preset range division threshold. The preset range division threshold can be set by the user and is not specifically limited here. Then, the distribution of pollution grids with second product values greater than the preset range division threshold is counted to determine whether this part of the pollution grids is concentrated on one side of the preset range. If so, the preset range is further divided according to the principle of dividing the preset range in half. Otherwise, the preset range is not adjusted. Figure 2 In the middle of a preset range, a straight line parallel to the end edge is used to divide the preset range equally, that is, the preset range is divided into two halves. This allows for a more accurate image of the area to be cleaned, allowing for further execution of subsequent steps.

[0098] After obtaining the first cleaning pollution information set, in order to predict the pollution conditions of the remaining areas to be cleaned outside the preset range, the present application will then combine historical cleaning data and a preset pollution prediction model to determine a second cleaning pollution information set.

[0099] S102 : Determine a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model.

[0100] The second cleaning contamination information set corresponds to the remaining area to be cleaned outside the preset range.

[0101] In an embodiment of the present application, based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model, determining the second cleaning contamination information set specifically includes:

[0102] Determine whether the first cleaning pollution information set exists within a preset cleaned range based on the current location. If it is determined that the first cleaning pollution information set exists within the preset cleaned range, the corresponding cleaned photovoltaic panel surface image, historical cleaning pollution information set, and historical cleaning strategy are used as historical cleaning data and sent to the cloud server. Based on the comparison results of the cleaned photovoltaic panel surface image and several corresponding preset cleaning residual pollution sample images, a corresponding pollution level identification deviation value is output. The pollution level identification deviation value is used to correct one or more preset parameters in the preset pollution prediction model. After receiving the pollution level identification deviation value from the cloud server, the preset pollution prediction model is adjusted, and the first cleaning pollution information set is input into the adjusted preset pollution prediction model to determine the second cleaning pollution information set based on the output result.

[0103] Specifically, the preset cleaning range can be understood as the same row of photovoltaic panels. Figure 2 The first preset range is the cleaned range, and the MCU can determine this based on the current position of the PV cleaning robot. At this point, the MCU controls the image acquisition device to capture surface images of the first preset range again, obtaining cleaned PV panel surface images. The first set of cleaning contamination information corresponding to the first preset range is used as the historical cleaning contamination information set, and the cleaning strategy for the first preset range is used as the historical cleaning strategy to obtain historical cleaning data. This historical cleaning data is then sent to the cloud server, which performs the computationally intensive operations.

[0104] The cloud server can pre-store a number of preset cleaning legacy pollution sample maps, which have a corresponding relationship with the corresponding cleaning pollution information sample set and cleaning strategy sample. First, the cloud server determines the similar sample map that is closest to the surface image of the cleaned photovoltaic panel (such as the largest cosine similarity); then, the historical cleaning pollution information set, the historical cleaning strategy and the cleaning pollution information sample set and the cleaning strategy sample corresponding to the similar sample map are compared respectively, which can be obtained by calculating the Euclidean distance or the cosine distance. For example, if the Euclidean distance is greater than the preset deviation threshold, it means that there is a pollution degree identification deviation. The pollution degree identification deviation value is obtained by matching the calculated Euclidean distance with the preset deviation value comparison table.

[0105] The preset deviation threshold and the preset deviation comparison table can be set by the user according to actual use, and this application does not impose specific restrictions on this. The MCU can correct the preset parameters in the preset pollution prediction model by using the preset pollution level identification deviation value, wherein the correspondence between the preset pollution level identification deviation value and the preset parameters can be provided by an expert. The preset pollution level identification deviation value can be used to adjust the preset parameters, such as increasing or decreasing the parameters. Then, using the adjusted preset pollution prediction model and the first cleaning pollution information set, the second cleaning pollution information set is further generated.

[0106] The above-mentioned preset pollution prediction model can be a pre-trained machine learning model, which can generate a second cleaning pollution information set of the photovoltaic panel that needs to be cleaned next based on the first cleaning pollution information set.

[0107] In another embodiment of the present application, the parameter adjustment step S102 can be performed once after a preset time period, rather than after a preset range of cleaning, thereby reducing the frequency of interaction between the MCU and the cloud server. The preset time period can be set by the user, such as one day or one week, and is not specifically limited in this application.

[0108] Through the above solution, the preset pollution prediction model can be updated in a timely manner according to the actual cleaning situation, thereby improving the accuracy of generating the cleaning pollution information set.

[0109] Furthermore, if it is determined that the first set of cleaning contamination information does not contain a pre-set cleaned area, the image of the area to be cleaned is sent to the cloud server for matching against the historical case database. If a match is successful, the matched historical case is used as the historical cleaning data. Otherwise, the preset default cleaning data is used as the historical cleaning data.

[0110] In other words, the MCU determines that the PV cleaning robot has not yet cleaned the preset area, typically at its initial position. The MCU then sends the image of the area to be cleaned to the cloud server. The cloud server then compares the image of the area to be cleaned to a historical case database containing historical cleaning data samples corresponding to the image of the area to be cleaned. The cloud server then uses the matched historical cases to determine the second cleaning contamination information set. If a match fails, the cloud server uses the user-defined default cleaning data as the historical cleaning data.

[0111] Through the above scheme, combined with historical cleaning data, the first cleaning pollution information set that has been collected can be accurately predicted to accurately predict the second cleaning pollution information set of subsequent photovoltaic panels, so as to reasonably plan the robot power and perform charging management.

[0112] S103 , generating a dynamic power consumption curve of the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve.

[0113] In the embodiment of the present application, based on the second cleaning pollution information set and the preset pollution cleaning intensity curve and the preset pollution power generation impact curve, a dynamic power consumption curve of the photovoltaic cleaning robot is generated, specifically including:

[0114] Determine, in chronological order, the pollution level sequence corresponding to the second cleaning pollution information set within a preset cleaning cycle. Based on the pollution level sequence and a preset pollution cleaning intensity curve, generate a second cleaning intensity curve corresponding to the second cleaning pollution information set. The horizontal axis of the preset pollution cleaning intensity curve represents the pollution level, and the vertical axis represents the cleaning intensity value. The cleaning intensity value corresponds to at least the cleaning power. Based on the second cleaning intensity curve and a preset pollution power generation impact curve, generate a second power generation efficiency curve corresponding to the second cleaning pollution information set. The horizontal axis of the preset pollution power generation impact curve represents the cleanliness level, and the vertical axis represents the power generation efficiency value. The cleanliness level is determined based on the cleaning intensity value, pollution type, pollution level, and a preset cleaning cleanliness comparison table. There is a positive correlation between the cleanliness level and the power generation efficiency value.

[0115] It should be noted that each cleaning contamination information set corresponds to a preset range, and a cleaning contamination information set can contain multiple contamination sequences, thus also containing multiple contamination level values. This application can calculate the average of these multiple contamination levels for the same cleaning contamination information set to obtain the contamination level corresponding to the preset range. Furthermore, each preset range corresponds to the location of a photovoltaic panel, and the preset ranges have a chronological order before and after cleaning. This application can generate a contamination level sequence for all preset ranges contained in the second cleaning contamination information set in chronological order.

[0116] Subsequently, a pre-set pollution cleaning intensity curve is used to match the cleaning intensity value sequence corresponding to the pollution degree sequence, thereby generating a second cleaning intensity curve in chronological order. Next, the cleaning intensity values in the second cleaning intensity curve, along with the pollution types and pollution intensities within the preset range corresponding to the cleaning intensity values, are matched against a pre-set cleaning cleanliness comparison table. This table contains a mapping relationship between different pollution types, preset ranges of different pollution intensities, and cleaning intensity values during cleaning, and the cleanliness after cleaning. The power generation efficiency values corresponding to each preset cleanliness range are searched from the pre-set pollution power generation impact curve, and the second power generation efficiency curve is generated in the aforementioned chronological order.

[0117] Furthermore, the second cleaning intensity curve and the second power generation efficiency curve are input into the following formula to generate a dynamic power consumption curve:

[0118]

[0119] Where E represents the power consumption of the dynamic power consumption curve within the cleaning cycle T. I(t) represents the cleaning intensity value at time t. C1 represents the power consumption per unit time of cleaning. G(t) represents the power generation efficiency value at time t. C2 represents the power consumption per unit time of standby mode.

[0120] Through the above solution, the two influence curves, namely the second cleaning intensity curve and the second power generation efficiency curve, can be combined to more reasonably generate a dynamic power consumption curve.

[0121] S104, input the dynamic power consumption curve into the cleaning intensity optimization model, and after updating the cleaning strategy corresponding to the area to be cleaned based on the optimization result, control the photovoltaic cleaning robot to clean the photovoltaic panel until it reaches the cleaning end point of the corresponding photovoltaic panel for the first time, so as to determine whether to perform a second cleaning on the corresponding photovoltaic panel based on the remaining power.

[0122] The same cleaning strategy corresponds to the same preset range.

[0123] In the embodiment of the present application, after inputting the dynamic power consumption curve into the cleaning intensity optimization model, the method further includes:

[0124] A multi-objective optimization algorithm is established with the objective function of minimizing energy consumption and maximizing cleaning efficiency corresponding to the dynamic power consumption curve, and the constraint condition that the cleaning intensity is greater than the cleaning intensity threshold. The formula corresponding to maximizing cleaning efficiency is:

[0125]

[0126] Where η represents the cleaning efficiency. P(t) represents the actual cleaning power at time t. max Indicates the maximum cleaning power of the photovoltaic cleaning robot.

[0127] The multi-objective optimization algorithm is as follows:

[0128]

[0129] Based on a pre-set genetic algorithm, a multi-objective optimization algorithm is solved to obtain an optimized cleaning intensity curve. The optimized cleaning intensity curve is used as the optimization result to update the cleaning strategy. The cleaning strategy at least includes cleaning intensities for different pollution types and pollution levels of photovoltaic pollution.

[0130] In other words, the present application can implement the solution of the above-mentioned multi-objective optimization algorithm through iterative operations of a preset genetic algorithm, or can also implement the solution through other algorithms, which are not specifically limited here. Using the preset genetic algorithm of the above example, an optimized cleaning intensity curve can be solved, and the optimized cleaning intensity curve corresponds to the cleaning power used in different preset ranges, where a uniform cleaning power is used in the same preset range.

[0131] For example, for lightly contaminated photovoltaic panels, a lower cleaning intensity can be used to save energy. For heavily contaminated photovoltaic panels, a higher cleaning intensity is required to ensure cleaning efficiency. For different types of contaminants (such as dust and bird droppings), the cleaning intensity can be further adjusted based on their characteristics and impact on photovoltaic panel performance.

[0132] This application can achieve this by, for example, predicting that a spot of bird droppings requires 20% power to clean, but may not be able to complete subsequent tasks after cleaning. This application can pre-determine the power consumption required for subsequent cleaning tasks, thereby flexibly adjusting the intensity of the bird droppings cleaning. If the required intensity makes it impossible to complete the subsequent tasks, the bird droppings cleaning power is adjusted accordingly. In addition, after the subsequent tasks are completed, the remaining power and generated power are used to determine whether to proceed with the bird droppings treatment.

[0133] The pollution targeted by this application includes but is not limited to bird droppings, algae, leaves and debris, metal oxides, as well as pollutants containing chemical substances such as acid rain, industrial waste gas, and plaques formed by various microorganisms on the surface of photovoltaic panels.

[0134] After obtaining the cleaning strategy as described above, the photovoltaic cleaning robot can be controlled to clean the area to be cleaned in the first cleaning pollution information set. At the same time, the above steps S101-S104 will be executed in real time to update the cleaning strategy in time until the photovoltaic panels in the same row are cleaned and the return conditions are met.

[0135] In one embodiment of the present application, the above-mentioned determination of whether to perform secondary cleaning on the corresponding photovoltaic panel based on the remaining power specifically includes:

[0136] Calculate the difference between the remaining power and the planned power corresponding to the remaining amount of cleaning tasks as the surplus power. The planned power at least includes the power required for the photovoltaic cleaning robot to return to the initial position. Determine whether the residual pollution information set contains residual pollution with a pollution degree greater than the preset secondary cleaning threshold. In the case where it is determined that the residual pollution information set contains residual pollution with a pollution degree greater than the preset secondary cleaning threshold, and the surplus power meets the thorough cleaning intensity for the corresponding residual pollution, it is determined that the corresponding photovoltaic panel should be cleaned a second time. The thorough cleaning intensity at least includes the cleaning power required to completely clean the residual pollution. Otherwise, control the photovoltaic cleaning robot to return to the initial position to clean the next photovoltaic panel with the remaining amount of cleaning tasks or perform wireless automatic charging.

[0137] In other words, before the PV cleaning robot returns to its initial position, it will further determine whether to perform a second cleaning of the PV panels on its return journey based on the battery level. It should be noted that when a rail-mounted PV cleaning robot performs a cleaning task, it will clean the same row of PV panels and then return to its initial position, where the walking robot will replace the PV cleaning robot and clean the next row of PV panels.

[0138] The MCU of this application can obtain the remaining power of the battery management module set inside the photovoltaic cleaning robot in real time. At the same time, it can also predict the planned power corresponding to the subsequent remaining cleaning tasks based on the power consumption of the cleaned photovoltaic panels. For example, if the power consumption of the cleaned photovoltaic panels is 10% for one row of photovoltaic panels, and the remaining cleaning tasks are 7 rows of photovoltaic panels, then the planned power is 70% + M, where M is the preset power consumption of the robot returning to the initial position. If the remaining power is 90%, then the surplus power is 20% - MR. R is the charging warning power. When the remaining power of the photovoltaic cleaning robot is less than the charging warning power, it is determined that wireless charging is required. The charging warning power can be set by the user according to the actual usage scenario, and this application does not make specific restrictions on this.

[0139] At the same time, the present application also determines the residual pollution information set, which is obtained by collecting the cleaned area images of each preset range of the cleaned photovoltaic panel through the image acquisition device, and then processing each cleaned area image separately through the above-mentioned image recognition model and pollution feature extraction model. The MCU will also determine whether there is a pollution level value greater than the preset secondary cleaning threshold in the residual pollution information set corresponding to the same row of photovoltaic panels. The preset secondary cleaning threshold is a threshold pre-set by the user and is not specifically limited here. The residual pollution with a pollution level value greater than the preset secondary cleaning threshold is regarded as pollution that needs to be thoroughly cleaned. At the same time, the MCU also determines whether the surplus power meets the thorough cleaning intensity for thorough cleaning (i.e., the cleaning power for thorough cleaning). If so, the photovoltaic panel will be cleaned twice, otherwise it will return directly to the initial position. In addition to the cleaning power, the above-mentioned cleaning intensity can also include the cleaning time, such as the duration of continuous flushing or scrubbing of a certain position.

[0140] In addition, the present application can also wirelessly charge the robot immediately once the photovoltaic cleaning robot returns to its initial position, so that the robot can be kept in a fully charged state. For example, after performing a row of cleaning and returning to the initial position, it will be charged until it reaches the next row of photovoltaic panels. The robot is also wirelessly charged during the movement, which increases the charging frequency and enables the robot's power to meet the needs of performing multiple cleaning tasks a day, thereby cleaning the photovoltaic panels more flexibly.

[0141] S105, if it is determined that the photovoltaic panel needs to be cleaned for the second time, a cleaning return path is generated based on the residual pollution information set and the remaining power corresponding to the photovoltaic panel, so that the photovoltaic cleaning robot returns to the initial position after performing the second cleaning task along the cleaning return path, and determines whether to automatically charge through the wireless charging device based on the remaining power.

[0142] In the embodiment of the present application, based on the residual pollution information set and the remaining power corresponding to the photovoltaic panels, a cleaning return path is generated, specifically including:

[0143] Based on the image of the cleaned area corresponding to the cleaned photovoltaic panel, the residual contamination information set is determined. Based on the pollution level corresponding to each residual contamination in the residual contamination information set, a secondary cleaning priority sequence is generated in descending order of pollution level. Based on the remaining power and the planned power corresponding to the remaining cleaning tasks, the corresponding surplus power is determined. Based on the surplus power, the secondary cleaning priority sequence and the thorough cleaning intensity corresponding to each residual contamination, the corresponding secondary cleaning residual contamination and the corresponding residual contamination location are determined. Based on the distance between the residual contamination location and the cleaning end point, the residual contamination location is added to the return cleaning node of the cleaning return path in order from near to far.

[0144] That is to say, during the secondary cleaning, the present application will further divide the secondary cleaning priority according to the pollution degree of the residual pollution corresponding to each preset range. At the same time, the secondary cleaning plan is carried out according to the surplus power, the secondary cleaning priority sequence and the thorough cleaning intensity. The present application calculates the power consumption corresponding to each residual pollution in sequence according to the preset thorough cleaning time, the cleaning power and cleaning time corresponding to the thorough cleaning intensity, and generates a power consumption sequence in the order of the secondary cleaning priority sequence. Then, the present application accumulates and calculates the power consumption and value of the power consumption sequence in sequence until the power consumption and value are greater than the surplus power value for the first time. The last power consumption in the cumulative calculation is eliminated, and it is judged whether the power consumption and value after elimination are less than the surplus power value. If so, the residual pollution corresponding to the accumulated power consumption is regarded as the secondary cleaning residual pollution, and each secondary cleaning residual pollution is cleaned in order from near to far. If it is judged that the power consumption and value after elimination are still greater than the surplus power value, the elimination operation is continued until the power consumption and value after elimination are less than the surplus power value.

[0145] For example, the secondary cleaning priority sequence is [x2, x1, x3, x4, x5], where x2 represents the contamination level corresponding to the second preset range. If the surplus power is 10%, the power consumption sequence corresponding to the secondary cleaning priority sequence is [x2·t, x1·t, x3·t, x4·t, x5·t]. x2·t corresponds to 5% power, x1·t corresponds to 3% of the robot's battery power, and x3·t corresponds to 3% of the battery power. 5% + 3% = 8%, at which point the sum of the power consumption is less than the surplus power value. However, 5% + 3% + 3% = 11%, at which point the sum of the power consumption is greater than the surplus power value. Therefore, the residual contamination corresponding to x2 and x1 is considered the residual contamination from the secondary cleaning, and the locations L2 and L1 corresponding to the second and first preset ranges, respectively, are designated as return cleaning nodes. At this time, the second preset range corresponding to x2 is closer to the cleaning end point than the first preset range corresponding to x1, so the return cleaning path is: L2→L1, that is, the photovoltaic cleaning robot moves to the second preset range first, cleans the second preset range of the residual pollution from the secondary cleaning, and then moves to the first preset range for secondary cleaning.

[0146] Through the above technical solution, this application uses image recognition to determine the first set of cleaning pollution information, combines historical data with prediction models, and clarifies the pollution of the remaining area to be cleaned, providing a basis for accurate cleaning. At the same time, a dynamic power consumption curve is generated based on the pollution information and related curves, and input into the optimization model to achieve dynamic matching of cleaning intensity and power consumption, thereby improving power utilization efficiency. After the first cleaning, it can also determine whether to perform a second cleaning based on the remaining power and residual pollution information, and reasonably plan the return path to avoid ineffective work. After completing the task, the photovoltaic cleaning robot automatically decides whether to charge wirelessly based on the remaining power, realizing an intelligent closed loop of cleaning and charging.

[0147] The above solution significantly improves the intelligence and automation level of cleaning and charging of photovoltaic cleaning robots. Through intelligent, automated and refined management methods, it effectively solves the current technical problems of insufficient power management level of photovoltaic panel cleaning robots, simple and extensive charging solutions, and difficulty in achieving flexible, intelligent and efficient wireless automatic charging.

[0148] Figure 3 This is a schematic diagram of the structure of a wireless automatic charging device for a photovoltaic cleaning robot provided in an embodiment of the present application. Figure 3 As shown, the wireless automatic charging device 300 for the photovoltaic cleaning robot includes:

[0149] The first determination module 301 is configured to determine a first set of cleaning contamination information based on image recognition results for an image of the area to be cleaned within a preset range. The first set of cleaning contamination information includes one or more contamination sequences. A contamination sequence includes at least contamination type, frequency of occurrence, and degree of contamination. The second determination module 302 is configured to determine a second set of cleaning contamination information based on the first set of cleaning contamination information, historical cleaning data, and a preset contamination prediction model. The second set of cleaning contamination information corresponds to the remaining area to be cleaned outside the preset range. The first generation module 303 is configured to generate a dynamic power consumption curve for the photovoltaic cleaning robot based on the second set of cleaning contamination information, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve. The input module 304 is configured to input the dynamic power consumption curve into a cleaning intensity optimization model. After updating the cleaning strategy corresponding to the area to be cleaned based on the optimization results, the photovoltaic cleaning robot is controlled to clean the photovoltaic panel until it reaches the cleaning endpoint for the first time. The remaining power level is then used to determine whether to perform a second cleaning of the panel. The same cleaning strategy corresponds to the same preset range. The second generation module 305 is used to generate a cleaning return path based on the residual pollution information set and remaining power corresponding to the photovoltaic panel if it is determined that the photovoltaic panel needs to be cleaned for the second time, so as to return the photovoltaic cleaning robot to its initial position after performing the second cleaning task along the cleaning return path, and determine whether to automatically charge the photovoltaic panel through the wireless charging device based on the remaining power.

[0150] In an embodiment of the present application, the above-mentioned device is applied to a rail-mounted photovoltaic cleaning robot hung on the edge of a photovoltaic panel. Based on the image recognition results of the image of the area to be cleaned within a preset range, before determining the first cleaning contamination information set, the device can also:

[0151] Determine whether the photovoltaic cleaning robot is at the initial position or the end edge of the previous preset range. If so, obtain the surface image of the photovoltaic panel in the driving direction captured by the image acquisition device. The image acquisition device is set on the body of the photovoltaic cleaning robot. According to the size information corresponding to the preset range stored in advance, the current position of the photovoltaic cleaning robot is used as the starting edge reference point to crop the surface image of the photovoltaic panel to obtain an image of the area to be cleaned corresponding to the preset range. The shape and size of the area image of the area to be cleaned are consistent with the preset range. The area of the photovoltaic panel is composed of an integer multiple of the area corresponding to the preset range. The image of the area to be cleaned is input into a pre-trained image recognition model to determine the various contaminated grid areas within the preset range according to the output of the model. The contaminated grid area is obtained based on clustering the contaminated grids of each pollution type. A contaminated grid area includes several contaminated grids corresponding to the same pollution type. Based on each contaminated grid area and the pollution feature extraction model, a first cleaning pollution information set is determined.

[0152] The second determining module 302 is specifically configured to:

[0153] The pollution feature extraction model is used to determine the pollution feature information corresponding to each polluted grid. The pollution feature information includes at least color feature values, texture feature values, shape feature values, area feature values, and reflectivity feature values associated with the pollution level. The product and value of the pollution feature information of the polluted grid and the preset pollution feature weight are calculated to determine the corresponding pollution level coefficient. The pollution level coefficient is used to quantify the pollution level. The average value of each pollution level coefficient corresponding to the same polluted grid area is calculated, and the first product value of the average value and the pollution level weight of the corresponding pollution type is calculated as the pollution level assessment value. The pollution level assessment value is matched with multiple gradient threshold intervals to determine the pollution level corresponding to the corresponding matching gradient threshold interval as the pollution level of the pollution type, and added to the corresponding pollution sequence. The number of occurrences of polluted grids of each pollution type is counted separately to determine the pollution occurrence frequency based on the ratio of the number of occurrences to the total number of polluted grids, and added to the corresponding pollution sequence to obtain the first cleaned pollution information set.

[0154] The second determining module 302 is further configured to:

[0155] Based on the current location, determine whether the first cleaning pollution information set exists within a pre-set cleaned range. If so, the corresponding cleaned photovoltaic panel surface image, historical cleaning pollution information set, and historical cleaning strategy are sent to the cloud server as historical cleaning data. Based on the comparison results of the cleaned photovoltaic panel surface image with several corresponding pre-set cleaning residual pollution sample images, a corresponding pollution level identification deviation value is output. The pollution level identification deviation value is used to correct one or more pre-set parameters in the pre-set pollution prediction model. After receiving the pollution level identification deviation value from the cloud server, the pre-set pollution prediction model is adjusted, and the first cleaning pollution information set is input into the adjusted pre-set pollution prediction model to determine the second cleaning pollution information set based on the output results.

[0156] The device can also:

[0157] If it is determined that the first cleaning contamination information set does not contain a pre-set cleaned area, the image of the area to be cleaned is sent to the cloud server for matching against the historical case database. If a match is successful, the matched historical case is used as the historical cleaning data. Otherwise, the preset default cleaning data is used as the historical cleaning data.

[0158] The first generating module 303 is specifically configured to:

[0159] Determine, in chronological order, the pollution level sequence corresponding to the second cleaning pollution information set within a preset cleaning cycle. Based on the pollution level sequence and a preset pollution cleaning intensity curve, generate a second cleaning intensity curve corresponding to the second cleaning pollution information set. The horizontal axis of the preset pollution cleaning intensity curve represents the pollution level, and the vertical axis represents the cleaning intensity value. The cleaning intensity value corresponds to at least the cleaning power. Based on the second cleaning intensity curve and a preset pollution power generation impact curve, generate a second power generation efficiency curve corresponding to the second cleaning pollution information set. The horizontal axis of the preset pollution power generation impact curve represents the cleanliness level, and the vertical axis represents the power generation efficiency value. The cleanliness level is determined based on the cleaning intensity value, pollution type, pollution level, and a preset cleaning cleanliness comparison table. There is a positive correlation between the cleanliness level and the power generation efficiency value.

[0160] The second cleaning intensity curve and the second power generation efficiency curve are input into the following formula to generate a dynamic power consumption curve:

[0161]

[0162] Where E represents the power consumption of the dynamic power consumption curve within the cleaning cycle T. I(t) represents the cleaning intensity value at time t. C1 represents the power consumption per unit time of cleaning. G(t) represents the power generation efficiency value at time t. C2 represents the power consumption per unit time of standby mode.

[0163] The device can also:

[0164] A multi-objective optimization algorithm is established with the dynamic power consumption curve corresponding to minimizing energy consumption and maximizing cleaning efficiency as the objective function and the cleaning intensity being greater than the cleaning intensity threshold as the constraint condition.

[0165] Among them, the formula for maximizing cleaning efficiency is:

[0166]

[0167] Where η represents the cleaning efficiency. P(t) represents the actual cleaning power at time t. max represents the maximum cleaning power of the PV cleaning robot. Based on a pre-defined genetic algorithm, a multi-objective optimization algorithm is solved to obtain an optimized cleaning intensity curve. This optimized cleaning intensity curve is used as the optimization result to update the cleaning strategy. The cleaning strategy at least includes cleaning intensities for different pollution types and pollution levels of PV pollution.

[0168] The second generating module 305 is specifically configured to:

[0169] Based on the image of the cleaned area corresponding to the cleaned photovoltaic panel, the residual contamination information set is determined. Based on the pollution level corresponding to each residual contamination in the residual contamination information set, a secondary cleaning priority sequence is generated in descending order of pollution level. Based on the remaining power and the planned power corresponding to the remaining cleaning tasks, the corresponding surplus power is determined. Based on the surplus power, the secondary cleaning priority sequence and the thorough cleaning intensity corresponding to each residual contamination, the corresponding secondary cleaning residual contamination and the corresponding residual contamination location are determined. Based on the distance between the residual contamination location and the cleaning end point, the residual contamination location is added to the return cleaning node of the cleaning return path in order from near to far.

[0170] The input module 304 is further specifically used for:

[0171] Calculate the difference between the remaining power and the planned power corresponding to the remaining amount of cleaning tasks as the surplus power. The planned power at least includes the power required for the photovoltaic cleaning robot to return to the initial position. Determine whether the residual pollution information set contains residual pollution with a pollution level greater than the preset secondary cleaning threshold. If so, and the surplus power meets the thorough cleaning intensity for the corresponding residual pollution, determine to perform a secondary cleaning on the corresponding photovoltaic panel. The thorough cleaning intensity at least includes the cleaning power required to completely clean the residual pollution. Otherwise, control the photovoltaic cleaning robot to return to the initial position to clean the next photovoltaic panel with the remaining amount of cleaning tasks or perform wireless automatic charging.

[0172] The various embodiments in this application are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from other embodiments. In particular, the device embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the partial description of the method embodiments.

[0173] The device and method provided in the embodiments of the present application correspond one to one, and therefore, the device also has similar beneficial technical effects to its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device will not be repeated here.

[0174] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0175] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A wireless automatic charging method for a photovoltaic cleaning robot, characterized in that: The method comprises: Determining a first set of cleaning contamination information based on an image recognition result of an image of an area to be cleaned within a preset range; wherein the first set of cleaning contamination information includes one or more contamination sequences; the contamination sequence includes at least contamination type, contamination frequency, and contamination degree; Determining a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model; the second cleaning contamination information set corresponds to the remaining areas to be cleaned outside the preset range; Generating a dynamic power consumption curve for the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve specifically includes: determining, in chronological order, a pollution degree sequence corresponding to the second cleaning pollution information set within a preset cleaning cycle; generating a second cleaning intensity curve corresponding to the second cleaning pollution information set based on the pollution degree sequence and the preset pollution cleaning intensity curve; generating a second power generation efficiency curve corresponding to the second cleaning pollution information set based on the second cleaning intensity curve and the preset pollution power generation impact curve; and inputting the second cleaning intensity curve and the second power generation efficiency curve into the following formula to generate the dynamic power consumption curve: in, Indicates the cleaning cycle The power consumption of the dynamic power consumption curve within; Indicates time The cleaning intensity value; The power consumption per unit time for cleaning; Indicates time The power generation efficiency value; The power consumption per unit time in standby mode; The dynamic power consumption curve is input into the cleaning intensity optimization model, and a multi-objective optimization algorithm is established with the dynamic power consumption curve corresponding to minimizing energy consumption and maximizing cleaning efficiency as the objective function and the cleaning intensity being greater than the cleaning intensity threshold as the constraint condition. The formula corresponding to maximizing cleaning efficiency is: in, Indicates cleaning efficiency; Indicates time Actual cleaning power; represents the maximum cleaning power of the photovoltaic cleaning robot; based on a preset genetic algorithm, the multi-objective optimization algorithm is solved to obtain an optimized cleaning intensity curve, and the optimized cleaning intensity curve is used as the optimization result to update the cleaning strategy; the cleaning strategy at least includes the cleaning intensity for photovoltaic pollution of different pollution types and different pollution degrees; Controlling the photovoltaic cleaning robot to clean the photovoltaic panel until it reaches the cleaning end point of the corresponding photovoltaic panel for the first time, and determining whether to perform a second cleaning on the corresponding photovoltaic panel based on the remaining power; wherein the same cleaning strategy corresponds to the same preset range; If it is determined that the photovoltaic panel needs to be cleaned a second time, a cleaning return path is generated based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power, so that the photovoltaic cleaning robot returns to the initial position after performing the second cleaning task along the cleaning return path, and determines whether to automatically charge through the wireless charging device based on the remaining power.

2. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 1, characterized in that: The method is applied to a rail-mounted photovoltaic cleaning robot hung on the edge of a photovoltaic panel; before determining a first cleaning contamination information set based on an image recognition result of an image of an area to be cleaned within a preset range, the method further includes: Determining whether the photovoltaic cleaning robot is at the initial position or the end edge of the previous preset range; If yes, obtain an image of the photovoltaic panel surface in the driving direction captured by an image acquisition device; wherein the image acquisition device is set on the body of the photovoltaic cleaning robot; Based on the pre-stored size information corresponding to the preset range, the photovoltaic panel surface image is cropped with the current position of the photovoltaic cleaning robot as the starting edge reference point to obtain the image of the area to be cleaned corresponding to the preset range; wherein the shape and size of the area of the image of the area to be cleaned are consistent with the preset range; and the area of the photovoltaic panel is composed of an integer multiple of the area corresponding to the preset range; Inputting the image of the area to be cleaned into a pre-trained image recognition model to determine each contaminated grid area within the preset range based on the model output result; wherein the contaminated grid area is obtained by clustering contaminated grids of each contamination type; and one contaminated grid area includes multiple contaminated grids corresponding to the same contamination type; The first cleaning contamination information set is determined based on each of the contaminated grid areas and the contamination feature extraction model.

3. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 2, characterized in that: Determining the first cleaning pollution information set based on each of the pollution grid areas and the pollution feature extraction model specifically includes: Determining pollution feature information corresponding to each of the pollution grids using the pollution feature extraction model; wherein the pollution feature information includes at least a color feature value, a texture feature value, a shape feature value, an area feature value, and a reflectivity feature value associated with the pollution degree; Calculating the product sum of the pollution characteristic information of the pollution grid and the preset pollution characteristic weight to determine a corresponding pollution degree coefficient; the pollution degree coefficient is used to quantify the pollution degree; Calculating an average value of the pollution degree coefficients corresponding to the same pollution grid area, and calculating a first product value of the average value and the pollution degree weight of the corresponding pollution type as a pollution degree assessment value; Matching the pollution level assessment value with a plurality of gradient threshold intervals to determine that the pollution level corresponding to the corresponding matching gradient threshold interval is the pollution level of the pollution type, and adding the pollution level to the corresponding pollution sequence; The number of occurrences of the contaminated grids of each contamination type is counted respectively, and the contamination occurrence frequency is determined according to the ratio of the number of occurrences to the total number of contaminated grids, and is added to the corresponding contamination sequence to obtain the first cleaning contamination information set.

4. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 2, characterized in that: Determining a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model specifically includes: Determining, based on the current position, whether the first cleaning contamination information set has a cleaned preset range; If so, the corresponding cleaned photovoltaic panel surface image, the historical cleaning pollution information set, and the historical cleaning strategy are used as the historical cleaning data and sent to the cloud server to output a corresponding pollution degree identification deviation value based on the comparison result of the cleaned photovoltaic panel surface image with a number of corresponding preset cleaning residual pollution sample images; wherein the pollution degree identification deviation value is used to correct one or more preset parameters in the preset pollution prediction model; After receiving the pollution degree identification deviation value from the cloud server, the preset pollution prediction model is adjusted, and the first cleaning pollution information set is input into the adjusted preset pollution prediction model to determine the second cleaning pollution information set based on the output result.

5. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 4, characterized in that: The method further comprises: If it is determined that the first cleaning contamination information set does not contain the cleaned preset range, sending the image of the area to be cleaned to the cloud server so that the cloud server matches it in a historical case database; If the match is successful, the matched historical case is used as the historical cleansing data; Otherwise, the preset default cleaning data is used as the historical cleaning data.

6. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 1, characterized in that: generating a dynamic power consumption curve for the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve, wherein the abscissa of the preset pollution cleaning intensity curve represents the pollution degree and the ordinate represents the cleaning intensity value; the cleaning intensity value at least corresponds to the cleaning power; Among them, the horizontal axis of the preset pollution power generation impact curve is cleanliness, and the vertical axis is the power generation efficiency value; the cleanliness is obtained based on the cleaning intensity value, the pollution type, the pollution degree and the preset cleaning cleanliness comparison table; the cleanliness and the power generation efficiency value are positively correlated.

7. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 1, characterized in that: Based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power, a cleaning return path is generated, specifically including: determining the residual contamination information set based on the cleaned area image corresponding to the cleaned photovoltaic panel; Generate a secondary cleaning priority sequence according to the pollution degree corresponding to each residual pollution in the residual pollution information set and in descending order of pollution degree; Determining corresponding surplus power based on the remaining power and the planned power corresponding to the remaining cleaning task amount; Determining corresponding secondary cleaning residual contamination and corresponding residual contamination locations according to the surplus power, the secondary cleaning priority sequence, and the thorough cleaning intensity corresponding to each residual contamination; According to the distance between the remaining contamination location and the cleaning end point, the remaining contamination location is sequentially added to the return cleaning node of the cleaning return path in order from near to far.

8. The wireless automatic charging method for a photovoltaic cleaning robot according to claim 1, characterized in that: Determining whether to perform secondary cleaning on the corresponding photovoltaic panel based on the remaining power specifically includes: Calculating the difference between the remaining power and the planned power corresponding to the remaining cleaning task amount as the surplus power; wherein the planned power at least includes the power required for the photovoltaic cleaning robot to return to its initial position; Determining whether the residual contamination information set contains residual contamination with a contamination degree greater than a preset secondary cleaning threshold; If yes, and the surplus power meets the intensity of thorough cleaning of the corresponding residual contamination, it is determined that the corresponding photovoltaic panel should be cleaned twice; wherein the thorough cleaning intensity at least includes the cleaning power required to completely clean the residual contamination; Otherwise, the photovoltaic cleaning robot is controlled to return to the initial position to clean the next photovoltaic panel with the remaining cleaning task amount or perform wireless automatic charging.

9. A wireless automatic charging device for a photovoltaic cleaning robot, characterized in that: The device uses a wireless automatic charging method for a photovoltaic cleaning robot according to any one of claims 1 to 8; the device comprises: A first determination module is configured to determine a first set of cleaning contamination information based on an image recognition result of an image of an area to be cleaned within a preset range; wherein the first set of cleaning contamination information includes one or more contamination sequences; and the contamination sequence includes at least contamination type, contamination frequency, and contamination degree. a second determining module, configured to determine a second cleaning contamination information set based on the first cleaning contamination information set, historical cleaning data, and a preset contamination prediction model; wherein the second cleaning contamination information set corresponds to the remaining to-be-cleaned area outside the preset range; A first generating module is configured to generate a dynamic power consumption curve of the photovoltaic cleaning robot based on the second cleaning pollution information set, a preset pollution cleaning intensity curve, and a preset pollution power generation impact curve; an input module, configured to input the dynamic power consumption curve into a cleaning intensity optimization model, and after updating a cleaning strategy corresponding to the area to be cleaned based on the optimization result, control the photovoltaic cleaning robot to clean the photovoltaic panel until the cleaning endpoint of the corresponding photovoltaic panel is reached for the first time, and determine whether to perform a second cleaning of the corresponding photovoltaic panel based on the remaining power; wherein the same cleaning strategy corresponds to the same preset range; The second generation module is used to generate a cleaning return path based on the residual pollution information set corresponding to the photovoltaic panel and the remaining power if it is determined that the photovoltaic panel needs to be cleaned for the second time, so as to return the photovoltaic cleaning robot to its initial position after performing the second cleaning task along the cleaning return path, and determine whether to automatically charge through a wireless charging device based on the remaining power.

Images