A cleaning method for a photovoltaic cleaning robot

Abstract

The present invention provides a cleaning method for a photovoltaic cleaning robot, which mainly includes the steps of: collecting information of photovoltaic panels to be cleaned and extracting environmental features to select a matching cleaning mode; obtaining a prefabricated cleaning plan under the matching cleaning mode; selecting a stored cleaning plan and then obtaining a working cleaning plan by fitting. The photovoltaic cleaning robot performs cleaning tasks according to the working cleaning plan. The method of the present invention collects on-site conditions through multiple sensors, specifies a prefabricated plan in combination with prefabricated information and on-site conditions, and formulates a plan suitable for the current task by multi-dimensional fitting; at the same time, random adjustments are made based on dynamic weights to avoid overfitting in approximate scenes, thereby reducing the problem of falling into local optimality, and thus better implementation effects can be obtained without significantly increasing the difficulty of calculation.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a cleaning method of a photovoltaic cleaning robot. Background Art

[0002] The photovoltaic intelligent cleaning robot is a device specially designed for intelligent cleaning of photovoltaic panels. It has the characteristics of a high degree of automation, which can reduce the investment in labor costs and optimize the operations of management personnel; at the same time, it ensures cleaning efficiency and cleaning effects, and effectively improves the level of photovoltaic power generation.

[0003] The existing photovoltaic intelligent cleaning robots have strong obstacle-crossing and climbing capabilities. They not only have excellent endurance, but can also be remotely controlled through mobile terminals or the cloud. While being convenient and efficient, they greatly reduce the time cost of operation and maintenance personnel, help improve the level of automation in photovoltaic power station management, and promote the intelligent development of photovoltaic power station operation and maintenance management. For example, the public has designed and developed an intelligent cleaning robot that can walk intelligently and clean autonomously without water. It can clean foreign objects such as sand, dust, fallen leaves, and snow on the surface of photovoltaic panels in photovoltaic power stations. It is widely used in different types of power stations such as large ground power stations and distributed rooftop power stations. Another example is a cleaning method and a photovoltaic cleaning system for a photovoltaic cleaning robot proposed in Chinese patent application 202310623133.3. It can formulate corresponding strategies for photovoltaic cleaning robots and photovoltaic trackers to perform cleaning tasks according to various real-time climates, avoid performing cleaning tasks in extreme weather and environments that are not suitable for performing cleaning tasks, and ensure that adjacent photovoltaic panels can maintain the same cleaning angle for the cleaning robot to pass safely.

[0004] However, the current design cannot meet the diverse needs of the market very well. The specific analysis is as follows:

[0005] At present, the relevant robot categories are:

[0006] 1. Full-width straddle-mounted cleaning robot, the whole machine spans both sides of the photovoltaic panel, and after simple setting, it can regularly clean the single-sided photovoltaic panel; although the overall structure is simple, the scope of application is small, and facing a large number of photovoltaic panel splicing scenes or distributed setting scenes, its overall cost will increase significantly; although there is also a design of a single machine with a movable base station design, the existence of the base station will also lead to an increase in cost, and it is still not suitable for photovoltaic arrays used for large-area splicing.

[0007] 2. Stand-alone omnidirectional mobile cleaning robot. This type of robot is small in size and can complete complex cleaning tasks through reasonable cleaning strategies and adapt to different usage scenarios.

[0008] Therefore, the current mainstream photovoltaic panel cleaning uses a single omnidirectional mobile cleaning robot. However, in order to play a cleaning role, a reasonable cleaning strategy must be specified, including the motion parameters of the cleaning mechanism and the motion path of the whole machine.

[0009] In the prior art, cleaning methods for photovoltaic panel cleaning robots are mostly divided into two categories. The first category is to set the cleaning route and select the cleaning mode of the surrounding environment, and then perform cleaning according to a predetermined plan after a simple analysis. Although this setting reduces the amount of calculation of the system, there are problems of low cleaning effect or waste of resources. The second category is to collect data through multiple sensors and combine cutting-edge artificial intelligence technologies such as machine learning to plan the cleaning strategy in detail. This has a better cleaning effect and saves overall resources, and has become the mainstream research and development direction. However, the second type of design method still has the problem of large amount of calculation and excessive system resource occupation in practical application, which will cause the cleaning process to take a long time to analyze, and the overall cleaning efficiency is still reduced; accordingly, there are two ways to solve it. One is to build a platform with excellent computing power locally, so as to realize the rapid formulation of the corresponding cleaning method. However, this requires a large amount of money to be invested in server construction in the early stage and to purchase expensive microelectronic equipment (such as high-performance graphics processors), and when failures occur in later use, it is difficult to purchase and replace parts in the first time; the other is to configure the corresponding system through cloud services provided by a third party, thereby saving initial investment. However, the common deployment areas of existing large-scale photovoltaic power stations are urban suburbs in the central and western regions. The cost of configuring communication facilities that meet the requirements alone is too high, and the signal connection is unstable. In addition, it is necessary to continue to pay high cloud usage fees, which still affects the overall cleaning plan.

[0010] In summary, a cleaning method that balances system complexity, usage cost and cleaning effect is also needed for the above-mentioned scenarios to meet the different usage requirements of conventional cleaning robots in the market. Summary of the invention

[0011] In view of the problems existing in the prior art, the present invention provides a cleaning method for a photovoltaic cleaning robot which has low hardware requirements, ideal cleaning effect and adaptive strategy adjustment.

[0012] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0013] The present invention provides a cleaning method of a photovoltaic cleaning robot, which mainly comprises the following steps:

[0014] The photovoltaic panel parameters, environmental parameters and image information of the photovoltaic panel to be cleaned are collected through the sensors of the photovoltaic cleaning robot;

[0015] Extracting the environmental parameters and environmental features of the image information, and selecting a matching cleaning mode in a database based on the environmental features;

[0016] In the matching cleaning mode, according to the environmental characteristics and the photovoltaic panel parameters, a prefabricated cleaning plan is output through a path planning method and a prefabricated strategy; the prefabricated cleaning plan at least includes preset power parameters of the cleaning mechanism of the photovoltaic cleaning robot and a calculated cleaning path;

[0017] Selecting a matching stored cleaning plan in the database based on the environmental characteristics and the photovoltaic panel parameters; the stored cleaning plan at least includes stored power parameters and stored cleaning paths of the cleaning mechanism of the photovoltaic cleaning robot;

[0018] The preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism are fitted by a dynamic weighting algorithm to obtain the working power parameters of the cleaning mechanism; at least one dynamic random weight is provided in the dynamic weighting algorithm;

[0019] The calculated cleaning path and the stored cleaning path are used to obtain a working cleaning path by a path fitting method;

[0020] Replacing the preset power parameters of the cleaning mechanism and the calculated cleaning path in the prefabricated cleaning plan with the working power parameters of the cleaning mechanism and the working cleaning path, respectively, to obtain a working cleaning plan;

[0021] The photovoltaic cleaning robot performs a cleaning task according to the work cleaning plan, and after completing the cleaning task, collects surface information of the photovoltaic panel through the sensor of the photovoltaic cleaning robot;

[0022] Based on the surface information, the cleanliness of the photovoltaic panel is calculated by image analysis, and the environmental characteristics of this task, photovoltaic panel parameters, photovoltaic panel cleanliness, working power parameters of the cleaning mechanism and working cleaning path are associated and stored in the database.

[0023] Optionally, the photovoltaic panel parameters include photovoltaic panel attitude parameters and photovoltaic panel size parameters;

[0024] The environmental parameters include air temperature, humidity and wind speed;

[0025] The image information includes a surface image of the photovoltaic panel to be cleaned.

[0026] Optionally, the environmental characteristics include season information, meteorological information, dust coverage and type of obstruction;

[0027] The types of obstructions include sand, gravel, stagnant water, snow, and bird excrement;

[0028] The preset power parameters of the cleaning mechanism include at least one of the cleaning mechanism downforce, the cleaning mechanism working end rotation speed, the cleaning liquid consumption rate, and the steam temperature;

[0029] The preset power parameters of the cleaning mechanism are set in association with the environmental characteristics.

[0030] Optional, associated settings include:

[0031] If the type of obstruction is sand, bird excrement, and the information label is no rain or frost, the cleaning mechanism starts the cleaning roller to clean along the path and spray water for cleaning simultaneously;

[0032] If the obstruction is sand, bird excrement, or the information label indicates that there has been rain or frost within 15 minutes to 72 hours, the cleaning mechanism will first spray water along the route and let it stand for 3 minutes to 5 minutes, and then start the cleaning roller to clean along the reverse route;

[0033] If the type of obstruction is snow and the information tag shows that the temperature rises above zero degrees, the cleaning mechanism starts the cleaning roller to sweep along the path and simultaneously sprays high-temperature steam to melt snow and ice.

[0034] Optionally, the environmental characteristics also include past meteorological information;

[0035] The past weather information is set as an information tag of the type of the obstruction.

[0036] Optionally, the structure of the database is a multi-root hierarchical database structure;

[0037] The database is provided with a plurality of root nodes, and the identification of the root node includes the environmental feature;

[0038] A plurality of first-layer child nodes are provided under the root node, and the prefabricated strategies and prefabricated cleaning plan templates are stored in the first-layer child nodes;

[0039] A plurality of second-layer sub-nodes are arranged under the first-layer sub-node, and the second-layer sub-nodes store storage cleaning plans.

[0040] Optionally, the path planning method is a bow-shaped path calculation method or an S-shaped path calculation method.

[0041] Optionally, the storage cleaning plan matching method includes:

[0042] Calculate the similarity between the environmental characteristics of each storage cleaning plan in the database and the environmental characteristics of the current task, and select the storage cleaning plans whose similarity exceeds a threshold to form a storage plan set;

[0043] The elements in the storage plan set are sorted according to the cleanliness of the photovoltaic panels, and the first n storage cleaning plans with high cleanliness of the photovoltaic panels are selected as the matching storage cleaning plans; n is a natural number greater than or equal to 1.

[0044] Optionally, the step of fitting the dynamic weighted algorithm includes:

[0045] Extract the preset weight array in the cleaning mode corresponding to the current task;

[0046] Adding random disturbance to at least one group of weights in the weight array to obtain dynamic random weights and a corresponding dynamic weight array;

[0047] The preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism are weightedly fitted through the dynamic weight array.

[0048] Optionally, the path fitting method is a piecewise linear fitting method or a feature point fitting method.

[0049] Optionally, the image analysis method is based on a digital image processing method.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The method of the present invention collects on-site conditions through multiple sensors, specifies a prefabrication plan in combination with prefabrication information and on-site conditions, and formulates a plan suitable for the current task through multi-dimensional fitting; at the same time, random adjustments are made based on dynamic weights to avoid overfitting in approximate scenarios, thereby reducing the problem of falling into local optimality, and thus better implementation effects can be obtained without significantly increasing the difficulty of calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0056] In the description of the present invention, “plurality” means two or more than two, unless otherwise clearly and specifically defined.

[0057] It is worth noting that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products, and their sources are not specifically limited unless otherwise specified.

[0058] It should also be noted that the following embodiments are described in a certain order in order to facilitate the understanding of the technical solution. Some steps can be performed simultaneously or in an adjusted order, and the description order should not be understood as a limiting condition.

[0059] like Figure 1 As shown, this embodiment provides a cleaning method of a photovoltaic cleaning robot, which mainly includes the following steps:

[0060] First, according to the overall cleaning plan, the photovoltaic cleaning robot is deployed to the corresponding starting position, such as the photovoltaic panel near the edge of the photovoltaic array, by means of vehicle carrying, manual handling, self-propelled platform movement, etc. Then, the corresponding system is started.

[0061] The photovoltaic panel parameters, environmental parameters and image information of the photovoltaic panel to be cleaned are collected through the sensors of the photovoltaic cleaning robot; the photovoltaic cleaning robot is explained by taking a common universal movable robot as an example, and its main body is a crawler-type self-propelled platform with cleaning mechanisms installed at both ends. The main body of the cleaning mechanism is a cleaning roller arranged at the bottom, and a liquid spray port and an air jet port are arranged on the side. The liquid spray port is used to spray cleaning liquid or water; the air jet port is used to spray high-speed airflow or high-temperature steam. The sensors of the photovoltaic cleaning robot are commonly installed posture sensors, distance sensors, smart cameras, laser radars, etc.

[0062] Optionally, the photovoltaic panel parameters include photovoltaic panel attitude parameters and photovoltaic panel size parameters. Photovoltaic panel attitude parameters are mainly the tilt angle and tilt direction of the photovoltaic panel. Photovoltaic panel size parameters are the panel size of the photovoltaic panel, such as length, width and robot initial position. The above parameters can be obtained through sensor measurement data or retrieved from the corresponding database.

[0063] Furthermore, based on the above information, the system also extracts environmental parameters and image information, wherein the environmental parameters of this embodiment include the current surrounding temperature, humidity and wind speed collected by the photovoltaic cleaning robot through temperature and humidity sensors and anemometers, and image information collected through smart cameras, including surface images of the photovoltaic panels to be cleaned.

[0064] The system extracts environmental features based on the above information, including season information, meteorological information, dust coverage and types of obstructions. Specifically, in the process of extracting environmental features, the system determines the current season information based on the current time and the aforementioned environmental parameters. Although season information can be directly obtained by simply relying on time and geographical location, due to the abnormal weather conditions in local areas, similarity matching in the meteorological database combined with current meteorological parameters can obtain season information that is more conducive to the formulation of subsequent plans. Meteorological information is specifically the type of weather, such as rainy days, snowy days, and windy days. Dust coverage can be directly identified by image information. Through traditional image processing algorithms, such as threshold segmentation, one or more grayscale thresholds are set to divide the image into different areas. For dirt such as dust, due to the difference in grayscale value between it and the panel, a suitable threshold can be selected to separate the dirt from the background; its calculation is simple and fast, but the effect may not be good for images with uneven grayscale distribution. Therefore, it can be further identified in combination with edge detection method. The edge information in the image can be detected by edge detection operator to determine the position and shape of the dirt. Among them, common edge detection operators include Sobel operator, Canny operator, etc. Finally, after edge detection of the image, morphological processing (such as expansion, corrosion, etc.) can be used to further optimize the edge detection results and remove noise and false edges. The image features of the occluder are extracted by the above-mentioned image feature extraction method. After comparing the data, the types of occluders can be identified, including sand, gravel, stagnant water, snow, bird excrement, etc.

[0065] Optionally, under the conditions for obtaining the above-mentioned environmental characteristics, the environmental characteristics also include past meteorological information; the past meteorological information here refers to the meteorological conditions that occurred within a week before the current task, such as rainfall 24 hours ago; and the past meteorological information is set as an information label of the obstruction type. The reason for this design is that dust and other obstructions will present different states under different meteorological conditions, especially corresponding strategies are needed in the subsequent cleaning process.

[0066] According to the structure of the cleaning robot in this embodiment, the preset power parameters of the cleaning mechanism are designed to include the downward pressure of the cleaning mechanism, the rotation speed of the working end of the cleaning mechanism, the consumption rate of the cleaning liquid and the steam temperature, and the preset power parameters of the cleaning mechanism are associated with the environmental characteristics. Optionally, the associated settings include:

[0067] If the type of obstruction is sand, bird excrement, and the information label is no rain or frost, the cleaning mechanism starts the cleaning roller to clean along the path and spray water for cleaning simultaneously;

[0068] If the obstruction is sand, bird excrement, or the information label indicates that there has been rain or frost within 15 minutes to 72 hours, the cleaning mechanism will first spray water along the route and let it stand for 3 minutes to 5 minutes, and then start the cleaning roller to clean along the reverse route;

[0069] If the type of obstruction is snow and the information tag shows that the temperature rises above zero degrees, the cleaning mechanism starts the cleaning roller to sweep along the path and simultaneously sprays high-temperature steam to melt snow and ice.

[0070] The above shows some of the associated settings for explanation. There are other associated settings accordingly. The specific situation can be adjusted and set according to the meteorological characteristics of the installation area and past experience combined with common knowledge.

[0071] A matching cleaning mode is selected from the database based on environmental features; specifically, the database is built in a local server, a local area network is established between the cleaning robot and the server, and communication is performed through a wireless network. Optionally, the database of this embodiment adopts a multi-root hierarchical database structure; its structure is simple and convenient for indexing and extracting data. The database is provided with multiple root nodes as different cleaning modes, and the identification of the root node includes at least part of the environmental features, such as season information and meteorological information; a plurality of first-layer child nodes are provided under the root node, and the first-layer child nodes store prefabricated strategies and prefabricated cleaning plan templates; a plurality of second-layer child nodes are provided under the first-layer child nodes, and the second-layer child nodes store cleaning plans. The matching cleaning mode is selected by finding the closest root node among multiple root nodes based on known environmental features. For example, if the currently acquired season information is autumn and the meteorological information is sunny, the corresponding root node identified as autumn-sunny is found in the root node of the database, and the prefabricated strategy and prefabricated cleaning plan template corresponding to the first-layer child node are extracted. Among them, the prefabricated strategy includes the path type; the prefabricated cleaning plan template is a work template set up manually based on past cleaning experience and simulation calculations, including the power parameters of the cleaning mechanism, especially the down pressure and rotation speed of the cleaning roller.

[0072] Thus, in the matching cleaning mode, the system outputs the prefabricated cleaning plan according to the environmental characteristics and photovoltaic panel parameters through the path planning method and the extracted prefabricated strategy and prefabricated cleaning plan template. The prefabricated cleaning plan at least includes the preset power parameters of the cleaning mechanism of the photovoltaic cleaning robot and the calculated cleaning path; specifically, since the common photovoltaic panels are of regular shape, the path planning method adopted in this embodiment is also the common bow-shaped path calculation method or S-shaped path calculation method, that is, the cleaning robot moves back and forth on the photovoltaic panel to complete the full-width cleaning work. Further, in order to better complete the cleaning work, the preset power parameters of the cleaning mechanism and the calculated cleaning path are associated through the prefabricated strategy and prefabricated cleaning plan template that meet the scene, so as to avoid excessive repeated cleaning and obtain better cleaning effect. For example: in the autumn-sunny scene, the prefabricated strategy sets the spacing between adjacent parallel segments in the path to be equal to or slightly less than the width of the cleaning roller; in the summer-sunny scene, the prefabricated strategy sets the spacing between adjacent parallel segments in the path to be equal to or slightly less than half the width of the cleaning roller.

[0073] A matching storage cleaning plan is selected based on the corresponding second-layer subnodes in the database based on the environmental characteristics and photovoltaic panel parameters.

[0074] Optionally, the storage cleaning plan matching method includes:

[0075] Calculate the similarity between the environmental characteristics of each storage cleaning plan in the database and the environmental characteristics of this task, and select the storage cleaning plans whose similarity exceeds the threshold to form a storage plan set [storage cleaning plan 1, storage cleaning plan 2...storage cleaning plan m];

[0076] The elements in the storage plan set are sorted according to the cleanliness of the photovoltaic panels, and the first n storage cleaning plans with high cleanliness of the photovoltaic panels are selected as matching storage cleaning plans, where n is a natural number greater than or equal to 1.

[0077] The above stored cleaning plan at least includes stored power parameters of the cleaning mechanism of the photovoltaic cleaning robot and stored cleaning paths.

[0078] The preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism are fitted by a dynamic weighting algorithm to obtain the working power parameters of the cleaning mechanism, and at least one dynamic random weight is set in the dynamic weighting algorithm.

[0079] The steps of dynamic weighted algorithm fitting include:

[0080] It is known that the preset power parameters of the cleaning mechanism are [a1, a2, a3...a k ]; the storage power parameter of the cleaning mechanism is [a save,1 , a save,2 , a save,3 ……asave,k ].

[0081] S1, extract the preset weight array [w1, w2, w3...w 2k ]; where w 2i-1 and w 2i is a group, and w 2i-1 +w 2i =1,i=1,2,...,k.

[0082] S2. Add random disturbance to at least one group of weights in the weight array to obtain dynamic random weights and corresponding dynamic weight arrays. In this embodiment, a group of weights of the speed of the cleaning roller in the cleaning mechanism is taken as an example. Assume that the group of weights is w z and w z+1 , (z is a natural number), a percentage value j (j is a real number) in the range [0, p] (p is a positive integer) is randomly generated by a random number generator, and then the disturbance value is calculated as w z *j / 100, then the new set of weights is w z +w z *j / 100;w z+1 -w z *j / 100, replace the original corresponding weights w z and w z+1 , to obtain an array of dynamic weights.

[0083] S3, weighted fitting of the preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism through a dynamic weight array. The formula for the working power parameters of the cleaning mechanism is:

[0084] a work,i =a i *w 2i-1 +a save,i *w 2i ;

[0085] Thus, the working power parameters of the cleaning mechanism are obtained as [a work,1 , a work,2 , a work,3 ……a work,k]. In this way, the power parameters of the cleaning mechanism can be prevented from infinitely approaching the preset information after multiple fittings in subsequent work through the method of perturbation weighted fitting. Because the information first stored in the database is the past records entered by personnel, and it also serves as a basic template; and after multiple subsequent work, the power parameters of photovoltaic panels with high cleanliness will gradually converge, so it is easy to fall into the problem of local optimality. Therefore, by changing the weights through perturbation, on the one hand, controlling the degree of disturbance can ensure that the cleaning effect meets the minimum requirements, and on the other hand, the diversity of each setting is increased in the form of perturbation. It does not require the high computing power support required by machine learning, and can also continuously explore better cleaning effects in the same scenario through the aforementioned optimal fitting in a random manner to jump out of the local optimality.

[0086] Further, the cleaning path is calculated and stored to obtain a working cleaning path through a path fitting method; optionally, the path fitting method is a piecewise linear fitting method or a feature point fitting method. Taking the piecewise linear fitting method as an example, the bow-shaped path or the S-shaped path is divided into several shorter line segments, and a linear function is used to fit each line segment, that is, the coordinates of the two endpoints are determined and connected into a straight line segment.

[0087] The preset power parameters of the cleaning mechanism and the calculated cleaning path in the prefabricated cleaning plan are replaced with the working power parameters of the cleaning mechanism and the working cleaning path, respectively, to obtain the working cleaning plan.

[0088] The photovoltaic cleaning robot performs cleaning tasks according to the work cleaning plan, and after completing the cleaning tasks, the photovoltaic cleaning robot collects surface information of the photovoltaic panel through its sensor. Among them, the surface information is the surface image of the photovoltaic panel after cleaning, and based on the surface information, the cleanliness of the photovoltaic panel is calculated by image analysis. Optionally, the image analysis method of this embodiment is based on a digital image processing method, taking the region growing method as an example, a pixel point in a dirt area is selected as a seed point, and then the surrounding pixel points are merged into the dirt area according to the color, texture and other characteristics, that is, starting from one or more seed points, the pixels in the neighborhood are gradually merged into the seed area according to the pixel similarity criterion until a certain stop condition is met. This method is different from the aforementioned image processing method, and can be used to detect water stains after cleaning and a small amount of residual stains.

[0089] Finally, the environmental characteristics, photovoltaic panel parameters, photovoltaic panel cleanliness, cleaning mechanism working power parameters and working cleaning path of this task are associated and stored in the database to form the aforementioned storage cleaning plan for subsequent task calls.

[0090] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A cleaning method for a photovoltaic cleaning robot, characterized in that: The steps include: The photovoltaic panel parameters, environmental parameters and image information of the photovoltaic panel to be cleaned are collected through the sensors of the photovoltaic cleaning robot; Extracting the environmental parameters and environmental features of the image information, and selecting a matching cleaning mode in a database based on the environmental features; In the matching cleaning mode, according to the environmental characteristics and the photovoltaic panel parameters, a prefabricated cleaning plan is output through a path planning method and a prefabricated strategy; the prefabricated cleaning plan at least includes preset power parameters of the cleaning mechanism of the photovoltaic cleaning robot and a calculated cleaning path; Selecting a matching stored cleaning plan in the database based on the environmental characteristics and the photovoltaic panel parameters; the stored cleaning plan at least includes stored power parameters and stored cleaning paths of the cleaning mechanism of the photovoltaic cleaning robot; The preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism are fitted by a dynamic weighting algorithm to obtain the working power parameters of the cleaning mechanism; at least one dynamic random weight is provided in the dynamic weighting algorithm; The calculated cleaning path and the stored cleaning path are used to obtain a working cleaning path by a path fitting method; Replacing the preset power parameters of the cleaning mechanism and the calculated cleaning path in the prefabricated cleaning plan with the working power parameters of the cleaning mechanism and the working cleaning path, respectively, to obtain a working cleaning plan; The photovoltaic cleaning robot performs a cleaning task according to the work cleaning plan, and after completing the cleaning task, collects surface information of the photovoltaic panel through the sensor of the photovoltaic cleaning robot; Based on the surface information, the cleanliness of the photovoltaic panel is calculated by image analysis, and the environmental characteristics of this task, photovoltaic panel parameters, photovoltaic panel cleanliness, working power parameters of the cleaning mechanism and working cleaning path are associated and stored in the database.

2. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The photovoltaic panel parameters include photovoltaic panel attitude parameters and photovoltaic panel size parameters; The environmental parameters include air temperature, humidity and wind speed; The image information includes a surface image of the photovoltaic panel to be cleaned.

3. The cleaning method of the photovoltaic cleaning robot according to claim 1 or 2, characterized in that: The environmental characteristics include seasonal information, meteorological information, dust coverage and types of obstructions; The types of obstructions include sand, gravel, stagnant water, snow, and bird excrement; The preset power parameters of the cleaning mechanism include at least one of the cleaning mechanism downforce, the cleaning mechanism working end rotation speed, the cleaning liquid consumption rate, and the steam temperature; The preset power parameters of the cleaning mechanism are set in association with the environmental characteristics.

4. The cleaning method of the photovoltaic cleaning robot according to claim 3, characterized in that: The environmental characteristics also include past meteorological information; The past weather information is set as an information tag of the type of the obstruction.

5. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The structure of the database is a multi-root hierarchical database structure; The database is provided with a plurality of root nodes, and the identification of the root node includes the environmental feature; A plurality of first-layer child nodes are provided under the root node, and the prefabricated strategies and prefabricated cleaning plan templates are stored in the first-layer child nodes; A plurality of second-layer sub-nodes are arranged under the first-layer sub-node, and the second-layer sub-nodes store storage cleaning plans.

6. The cleaning method of the photovoltaic cleaning robot according to claim 5, characterized in that: The path planning method is a bow-shaped path calculation method or an S-shaped path calculation method.

7. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The matching methods for storage cleaning plan include: Calculate the similarity between the environmental characteristics of each storage cleaning plan in the database and the environmental characteristics of the current task, and select the storage cleaning plans whose similarity exceeds a threshold to form a storage plan set; The elements in the storage plan set are sorted according to the cleanliness of the photovoltaic panels, and the first n storage cleaning plans with high cleanliness of the photovoltaic panels are selected as the matching storage cleaning plans; n is a natural number greater than or equal to 1.

8. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The steps of dynamic weighted algorithm fitting include: Extract the preset weight array in the cleaning mode corresponding to the current task; Adding random disturbance to at least one group of weights in the weight array to obtain dynamic random weights and a corresponding dynamic weight array; The preset power parameters of the cleaning mechanism and the stored power parameters of the cleaning mechanism are weightedly fitted through the dynamic weight array.

9. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The path fitting method is a piecewise linear fitting method or a feature point fitting method.

10. The cleaning method of the photovoltaic cleaning robot according to claim 1, characterized in that: The image analysis method is based on digital image processing method.

Images