Photovoltaic panel cleaner and control method thereof

Abstract

The invention discloses a photovoltaic panel cleaning device and aims to provide a photovoltaic panel cleaner which is high in cleaning efficiency, good in cleaning effect, high in automation degree and capable of monitoring the state of a photovoltaic panel and a control method of the photovoltaic panel cleaner. The power generation efficiency of the photovoltaic panel is detected through the control system, on the basis that the cleaning robot cleans the photovoltaic panel at the fixed frequency, the cleaning frequency is additionally increased according to the change of the power generation efficiency of the photovoltaic panel, the power generation efficiency of the photovoltaic panel is guaranteed, and meanwhile resource waste caused by too frequent cleaning can be prevented; and meanwhile, the monitoring system can replace manual work to confirm the cleaning effect after the cleaning robot completes cleaning, the cleaning effect is checked and confirmed under the condition that no person exists, the detection process can be shortened, and the photovoltaic power station cleaning system and method are suitable for the technical field of photovoltaic power stations.

Description

Technical Field

[0001] The present invention relates to a photovoltaic panel cleaning device, and more particularly, to a photovoltaic panel cleaner and a control method thereof. Background Art

[0002] Dust, dirt and other pollutants on the surface of photovoltaic panels will reduce the power generation efficiency of photovoltaic panels and affect the absorption and conversion of light energy. Traditional methods of cleaning photovoltaic panels usually require manual cleaning, which is time-consuming, labor-intensive and inefficient. Photovoltaic power stations are usually located outdoors and are easily affected by factors such as climate and environment, resulting in serious surface pollution of photovoltaic panels. If not cleaned in time, it will not only affect the power generation efficiency of photovoltaic panels, but also the accumulation of pollutants will damage photovoltaic panels and reduce the reliability and stability of photovoltaic power stations. Clean energy is the trend of future energy development. Photovoltaic power stations, as a clean energy power generation method, have less environmental pollution. Therefore, it is urgent to improve the operation and maintenance effect of photovoltaic panels and improve the power generation efficiency of photovoltaic panels through intelligent equipment and digital means. Summary of the invention

[0003] In view of the deficiencies in the prior art, the object of the present invention is to provide a photovoltaic panel cleaner and a control method thereof which has high cleaning efficiency, good cleaning effect, high degree of automation and can monitor the status of the photovoltaic panel.

[0004] To achieve the above-mentioned object, the present invention provides the following technical solutions: A photovoltaic panel cleaner, comprising a cleaning robot: used for cleaning the surface of a photovoltaic panel;

[0005] Micro weather station: used to obtain real-time environmental indicators to evaluate power generation efficiency and cleanliness of photovoltaic panels;

[0006] Robot control system: used to adaptively adjust the cleaning robot to improve cleaning efficiency;

[0007] Monitoring system: used to monitor photovoltaic panels and improve power generation efficiency.

[0008] The present invention is further configured as follows: the cleaning robot includes at least two groups of roller brushes, which are arranged on both sides of the cleaning robot and distributed along the forward direction of the cleaning robot, the roller brush includes bristles and a rotating rod, the bristles include a first brush, a second brush and a third brush, and the length of the third brush is less than the length of the second brush and less than the length of the first brush, a pressure sensor is also provided on the top of the roller brush, and the pressure sensor is communicatively connected to the robot control system, and a driving mechanism is also provided inside the cleaning robot, and the driving mechanism includes a forward driving mechanism and a side shifting driving mechanism, which is used to drive the cleaning robot to move on the photovoltaic panel.

[0009] The present invention is further configured as follows: the micro weather station includes a plurality of sensors, which are communicatively connected to the robot control system for detecting environmental data and comparing the environmental data with the working data of the photovoltaic panels to determine whether the photovoltaic panels need additional cleaning.

[0010] The present invention is further configured as follows: the detection method of the photovoltaic panel cleaning device comprises the following steps:

[0011] S1-1, the micro-weather station obtains the current environmental indicators of the photovoltaic panels;

[0012] S1-2, the micro-weather station calculates the standard power generation data of the photovoltaic panel according to the light data within the current environmental indicators;

[0013] S1-3. According to the standard power generation data, the power generation is divided into the normal upper limit K MAX K MIN ;

[0014] S1-4, detect the power generation K of the photovoltaic panel and compare the power generation with the standard power generation data. If K MIN ≤K≤K MAX , it is judged that the power generation is normal, and the cleaning robot cleans the photovoltaic panels according to the cycle. Otherwise, it jumps to step S1-5 to continue the judgment;

[0015] S1-5, if K<K MIN , then the photovoltaic panel power generation efficiency is judged to be too low, and the cleaning robot performs additional cleaning tasks on the photovoltaic panel. If K>K MAX , the sensor is judged to be abnormal and the staff is reminded to repair it.

[0016] The present invention is further configured as follows: the cleaning robot control method comprises the following steps:

[0017] S2-1, data collection, collecting cleaning data of photovoltaic panels through passive sensor terminals;

[0018] S2-2, data reception and preprocessing, processing and cleaning the collected data;

[0019] S2-3, feature extraction, converting the processed data into feature vectors for subsequent neural network training and prediction;

[0020] S2-4. Neural network model design, design a multi-layer neural network model;

[0021] S2-5, model training, use the labeled data to train the neural network model and optimize the model;

[0022] S2-6, model application, applying the trained neural network model to the control of the cleaning robot to achieve adaptive adjustment of the cleaning robot.

[0023] Preferably, the passive sensor in step S2-1 includes an energy collection management module, an energy storage unit, a wireless radio frequency module, a sensor module and a radio frequency antenna, the radio frequency antenna includes a collection antenna and a receiving and transmitting antenna, and the working method of the passive sensor includes the following steps:

[0024] S2-1-1, collecting radio frequency signals through radio frequency antennas, converting the radio frequency signals into direct current, and storing them in energy storage units;

[0025] S2-1-2, detect the voltage of the energy storage unit. If the detected voltage value reaches the set threshold, the energy storage unit discharges to provide working current for the wireless RF module, and jumps back to step S2-1-1 to perform energy storage again;

[0026] Preferably, the neural network model in step S2-4 includes an input layer, a hidden layer and an output layer, the input layer is used to receive preprocessed data, the hidden layer is used to process and extract data from the input layer, and the output layer is used to output control instructions for the cleaning robot.

[0027] Preferably, the control method of the cleaning robot further includes cloud sensors, and the cloud sensors include current and voltage sensors, wind volume sensors, rainfall sensors, wind pressure sensors, light sensors and photovoltaic panel angle sensors, and the sensors are connected via RF radio frequency communication.

[0028] The present invention is further configured as follows: the monitoring system includes an optical camera, and the optical camera is used to monitor hot spots on photovoltaic panels, including the following steps:

[0029] S3-1, setting the hot spot area threshold on the photovoltaic panel to S, the optical camera detects the hot spot area on the photovoltaic panel, and the detection result is S1;

[0030] S3-2, compare S1 with S. If S1>S, the photovoltaic panel is judged to be faulty, and a fault work order is generated and sent to the manual end. Otherwise, no work order is generated and stored as a tracking task.

[0031] S3-3. Start the photovoltaic drone to inspect the photovoltaic panels in the area according to the stored tracking task. The detected hot spot area is S2. The expansion rate of the hot spot is calculated based on S2 and S1. If the calculation result is greater than S, a fault work order is generated. Otherwise, no work order is generated. The inspection frequency of the drone is calculated based on the expansion rate.

[0032] Preferably, the calculation formula of the hot spot expansion rate and the inspection frequency in step S3-3 is as follows:

[0033] H=(Au1 / Au0-Au2 / Au0) / X

[0034] X=T0×[1-(Au1 / Au0-Au2 / Au0)×e(Au1-Au2) / Au0-1]

[0035] Among them, X is the inspection time interval, H is the expansion rate of the hot spot, Au0 is the area of ​​a single photovoltaic panel, Au1 is the hot spot area of ​​the photovoltaic panel that was initially identified but not processed, and Au2 is the hot spot area of ​​the photovoltaic panel after the most recent inspection.

[0036] By adopting the above technical solution, the micro-weather station monitors the environment around the photovoltaic panels and detects the power generation efficiency of the photovoltaic panels through the control system. On the basis of the cleaning robot cleaning the photovoltaic panels at a fixed frequency, the number of cleanings is increased according to the changes in the power generation efficiency of the photovoltaic panels. While ensuring the power generation efficiency of the photovoltaic panels, it can also prevent the waste of resources caused by too frequent cleaning. At the same time, the monitoring system can replace manual confirmation of the cleaning effect after the cleaning robot completes cleaning, so that the cleaning effect can be checked and confirmed without anyone present, and the detection process can be shortened.

[0037] Furthermore, in order to improve the cleaning effect of the cleaning robot on the surface of the photovoltaic panel, roller brushes are symmetrically arranged on both sides of the cleaning robot and distributed along the forward direction of the cleaning robot. The roller brush includes bristles and a rotating rod. The bristles include a first brush, a second brush and a third brush, and the length of the third brush is less than the length of the second brush and less than the length of the first brush. Generally speaking, under the same material, the shorter the length of the brush, the harder the hardness of the brush, and the longer the length of the brush, the softer the hardness of the brush. The bristles are arranged in three lengths, so that the bristles form different hardness levels. The photovoltaic panel is cleaned by the contact between the brushes of different hardness and the photovoltaic panel. A pressure sensor is also provided on the top of the roller brush. The pressure sensor is communicatively connected to the robot control system. A driving mechanism is also provided inside the cleaning robot. The driving mechanism includes a forward driving mechanism and a side shifting driving mechanism for driving the brush. A mobile cleaning robot moves on the photovoltaic panel, the positive direction of the cleaning robot's forward direction is set as the front end of the cleaning robot, and the negative direction of the cleaning robot's forward direction is set as the rear end of the cleaning robot, the roller brush arranged at the front end of the cleaning robot is located in the area to be cleaned, and the covering is cleaned intensively, and the roller brush arranged at the rear end of the cleaning robot is located in the cleaned area, and the surface of the photovoltaic panel is cleaned for the second time, and the traces left by the forward driving mechanism of the cleaning robot during its forward movement are removed to prevent secondary pollution, and the contact pressure between the roller brush and the photovoltaic panel is detected by a pressure sensor on the top of the roller brush, and the pressure sensor is communicated with the robot control system, so as to realize different cleaning modes for different types of coverings, and the real-time perception of the contact pressure between the roller brush and the photovoltaic panel is completed through the cooperation of the mechanical structure and the sensing module, so as to realize the real-time control of the cleaning force by the cleaning robot, which not only ensures the cleaning effect but also prevents damage to the photovoltaic panel.

[0038] At the same time, the micro-weather station includes a number of sensors, which are connected to the robot control system for communication and are used to detect environmental data and compare the environmental data with the working data of the photovoltaic panels to determine whether the photovoltaic panels need additional cleaning. The sensors include but are not limited to wind volume sensors, rainfall sensors, wind pressure sensors, and light sensors. The micro-weather station is connected to the system to obtain real-time current environmental indicators to evaluate the power generation efficiency and the cleanliness of the photovoltaic panels in the system. The power generation value of the light data obtained by the micro-weather station under the current light intensity is used to make an intelligent judgment on whether the photovoltaic panels are normal and whether they need to be cleaned, including the following steps:

[0039] S1-1, the micro-weather station obtains the current environmental indicators of the photovoltaic panels;

[0040] S1-2, the micro-weather station calculates the standard power generation data of the photovoltaic panel according to the light data within the current environmental indicators;

[0041] S1-3. According to the standard power generation data, the power generation is divided into the normal upper limit K MAX K MIN ;

[0042] S1-4, detect the power generation K of the photovoltaic panel and compare the power generation with the standard power generation data. If K MIN ≤K≤K MAX , it is judged that the power generation is normal, and the cleaning robot cleans the photovoltaic panels according to the cycle. Otherwise, it jumps to step S1-5 to continue the judgment;

[0043] S1-5, if K<K MIN , then the photovoltaic panel power generation efficiency is judged to be too low, and the cleaning robot performs additional cleaning tasks on the photovoltaic panel. If K>K MAX , the sensor is judged to be abnormal and the staff is reminded to repair it.

[0044] Through the above steps, the optimal cleaning frequency algorithm can be intelligently determined based on real-time data and preset indicators, ensuring that the photovoltaic panels always operate in the best condition. While ensuring the cleanliness of the photovoltaic panels, it avoids excessive or insufficient maintenance and prevents waste of resources, which helps to reduce long-term maintenance and operating costs and extend equipment life.

[0045] In order to control the cleaning robot, a control algorithm of the robot is set in the robot control system, including the following steps:

[0046] S2-1, data collection, collecting cleaning data of photovoltaic panels through passive sensor terminals;

[0047] S2-2, data reception and preprocessing, processing and cleaning the collected data;

[0048] S2-3, feature extraction, converting the processed data into feature vectors for subsequent neural network training and prediction;

[0049] S2-4. Neural network model design, design a multi-layer neural network model;

[0050] S2-5, model training, use the labeled data to train the neural network model and optimize the model;

[0051] S2-6, model application, applying the trained neural network model to the control of the cleaning robot to achieve adaptive adjustment of the cleaning robot.

[0052] Specifically, the robot's motion trajectory is adaptively adjusted using a trained model, so that the robot can better adapt to the cleaning needs in different environments, thereby improving the cleaning efficiency and cleaning accuracy. The trained neural network model will be applied to the control of the actual photovoltaic cleaning robot. This includes adaptive adjustment of the robot's motion trajectory, cleaning speed and other parameters. Through the application of the model, the robot can intelligently select the best cleaning strategy according to the current environmental conditions and task requirements, thereby improving the cleaning efficiency and cleaning accuracy. At the same time, the passive sensor in step S2-1 includes an energy collection management module, an energy storage unit, a wireless radio frequency module, a sensor module and a radio frequency antenna. The radio frequency antenna includes a collecting antenna and a receiving and transmitting antenna. The working method of the passive sensor includes the following steps:

[0053] S2-1-1, collecting radio frequency signals through radio frequency antennas, converting the radio frequency signals into direct current, and storing them in energy storage units;

[0054] S2-1-2, detect the voltage of the energy storage unit. If the detected voltage value reaches the set threshold, the energy storage unit discharges to provide working current for the wireless RF module, and jumps back to step S2-1-1 to perform energy storage again;

[0055] At the same time, the neural network model in step S2-4 includes an input layer, a hidden layer and an output layer. The input layer is used to receive preprocessed data, the hidden layer is used to process and extract the data of the input layer, so as to realize adaptive control of the robot's motion state and environmental changes, and the output layer is used to output control instructions for the cleaning robot. The activation function of each neuron adopts the ReLU function, and the neurons of the hidden layer and the output layer are connected using the tanh function.

[0056] In addition, the cloud-based sensors include current and voltage sensors, wind volume sensors, rainfall sensors, wind pressure sensors, light sensors and photovoltaic panel angle sensors. The sensors are connected to each other via RF radio frequency communication. The background receives current and voltage data collected by the current and voltage sensors, and reads the light data received simultaneously, converts the current and voltage data into current, voltage and power data, and associates the current, voltage and power data with the light data for storage; the operation and maintenance platform receives wind volume data, and reads the wind pressure data and tilt data received simultaneously, and associates the wind volume data with the wind pressure data and tilt data for storage; the operation and maintenance platform receives rainfall data, and reads the tilt data received simultaneously, and associates the rainfall data with the tilt data for storage.

[0057] By adopting the above control scheme, the cleaning robot can automatically clean and inspect without human intervention, saving human resources. At the same time, it can provide continuous data feedback through regular inspection of the cleaning robot, which helps to detect problems in time. It can also replace personnel to work in high temperature and high pressure environments, greatly improving safety.

[0058] In order to monitor the status of the photovoltaic panel, an optical camera is connected to the monitoring system to monitor the hot spots of the photovoltaic panel, including the following steps:

[0059] S3-1, setting the hot spot area threshold on the photovoltaic panel to S, the optical camera detects the hot spot area on the photovoltaic panel, and the detection result is S1;

[0060] S3-2, compare S1 with S. If S1>S, the photovoltaic panel is judged to be faulty, and a fault work order is generated and sent to the manual end. Otherwise, no work order is generated and stored as a tracking task.

[0061] S3-3. Start the photovoltaic drone to inspect the photovoltaic panels in the area according to the stored tracking task. The detected hot spot area is S2. The expansion rate of the hot spot is calculated based on S2 and S1. If the calculation result is greater than S, a fault work order is generated. Otherwise, no work order is generated. The inspection frequency of the drone is calculated based on the expansion rate.

[0062] In order to better complete the hot spot monitoring, conduct fault statistical analysis and formulate a reasonable inspection frequency to achieve a balance between inspection input and output, the inspection frequency is fed back through the hot spot expansion rate. The larger the expansion rate, the higher the frequency of regular autonomous inspections by drones, and the shorter the inspection interval. The calculation formula of expansion rate and inspection frequency is as follows:

[0063] H=(Au1 / Au0-Au2 / Au0) / X

[0064] X=T0×[1-(Au1 / Au0-Au2 / Au0)×e(Au1-Au2) / Au0-1]

[0065] Among them, X is the inspection time interval, H is the expansion rate of the hot spot, Au0 is the area of ​​a single photovoltaic panel, Au1 is the hot spot area of ​​the photovoltaic panel that was initially identified but not processed, and Au2 is the hot spot area of ​​the photovoltaic panel after the most recent inspection.

[0066] Preferably, in step S3-1, since the power generation of the photovoltaic panel will decay with age, the normal decay of power generation will affect the abnormal judgment, and the threshold curve is further corrected based on the age value, and the correction formula is as follows:

[0067] d=(1-ηt / η0)×100% / t

[0068] Among them, d is the annual attenuation rate of the photovoltaic panel, that is, the photovoltaic panel power generation efficiency attenuation rate refers to the rate at which the power generation efficiency of the photovoltaic panel decays after a certain period of use, t is the design service life of the photovoltaic panel, η0 is the initial power generation efficiency of the photovoltaic panel, and ηt is the current power generation efficiency of the photovoltaic panel. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A side view of a cleaning robot according to an embodiment of a photovoltaic panel cleaner and a control method thereof of the present invention;

[0070] Figure 2 A diagram of the hair implantation structure of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0071] Figure 3 A cleaning mode control diagram of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0072] Figure 4 A flow chart of the detection of photovoltaic panels by a micro-weather station according to an embodiment of a photovoltaic panel cleaner and a control method thereof of the present invention;

[0073] Figure 5 A coordinate diagram of illumination and power generation of a photovoltaic panel cleaner and a control method thereof according to an embodiment of the present invention;

[0074] Figure 6 A cleaning robot control flow chart of a photovoltaic panel cleaner and a control method thereof according to an embodiment of the present invention;

[0075] Figure 7 A cloud sensor working diagram of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0076] Figure 8 It is a working diagram of an energy storage unit of an embodiment of a photovoltaic panel cleaner and a control method thereof of the present invention;

[0077] Fig. 9 A multi-layer neural network structure diagram of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0078] Fig.10 A flow chart of hot spot detection of photovoltaic panels by a monitoring system of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0079] Fig.11 A power generation threshold correction curve diagram of an embodiment of a photovoltaic panel cleaner and a control method thereof according to the present invention;

[0080] The reference numerals in the figure are: 1, roller brush; 11, hair planting; 12, rotating rod; 2, first brush; 3, second brush; 4, third brush; 5, driving mechanism. DETAILED DESCRIPTION

[0081] Reference Figures 1 to 11 A photovoltaic panel cleaner and a control method thereof according to an embodiment of the present invention are further described.

[0082] For ease of explanation, spatial relative terms such as "upper", "lower", "left", "right" and the like are used in the embodiments to illustrate the relationship of one element or feature shown in the figure relative to another element or feature. It should be understood that, in addition to the orientation shown in the figure, the spatial terms are intended to include different orientations of the device in use or operation. For example, if the device in the figure is inverted, the element described as being "under" other elements or features will be positioned "on" other elements or features. Therefore, the exemplary term "under" can include both upper and lower orientations. The device can be positioned in other ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used here can be interpreted accordingly.

[0083] Furthermore, relational terms such as “first” and “second” and the like are merely used to distinguish one component from another having the same name, but do not necessarily require or imply any such actual relationship or order between these components.

[0084] A photovoltaic panel cleaner comprises a cleaning robot: used for cleaning the surface of a photovoltaic panel;

[0085] Micro weather station: used to obtain real-time environmental indicators to evaluate power generation efficiency and cleanliness of photovoltaic panels;

[0086] Robot control system: used to adaptively adjust the cleaning robot to improve cleaning efficiency;

[0087] Monitoring system: used to monitor photovoltaic panels and improve the efficiency of power generation. The present invention monitors the environment around the photovoltaic panels through a micro-weather station and detects the power generation efficiency of the photovoltaic panels through a control system. On the basis of the cleaning robot cleaning the photovoltaic panels at a fixed frequency, the number of cleanings is increased according to the changes in the power generation efficiency of the photovoltaic panels. While ensuring the power generation efficiency of the photovoltaic panels, it can also prevent the waste of resources caused by too frequent cleaning. At the same time, the monitoring system can replace manual confirmation of the cleaning effect after the cleaning robot completes cleaning, thereby realizing the inspection and confirmation of the cleaning effect without anyone present, and can shorten the detection process.

[0088] Furthermore, in order to improve the cleaning effect of the cleaning robot on the surface of the photovoltaic panel, a roller brush 1 is symmetrically arranged on both sides of the cleaning robot and distributed along the forward direction of the cleaning robot. The roller brush 1 includes a hair transplant 11 and a rotating rod 12. The hair transplant 11 includes a first brush 2, a second brush 3 and a third brush 4, and the length of the third brush 4 is less than the length of the second brush 3 and less than the length of the first brush 2. Generally speaking, under the same material, the shorter the length of the brush, the harder the hardness of the brush is, and the longer the length of the brush, the softer the hardness of the brush is. The hair transplant 11 forms different hardness levels by setting brushes of three lengths. The photovoltaic panel is cleaned by the contact between the brushes of different hardness and the photovoltaic panel. A pressure sensor is also provided on the top of the roller brush 1. The pressure sensor is communicated with the robot control system. A driving mechanism 5 is also provided inside the cleaning robot. The driving mechanism 5 includes a forward driving mechanism 5 and a side shift driving mechanism 5. The driving mechanism 5 is used to drive the cleaning robot to move on the photovoltaic panel. The positive direction of the forward direction of the cleaning robot is set as the front end of the cleaning robot, and the negative direction of the forward direction of the cleaning robot is set as the rear end of the cleaning robot. The roller brush 1 arranged at the front end of the cleaning robot is located in the area to be cleaned, and the covering is cleaned intensively. The roller brush 1 arranged at the rear end of the cleaning robot is located in the cleaned area, and the surface of the photovoltaic panel is cleaned for a second time to remove the traces left by the forward driving mechanism 5 during the forward movement of the cleaning robot to prevent secondary pollution. The contact pressure between the roller brush 1 and the photovoltaic panel is detected by a pressure sensor on the top of the roller brush 1. The pressure sensor is communicated with the robot control system to realize different cleaning modes for different types of coverings. The real-time perception of the contact pressure between the roller brush 1 and the photovoltaic panel is completed through the cooperation of the mechanical structure and the sensing module, so that the cleaning robot can control the cleaning force in real time, which not only ensures the cleaning effect but also prevents damage to the photovoltaic panel.

[0089] At the same time, the pressure sensor can also set a variety of cleaning modes according to the types of coverings on the photovoltaic panel. According to the hardness of the implanted bristles 11, it can be divided into light pressure brush, medium pressure brush and heavy pressure brush. The light pressure brush refers to the cleaning mode when the first bristles are in contact with the photovoltaic panel, and the contact depth of the roller brush 1 with the photovoltaic panel is small, which is used to clean dust and other particles on the photovoltaic panel. At this time, the pressure data detected by the pressure sensor is small. The medium pressure brush refers to the cleaning mode when the second bristles are in initial contact with the photovoltaic panel, and the contact depth of the roller brush 1 with the photovoltaic panel is medium, which is used to clean large-volume and non-sticky coverings such as leaves on the photovoltaic panel. At this time, the pressure data detected by the pressure sensor is medium. The heavy pressure brush refers to the cleaning mode in which the second bristles are in full contact with the photovoltaic panel, and the contact depth of the roller brush 1 with the photovoltaic panel is large, which is used to effectively clean coverings such as bird droppings on the photovoltaic panel. The cleaning modes of the medium pressure brush and the heavy pressure brush can achieve the purpose of fixed-point cleaning of the cleaning robot. Preferably, the pressure sensor can further divide the cleaning mode into high-speed cleaning mode and low-speed cleaning mode according to the rotation speed of the roller brush 1.

[0090] Furthermore, the photovoltaic panels are generally tilted to better receive light. When the cleaning robot is running, it has a tendency to slide down along the photovoltaic panels due to the effect of gravity. Through the high-speed rotation of the rotating rod 12 and the implanted hair 11, a longitudinal force opposite to the sliding direction of the cleaning robot is formed, thereby forming a unified longitudinal lift. The longitudinal lift and gravity achieve a balance effect during their interaction, effectively preventing the cleaning robot from sliding down and self-locking.

[0091] At the same time, the micro-weather station includes a number of sensors, which are connected to the robot control system for communication and are used to detect environmental data and compare the environmental data with the working data of the photovoltaic panels to determine whether the photovoltaic panels need additional cleaning. The sensors include but are not limited to wind volume sensors, rainfall sensors, wind pressure sensors, and light sensors. The micro-weather station is connected to the system to obtain real-time current environmental indicators to evaluate the power generation efficiency and the cleanliness of the photovoltaic panels in the system. The power generation value of the light data obtained by the micro-weather station under the current light intensity is used to make an intelligent judgment on whether the photovoltaic panels are normal and whether they need to be cleaned, including the following steps:

[0092] S1-1, the micro-weather station obtains the current environmental indicators of the photovoltaic panels;

[0093] S1-2, the micro-weather station calculates the standard power generation data of the photovoltaic panel according to the light data within the current environmental indicators;

[0094] S1-3. According to the standard power generation data, the power generation is divided into the normal upper limit K MAX K MIN ;

[0095] S1-4, detect the power generation K of the photovoltaic panel and compare the power generation with the standard power generation data. If K MIN ≤K≤K MAX , it is judged that the power generation is normal, and the cleaning robot cleans the photovoltaic panels according to the cycle. Otherwise, it jumps to step S1-5 to continue the judgment;

[0096] S1-5, if K<K MIN , then the photovoltaic panel power generation efficiency is judged to be too low, and the cleaning robot performs additional cleaning tasks on the photovoltaic panel. If K>K MAX , the sensor is judged to be abnormal and the staff is reminded to repair it.

[0097] Through the above steps, the optimal cleaning frequency algorithm can be intelligently determined based on real-time data and preset indicators, ensuring that the photovoltaic panels always operate in the best condition. While ensuring the cleanliness of the photovoltaic panels, it avoids excessive or insufficient maintenance and prevents waste of resources, which helps to reduce long-term maintenance and operating costs and extend equipment life.

[0098] Refer to the figure

[0099] In order to control the cleaning robot, a control algorithm of the robot is set in the robot control system, including the following steps:

[0100] S2-1, data collection, collecting cleaning data of photovoltaic panels through passive sensor terminals;

[0101] S2-2, data reception and preprocessing, processing and cleaning the collected data;

[0102] S2-3, feature extraction, converting the processed data into feature vectors for subsequent neural network training and prediction;

[0103] S2-4. Neural network model design, design a multi-layer neural network model;

[0104] S2-5, model training, use the labeled data to train the neural network model and optimize the model;

[0105] S2-6, model application, applying the trained neural network model to the control of the cleaning robot to achieve adaptive adjustment of the cleaning robot.

[0106] Specifically, the robot's motion trajectory is adaptively adjusted using a trained model, so that the robot can better adapt to the cleaning needs in different environments, thereby improving the cleaning efficiency and cleaning accuracy. The trained neural network model will be applied to the control of the actual photovoltaic cleaning robot. This includes adaptive adjustment of the robot's motion trajectory, cleaning speed and other parameters. Through the application of the model, the robot can intelligently select the best cleaning strategy according to the current environmental conditions and task requirements, thereby improving the cleaning efficiency and cleaning accuracy. At the same time, the passive sensor in step S2-1 includes an energy collection management module, an energy storage unit, a wireless radio frequency module, a sensor module and a radio frequency antenna. The radio frequency antenna includes a collecting antenna and a receiving and transmitting antenna. The working method of the passive sensor includes the following steps:

[0107] S2-1-1, collecting radio frequency signals through radio frequency antennas, converting the radio frequency signals into direct current, and storing them in energy storage units;

[0108] S2-1-2, detect the voltage of the energy storage unit. If the detected voltage value reaches the set threshold, the energy storage unit discharges to provide working current for the wireless RF module, and jumps back to step S2-1-1 to perform energy storage again;

[0109] The above energy storage unit is used to store electricity and can release electricity quickly. According to the needs, supercapacitors of different capacities are generally used as energy storage units. Supercapacitors rely on double-layer and redox pseudocapacitor charges to store electrical energy, and the energy storage process does not produce chemical reactions. This energy storage process is reversible and can be repeatedly charged and discharged hundreds of thousands of times.

[0110] At the same time, the neural network model in step S2-4 includes an input layer, a hidden layer and an output layer. The input layer is used to receive preprocessed data, the hidden layer is used to process and extract the data of the input layer, so as to realize adaptive control of the robot's motion state and environmental changes, and the output layer is used to output control instructions for the cleaning robot. The activation function of each neuron adopts the ReLU function, and the neurons of the hidden layer and the output layer are connected using the tanh function.

[0111] The multi-layer neural network input layer to output layer is forward propagation, and the formula for forward propagation is as follows:

[0112] a (l) =σ(z (l) )=σ(w (l) a (l-1) +b (l) )

[0113] Among them, a is the activation value, w is the weight, b is the bias, and w (l) a(l-1) +b (l) The weight input for each layer, the activation value of each input layer is the feature vector of the input of that layer.

[0114] The basic framework of the cleaning robot controller is:

[0115] Input: state quantity x, expected output r, error signal e.

[0116] Processing: Calculate the control signal u=f(x,r,e).

[0117] Output: control signal u.

[0118] Wherein, f is the adaptive algorithm of the controller, and the adaptive algorithm is the forward propagation process of the neural network, and the formula is as follows:

[0119] u=f(x,r,e)=σ(w (L) a (L-1) +b (L) )

[0120] Where L is the number of layers in the neural network, w (L) is the output layer weight of the neural network, b (L) is the output layer bias, a (L-1) is the activation value of the input layer.

[0121] Furthermore, in the model training, specifically, cross entropy is used as the loss function, and the Adam optimizer is used to update the model parameters. Commonly used loss functions include mean square error (MSE), among which MSE is the most commonly used and is defined as:

[0122]

[0123] Among them, yi is the output of the neural network, ti ​​is the expected output, and N is the number of data samples.

[0124] At the same time, in order to calculate the gradients of weights and biases in the neural network, the back propagation algorithm is used so that the staff can minimize the loss function. The basic formula of the back propagation algorithm is as follows:

[0125]

[0126] δ (l) =((w (l+1) ) T δ (l+1) )⊙σ'(z (l) )

[0127]

[0128] Among them, L is the loss function, a is the activation function, z is the weighted sum, w is the weight, b is the bias, is the error signal.

[0129] In addition, the cloud-based sensors include current and voltage sensors, wind volume sensors, rainfall sensors, wind pressure sensors, light sensors and photovoltaic panel angle sensors. The sensors are connected to each other via RF radio frequency communication. The background receives current and voltage data collected by the current and voltage sensors, and reads the light data received simultaneously, converts the current and voltage data into current, voltage and power data, and associates the current, voltage and power data with the light data for storage; the operation and maintenance platform receives wind volume data, and reads the wind pressure data and tilt data received simultaneously, and associates the wind volume data with the wind pressure data and tilt data for storage; the operation and maintenance platform receives rainfall data, and reads the tilt data received simultaneously, and associates the rainfall data with the tilt data for storage.

[0130] By adopting the above control scheme, the cleaning robot can automatically clean and inspect without human intervention, saving human resources. At the same time, it can provide continuous data feedback through regular inspection of the cleaning robot, which helps to detect problems in time. It can also replace personnel to work in high temperature and high pressure environments, greatly improving safety.

[0131] In the process of applying the above model, the trained neural network model is applied to the control of the actual photovoltaic cleaning robot to achieve adaptive adjustment of the robot's motion trajectory, cleaning speed and other parameters. Specifically, the robot's motion trajectory is adaptively adjusted using the trained model, so that the robot can better adapt to the cleaning needs in different environments, thereby improving the cleaning efficiency and cleaning accuracy.

[0132] In order to monitor the status of the photovoltaic panel, an optical camera is connected to the monitoring system to monitor the hot spots of the photovoltaic panel, including the following steps:

[0133] S3-1, setting the hot spot area threshold on the photovoltaic panel to S, the optical camera detects the hot spot area on the photovoltaic panel, and the detection result is S1;

[0134] S3-2, compare S1 with S. If S1>S, the photovoltaic panel is judged to be faulty, and a fault work order is generated and sent to the manual end. Otherwise, no work order is generated and stored as a tracking task.

[0135] S3-3. Start the photovoltaic drone to inspect the photovoltaic panels in the area according to the stored tracking task. The detected hot spot area is S2. The expansion rate of the hot spot is calculated based on S2 and S1. If the calculation result is greater than S, a fault work order is generated. Otherwise, no work order is generated. The inspection frequency of the drone is calculated based on the expansion rate.

[0136] In order to better complete the hot spot monitoring, conduct fault statistical analysis and formulate a reasonable inspection frequency to achieve a balance between inspection input and output, the inspection frequency is fed back through the hot spot expansion rate. The larger the expansion rate, the higher the frequency of regular autonomous inspections by drones, and the shorter the inspection interval. The calculation formula of expansion rate and inspection frequency is as follows:

[0137] H=(Au1 / Au0-Au2 / Au0) / X

[0138] X=T0×[1-(Au1 / Au0-Au2 / Au0)×e(Au1-Au2) / Au0-1]

[0139] Among them, X is the inspection time interval, H is the expansion rate of the hot spot, Au0 is the area of ​​a single photovoltaic panel, Au1 is the hot spot area of ​​the photovoltaic panel that was initially identified but not processed, and Au2 is the hot spot area of ​​the photovoltaic panel after the most recent inspection.

[0140] Preferably, in step S3-1, since the power generation of the photovoltaic panel will decay with age, the normal decay of power generation will affect the abnormal judgment, and the threshold curve is further corrected based on the age value, and the correction formula is as follows:

[0141] d=(1-ηt / η0)×100% / t

[0142] Among them, d is the annual attenuation rate of the photovoltaic panel, that is, the photovoltaic panel power generation efficiency attenuation rate refers to the rate at which the power generation efficiency of the photovoltaic panel decays after a certain period of use, t is the design service life of the photovoltaic panel, η0 is the initial power generation efficiency of the photovoltaic panel, and ηt is the current power generation efficiency of the photovoltaic panel.

[0143] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Common changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic panel cleaner, characterized in that: Including cleaning robots: used to clean the surface of photovoltaic panels; Micro weather station: used to obtain real-time environmental indicators to evaluate power generation efficiency and cleanliness of photovoltaic panels; Robot control system: used to adaptively adjust the cleaning robot to improve cleaning efficiency; Monitoring system: used to monitor photovoltaic panels and improve power generation efficiency.

2. A photovoltaic panel cleaner according to claim 1, characterized in that: The cleaning robot comprises at least two groups of roller brushes (1), the roller brushes (1) are arranged on both sides of the cleaning robot and distributed along the forward direction of the cleaning robot, the roller brush (1) comprises a hair graft (11) and a rotating rod (12), the hair graft (11) comprises a first brush (2), a second brush (3) and a third brush (4), and the length of the third brush (4) is less than the length of the second brush (3) and less than the length of the first brush (2), a pressure sensor is also arranged on the top of the roller brush (1), and the pressure sensor is communicatively connected with the robot control system, and a driving mechanism (5) is also arranged inside the cleaning robot, and the driving mechanism (5) comprises a forward driving mechanism (5) and a side shift driving mechanism (5), which is used to drive the cleaning robot to move on the photovoltaic panel.

3. A photovoltaic panel cleaner according to claim 1, characterized in that: The micro weather station includes a plurality of sensors, which are communicatively connected with the robot control system and are used to detect environmental data and compare the environmental data with the working data of the photovoltaic panels to determine whether the photovoltaic panels need additional cleaning.

4. A detection method for a photovoltaic panel cleaning device according to claim 3, characterized in that: The following steps are involved: S1-1, the micro-weather station obtains the current environmental indicators of the photovoltaic panels; S1-2, the micro-weather station calculates the standard power generation data of the photovoltaic panel according to the light data within the current environmental indicators; S1-3. According to the standard power generation data, the power generation is divided into the normal upper limit K MAX K MIN ; S1-4, detect the power generation K of the photovoltaic panel and compare the power generation with the standard power generation data. If K MIN ≤K≤K MAX , it is judged that the power generation is normal, and the cleaning robot cleans the photovoltaic panels according to the cycle. Otherwise, it jumps to step S1-5 to continue the judgment; S1-5, if K<K MIN , then the photovoltaic panel power generation efficiency is judged to be too low, and the cleaning robot performs additional cleaning tasks on the photovoltaic panel. If K>K MAX , the sensor is judged to be abnormal and the staff is reminded to repair it.

5. A cleaning robot control method according to claim 1, characterized in that: The following steps are involved: S2-1, data collection, collecting cleaning data of photovoltaic panels through passive sensor terminals; S2-2, data reception and preprocessing, processing and cleaning the collected data; S2-3, feature extraction, converting the processed data into feature vectors for subsequent neural network training and prediction; S2-4. Neural network model design, design a multi-layer neural network model; S2-5, model training, use the labeled data to train the neural network model and optimize the model; S2-6, model application, applying the trained neural network model to the control of the cleaning robot to achieve adaptive adjustment of the cleaning robot.

6. The control method of the cleaning robot according to claim 5, characterized in that: The passive sensor in step S2-1 includes an energy collection management module, an energy storage unit, a wireless radio frequency module, a sensor module and a radio frequency antenna, wherein the radio frequency antenna includes a collection antenna and a receiving and transmitting antenna. The working method of the passive sensor includes the following steps: S2-1-1, collecting radio frequency signals through radio frequency antennas, converting the radio frequency signals into direct current, and storing them in energy storage units; S2-1-2, detect the voltage of the energy storage unit. If the detected voltage value reaches the set threshold, the energy storage unit discharges to provide working current for the wireless RF module, and jumps back to step S2-1-1 to perform energy storage again; 7. The control method of the cleaning robot according to claim 5, characterized in that: The neural network model in step S2-4 includes an input layer, a hidden layer and an output layer. The input layer is used to receive preprocessed data, the hidden layer is used to process and extract data from the input layer, and the output layer is used to output control instructions for the cleaning robot.

8. The control method of the cleaning robot according to claim 5, characterized in that: It also includes cloud sensors, which include current and voltage sensors, wind volume sensors, rainfall sensors, wind pressure sensors, light sensors, and photovoltaic panel angle sensors. The sensors are connected via RF radio frequency communication.

9. A photovoltaic panel cleaner according to claim 1, characterized in that: The monitoring system includes an optical camera, which is used to monitor hot spots on photovoltaic panels, and includes the following steps: S3-1, setting the hot spot area threshold on the photovoltaic panel to S, the optical camera detects the hot spot area on the photovoltaic panel, and the detection result is S1; S3-2, compare S1 with S. If S1>S, the photovoltaic panel is judged to be faulty, and a fault work order is generated and sent to the manual end. Otherwise, no work order is generated and stored as a tracking task. S3-3. Start the photovoltaic drone to inspect the photovoltaic panels in the area according to the stored tracking task. The detected hot spot area is S2. The expansion rate of the hot spot is calculated based on S2 and S1. If the calculation result is greater than S, a fault work order is generated. Otherwise, no work order is generated. The inspection frequency of the drone is calculated based on the expansion rate.

10. The photovoltaic panel hot spot monitoring method according to claim 9, characterized in that: The calculation formula of the hot spot expansion rate and the inspection frequency in step S3-3 is as follows: H=(Au1 / Au0-Au2 / Au0) / X X=T0×[1-(Au1 / Au0-Au2 / Au0)×e(Au1-Au2) / Au0-1] Among them, X is the inspection time interval, H is the expansion rate of the hot spot, Au0 is the area of ​​a single photovoltaic panel, Au1 is the hot spot area of ​​the photovoltaic panel that was initially identified but not processed, and Au2 is the hot spot area of ​​the photovoltaic panel after the most recent inspection.

Images