An unmanned aerial vehicle inspection method for monitoring the pollution state of a photovoltaic module

Abstract

The present application relates to the technical field of unmanned aerial vehicle, especially relates to a kind of unmanned aerial vehicle inspection method for monitoring the pollution state of photovoltaic module, including reading photovoltaic module position setting information from power grid data platform, satellite GPS generates monitoring target template, and generates initial image data according to monitoring target template and the inclination parameter of photovoltaic module setting, establishes at least one photovoltaic module three-dimensional rectangular coordinate system, generates inspection map according to photovoltaic module three-dimensional rectangular coordinate system, photovoltaic module three-dimensional rectangular coordinate system includes X axis, Y axis and H axis, at least one pollutant three-dimensional rectangular coordinate system, generates sub-inspection map according to pollutant three-dimensional rectangular coordinate system, and pollutant three-dimensional rectangular coordinate system includes X1 axis, Y1 axis and H1 axis.The present application has the advantages that the accuracy of pollutant collection data can be improved, the full life cycle tracking of photovoltaic module is realized, the processing priority rationality is improved, and the photovoltaic module light utilization rate is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicles, and particularly relates to a method for monitoring pollution states of photovoltaic components by unmanned aerial vehicle inspection. BACKGROUND

[0002] With the wide application of unmanned aerial vehicles in power inspection, the outdoor transmission and distribution line inspection by unmanned aerial vehicles has developed from manual inspection to automatic inspection with preset flight route planning.

[0003] However, the existing method is not applicable to some areas with high layout intensification, for example, in the process of monitoring a large number of photovoltaic components, the photovoltaic components are set to different inclinations due to different latitudes and altitudes and different solar illumination angles, and the photovoltaic components are contaminated by pollutants such as bird droppings, fallen leaves and floating dust, thereby reducing the light utilization rate. Therefore, the unmanned aerial vehicle needs to adopt different shooting angles for photovoltaic components with different inclinations to obtain clearer images of the contaminated pollutants, so as to analyze and dispatch maintenance orders.

[0004] In addition, for photovoltaic components with different inclinations, the planning of the unmanned aerial vehicle inspection path does not change, which will result in low inspection efficiency and inconvenient determination of the inspection order, and is easy to cause poor inspection effect and inconvenient rapid collection of inspection images. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the deficiencies of the prior art and provide a method for monitoring the pollution state of photovoltaic components by unmanned aerial vehicle inspection, which can improve the accuracy of pollution collection data, realize the full life cycle tracking of photovoltaic components, improve the rationality of processing priority, and ensure the light utilization rate of photovoltaic components.

[0006] To solve the above technical problems, the present application adopts the following technical solutions:

[0007] A method for monitoring the pollution state of photovoltaic components by unmanned aerial vehicle inspection, comprising

[0008] S101: generating a monitoring target template from the photovoltaic component position setting information read from the power grid data platform and satellite GPS, and generating initial image data according to the monitoring target template and the inclination parameter of the photovoltaic component setting;

[0009] S102: Establish at least one photovoltaic component three-dimensional rectangular coordinate system, generate an inspection map according to the photovoltaic component three-dimensional rectangular coordinate system, the photovoltaic component three-dimensional rectangular coordinate system includes X axis, Y axis and H axis, use several standard division areas to tile the photovoltaic component three-dimensional rectangular coordinate system, keep the long side of each standard division area parallel to the X axis, the width of each standard division area is parallel to the Y axis, and the height of each standard division area is parallel to the H axis;

[0010] S103: The inspection map is divided to obtain a plurality of sub-inspection areas, and the inspection path is set according to the plurality of sub-inspection areas after division, and further comprises: at least one pollution three-dimensional rectangular coordinate system, generate a sub-inspection map according to the pollution three-dimensional rectangular coordinate system, the pollution three-dimensional rectangular coordinate system includes X1 axis, Y1 axis and H1 axis, use several standard division areas to tile the pollution three-dimensional rectangular coordinate system, keep the long side of each standard division area parallel to the X1 axis, the width of each standard division area is parallel to the Y1 axis, and the height of each standard division area is parallel to the H1 axis;

[0011] S104: Sort a plurality of groups of sub-inspection areas according to the H axis coordinate from small to large, and / or sort a plurality of groups of sub-inspection areas according to the H1 axis coordinate from small to large, to obtain an inspection group sorting;

[0012] S105: Complete the unmanned aerial vehicle inspection operation preparation and unmanned aerial vehicle inspection operation analysis according to the inspection group sorting, and perform the unmanned aerial vehicle inspection operation; collect image data and perform data preprocessing; construct a corresponding inspection report according to the inspection data to perform operation and maintenance ordering.

[0013] As preferred, the photovoltaic component image data acquisition process includes: acquiring a plurality of sample images by using a real-time shooting method for X axis, Y axis and H axis coordinate axes;

[0014] Get the position of the photovoltaic component in each sample image, remove the blurred photovoltaic component position from the sample image according to the blurred photo, and then superimpose the remaining part of the sample image to obtain a superimposed image;

[0015] Determine whether the superimposed image is damaged, if so, instruct the unmanned aerial vehicle to acquire a plurality of sample images by using a real-time shooting method.

[0016] As preferred, the pollution image data acquisition process includes: acquiring a plurality of sample images by using a real-time shooting method for X1 axis, Y1 axis and H1 axis coordinate axes;

[0017] Get the position of the pollution in each sample image, remove the blurred pollution position from the sample image according to the blurred photo, and then superimpose the remaining part of the sample image to obtain a superimposed image;

[0018] If the superimposed image is determined to have a defect, the UAV is instructed to acquire a plurality of sample images by real-time shooting.

[0019] As a preferred embodiment, the method further comprises meteorological monitoring data, the meteorological monitoring data comprising meteorological observation monitoring data, meteorological satellite monitoring data and meteorological radar monitoring data, wherein the meteorological observation monitoring data comprises temperature data, humidity data, wind speed data and precipitation data, the meteorological satellite monitoring data comprises infrared cloud image data and visible light cloud image data, and the meteorological radar monitoring data comprises basic reflectivity data and precipitation data.

[0020] As a preferred embodiment, the method further comprises photovoltaic module placement monitoring, which is used to monitor whether the photovoltaic module has a position deviation; if so, the UAV is instructed to acquire new position information of the photovoltaic module by real-time shooting.

[0021] As a preferred embodiment, before generating the inspection report according to the image data, the method further comprises comparing, according to the meteorological monitoring data, the second image data with the initial image data, if the meteorological monitoring data is greater than a severe weather value, the UAV is instructed to acquire the second image data by real-time shooting.

[0022] As a preferred embodiment, the method further comprises, after the operation and maintenance order is completed, the UAV is instructed to acquire third image data by real-time shooting, and the third image data is compared with the initial image data.

[0023] As a preferred embodiment, the method further comprises constructing a data preprocessing model, and the data preprocessing model is trained by using a core algorithm, the core algorithm comprising: researching a basic sparse coding model of image information, researching an image sparse coding algorithm based on an optimized dictionary, and researching a convolutional neural network algorithm.

[0024] A computer readable storage medium, the computer readable storage medium comprising a stored program, wherein the program, when executed, controls a device in which the computer readable storage medium is located to perform the UAV inspection method for monitoring a pollution state of a photovoltaic module as described above.

[0025] A processor, the processor being configured to execute a program, wherein the program, when executed, performs the UAV inspection method for monitoring a pollution state of a photovoltaic module as described above.

[0026] The present application has the following advantages:

[0027] 1. In the operation and maintenance of photovoltaic components of photovoltaic power stations, the unmanned aerial vehicle is used to improve the inspection efficiency, efficiently handle the pollution of photovoltaic components, and improve the light utilization rate of photovoltaic components. The unmanned aerial vehicle has high inspection frequency, high system automation degree, and does not need human intervention, thereby improving the accuracy of judgment. The pollution has seasonality and independence. The probability of leaf blocking increases in autumn, and the monthly situation of bird droplet pollution caused by migration of birds and other reasons also changes. The whole life cycle of the photovoltaic component is tracked, and the efficiency of handling the pollution of the photovoltaic component is improved.

[0028] 2. Considering the influence factors of extreme weather, the photovoltaic components are timely maintained, the service life of the photovoltaic components is prolonged, and the cost is reduced.

[0029] 3. Different types of pollutants are identified and guided to clean the photovoltaic components. The large and high blocking rate need to be immediately maintained and processed. The dust pollution produced by the factory exhaust is affected by the monthly order and production process and fluctuates. It generally needs a long time to deposit and then clean. Accurate identification of the pollution degree and the main pollutants can enable the cleaning team to adopt a targeted cleaning scheme, such as using tools such as clamps for fallen leaves, tools such as spades and rags for bird droplets, and tools such as rags for dust. The cleaning efficiency is improved.

[0030] 4. The data processing priority of the photovoltaic component pollution is reasonable, the photovoltaic component pollution is efficiently cleaned, the cleanliness of the photovoltaic component is maintained, and the light utilization rate of the photovoltaic component is improved.

[0031] 5. The second convolutional neural network model is a convolutional neural network model with a residual mechanism. Through the residual mechanism, the learned features are more accurate during the training of the convolutional neural network model, and thus it is more suitable for performing complex identification tasks. It is worth emphasizing that the first convolutional neural network model is set as a general convolutional neural network model, and the second convolutional neural network model is a convolutional neural network model with a residual mechanism. Through such a setting, the dirt image in an image can be accurately identified through a simple model, and the dirt image in an image can be accurately identified through a complex model, thereby realizing the unity of efficiency and accuracy.

[0032] These features and advantages of the present application will be disclosed in detail in the following specific embodiments and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0033] The present application will be further described below in conjunction with the drawings:

[0034] Fig. 1 The flowchart of the present application is shown in the figure.

[0035] Fig. 2 The unmanned aerial vehicle inspection process is shown in the figure. DETAILED DESCRIPTION

[0036] In one embodiment of the present application, an unmanned aerial vehicle inspection method for monitoring the pollution state of a photovoltaic module, as shown in the accompanying drawings, comprises the following steps: Figs. 1-2

[0037] S101: Read the photovoltaic module position setting information from the power grid data platform and satellite GPS to generate an inspection target template, and generate initial image data according to the inspection target template and the inclination parameter of the photovoltaic module setting;

[0038] S102: Establish at least one photovoltaic module three-dimensional rectangular coordinate system, generate an inspection map according to the photovoltaic module three-dimensional rectangular coordinate system, the photovoltaic module three-dimensional rectangular coordinate system includes X axis, Y axis and H axis, use several division standard areas to tile the photovoltaic module three-dimensional rectangular coordinate system, keep the long side of each division standard area parallel to the X axis, the width of each division standard area parallel to the Y axis, and the height of each division standard area parallel to the H axis;

[0039] S1021: Each division standard area on the X axis, Y axis and H axis is aligned with the photovoltaic module one by one, using the X axis and Y axis as the main mark, the photovoltaic module images located on the same X axis, Y axis and H axis are numbered as XYHM, and M increases by number 0 with the number of shots;

[0040] S1022: The photovoltaic module images located on the same X axis and Y axis are numbered as XYM, M increases by number 0 with the number of shots, and the photovoltaic modules with different H heights are imaged separately to improve the imaging efficiency.

[0041] S103: Divide the inspection map to obtain several sub-inspection areas, set the inspection path according to the several sub-inspection areas after division, and further comprise at least one pollution three-dimensional rectangular coordinate system, generate a sub-inspection map according to the pollution three-dimensional rectangular coordinate system, the pollution three-dimensional rectangular coordinate system includes X1 axis, Y1 axis and H1 axis, use several division standard areas to tile the pollution three-dimensional rectangular coordinate system, keep the long side of each division standard area parallel to the X1 axis, the width of each division standard area parallel to the Y1 axis, and the height of each division standard area parallel to the H1 axis;

[0042] S1031: The photovoltaic panels on the same photovoltaic module are at the same X1 axis, Y1 axis and H1 axis height, the unmanned aerial vehicle separately shoots the pollution position, the pollution includes bird droppings, leaves, dust, etc., the X1 axis and Y1 axis shooting can measure the area of the pollution, and the H1 axis shooting can measure the thickness of the pollution, improving the imaging clarity.

[0043] ​S1032: Take the X1 axis, Y1 axis as the main mark, take the pollution image and number it as X1Y1M, M starts from 0 and increases with the number of shots;

[0044] It is worth noting that X1Y1 is multiplied, the larger the number represents the larger the pollution area, the higher the processing priority.

[0045] S1033: Take the pollution H1 axis height, the lower the H1 height, the greater the adhesion, the higher the processing priority.

[0046] For example: when bird droppings fall on the photovoltaic panel, the coverage area is determined according to the inclination of the photovoltaic panel, and the thickness is small; when leaves fall on the photovoltaic panel, the coverage area is determined according to the size of the leaves themselves, and the thickness is large; dust accumulates over the years, the coverage area is small, and the thickness is also small. The priority of processing pollutants decreases in turn from bird droppings, leaves, and dust.

[0047] Of course, the pollutants include not only bird droppings, leaves, dust, etc., but also other pollutants within the scope of protection.

[0048] S1034: Input the dirt image in the area into the trained dirt analysis model to determine the type of dirt in each dirt image.

[0049] Or, input the dirt image into the convolutional neural network model to obtain the dirt type and quantity.

[0050] The training process of the dirt analysis model includes: collecting multiple dirt images in advance, labeling each dirt image as a training image, and labeling the type of dirt in each training image; different dirt types are converted into digital labels; the labeled training images are divided into a training set and a test set; the training set is used to train the dirt analysis model, and the test set is used to test the dirt analysis model; a preset error threshold is set, when the average prediction error of all training images in the test set is less than the error threshold, the dirt analysis model is output; the dirt analysis model is a convolutional neural network model.

[0051] Specifically, the dirt analysis model is a convolutional neural network model; convolutional neural network (convolutional neural network, CNN or ConvNet) is a kind of multi-layer feedforward artificial neural network. Convolutional neural network can be regarded as a special case of feedforward network, which mainly simplifies and improves the feedforward network in network structure. In theory, back propagation algorithm can also be used to train convolutional neural network.

[0052] The convolutional neural network is a multi-layer feedforward network, and each layer input is a pattern of multiple two-dimensional matrices. The convolutional neural network model is composed of three parts, namely, an input layer, an intermediate layer, and an output layer. The intermediate layer is alternately composed of a convolutional layer and a pooling layer.

[0053] The input layer of the convolutional neural network directly receives a two-dimensional visual pattern, such as a two-dimensional image. No additional human intervention process is required to select or design suitable image features as input. The convolutional neural network can automatically extract features from raw image data and learn a classifier. The convolutional neural network can greatly reduce the manual preprocessing process and help to learn the most effective visual features for the current image processing task.

[0054] The intermediate layer is a feature extraction layer. Each convolutional layer includes multiple convolutional neurons. Each convolutional neuron is connected to a local receptive field of a corresponding position in the previous layer network and extracts image features of the part. The specific features extracted depend on the connection weight of the neuron and the local receptive field of the previous layer. Different connection weights extract different features.

[0055] In order to further reduce the network parameters, the convolutional neural network limits the weights of different neurons in the same convolutional layer connected to different positions in the previous layer network to be equal, that is, a convolutional layer is only used to extract the same features at different positions in the previous layer network. This restriction strategy is called weight sharing. By designing multiple convolutional layers, the convolutional neural network can extract multiple different features for the final image processing task. In actual use scenarios, the number of convolutional layers and the number of convolutional neurons in each convolutional layer should be determined according to the specific task.

[0056] The pooling layer belongs to the intermediate layer and is also a feature mapping layer. Each pooling layer includes multiple pooling neurons. The pooling neurons are only connected to the local receptive field of the corresponding position in the previous layer network. Unlike the convolutional neurons, all the weights of each pooling neuron connected to the local receptive field of the previous layer network are fixed to a specific value and are no longer iteratively updated during the network training process. The current pooling layer network not only no longer generates new training parameters, but also down-samples the features extracted by the previous layer network, further reducing the network size. Through the down-sampling of the local receptive field of the previous layer network, the network can more quickly recognize the potential deformation of the input pattern.

[0057] The output layer of the convolutional neural network is connected in a full connection manner like a common feedforward network. The two-dimensional feature pattern obtained by the last connection layer is stretched into a vector and connected to the output layer in a full connection manner. This structure can fully exploit the mapping relationship between the last extracted features of the network and the output class labels. In complex applications, the output layer can be designed as multiple full connection layers.

[0058] Specifically, in S1034, the convolutional neural network model comprises a first convolutional neural network model and a second convolutional neural network model, wherein the first convolutional neural network model is used for image recognition on a dirt image of a whole row of photovoltaic components to obtain a dirt type and quantity; the second convolutional neural network model is used for image recognition on a to-be-recognized image of a dirt image category of a single photovoltaic component to obtain a dirt type and quantity; in this embodiment, different training sets are used to train respective models, thereby improving the accuracy of the models in identifying a dirt type and quantity.

[0059] Further, the second convolutional neural network model is a convolutional neural network model with a residual mechanism, so that the learned features are more accurate in the training process of the convolutional neural network model, and thus the convolutional neural network model is more suitable for performing complex recognition tasks. It is worth emphasizing that the first convolutional neural network model is set as a common convolutional neural network model, and the second convolutional neural network model is a convolutional neural network model with a residual mechanism. Through such a setting, the dirt image in an image can be accurately recognized by a simple model, and the dirt image in an image can be accurately recognized by a complex model, thereby realizing the unification of efficiency and accuracy.

[0060] It is worth mentioning that S1034 can be a further step of S103, or can directly replace S102-S104.

[0061] S104: sorting a plurality of groups of sub-inspection areas according to H-axis coordinates from small to large, and / or sorting a plurality of groups of sub-inspection areas according to H1-axis coordinates from small to large to obtain an inspection group sorting;

[0062] The plurality of groups of sub-inspection areas are sorted according to H-axis coordinates from small to large, and the smaller the H height value, the higher the processing priority. The smaller the H height value, the lower the inclination of the photovoltaic component, and the lower the possibility of self-cleaning of the pollutants due to their own gravity, wind and other factors, and the photovoltaic component needs to be processed in priority.

[0063] The plurality of groups of sub-inspection areas are sorted according to H1-axis coordinates from small to large, and the smaller the H1 height value, the higher the processing priority. The smaller the H1 height value, the higher the probability of the pollutants being adhesives, and the adhesives need to be processed in time. The longer the delay time, the more difficult it is to clean, which affects the utilization rate of light and the scientific judgment of the pollution condition of the power station. According to the relevant priority, the photovoltaic component is processed for operation and maintenance.

[0064] S105: completing unmanned aerial vehicle inspection operation preparation and unmanned aerial vehicle inspection operation analysis according to the inspection group sorting, and performing unmanned aerial vehicle inspection operation; collecting image data and performing data preprocessing; constructing a corresponding inspection report according to the inspection data to perform operation and maintenance dispatching.

[0065] S1051: corresponding and recording cleaning parameters in the region, the cleaning parameter generation method comprising: taking the instant environmental data, the dirt type and the dirt coefficient corresponding to a region as a set of analysis data, a set of analysis data corresponding to a region; the cleaning robot obtains the corresponding cleaning parameters under the condition of the analysis data, and the cleaning parameters correspond one by one to the analysis data; the cleaning parameters include water consumption, water pressure, cleaning angle, cleaning time and cleaning agent type.

[0066] The analysis data is input into the trained first parameter acquisition model to predict the corresponding water consumption; the analysis data is input into the trained second parameter acquisition model to predict the corresponding water pressure; the analysis data is input into the trained third parameter acquisition model to predict the corresponding cleaning angle; the analysis data is input into the trained fourth parameter acquisition model to predict the corresponding cleaning time; and the analysis data is input into the trained fifth parameter acquisition model to predict the corresponding cleaning agent type.

[0067] Different types of pollutants are identified and guided to clean the photovoltaic module. Large and high shielding rate need to be immediately maintained; dust pollution from factory exhaust also fluctuates due to monthly orders and production process, which generally needs a long time to deposit and clean. Accurate identification of pollution degree and main pollutants can enable the cleaning team to adopt targeted cleaning solutions, such as fallen leaves requiring clamps and other tools, bird droppings requiring shovels and cloths, and dust requiring cloths. Different cleaning agents are used for different pollutants to improve cleaning efficiency.

[0068] H-axis measures the height of the photovoltaic module, i.e. the inclination of the photovoltaic module; H1-axis measures the thickness of the pollutant, i.e. the thickness value of the pollutant adhering to the inclined photovoltaic module.

[0069] In the photovoltaic module maintenance work of the photovoltaic power station, the unmanned aerial vehicle is used to improve the inspection efficiency, efficiently handle the photovoltaic module pollutants, and improve the light utilization rate of the photovoltaic module. The unmanned aerial vehicle has high inspection frequency and high system automation degree, without human intervention, improving the accuracy of judgment and realizing the whole life cycle tracking of the photovoltaic module.

[0070] The photovoltaic module pollutant data processing priority is reasonable, the photovoltaic module pollutants are efficiently cleaned, the photovoltaic module cleanliness is maintained, and the light utilization rate of the photovoltaic module is improved.

[0071] The process of obtaining photovoltaic module image data includes: using real-time shooting method to obtain a plurality of sampling images for X-axis, Y-axis and H-axis coordinate axes.

[0072] Obtaining the position of the photovoltaic module in each sample image, removing the blurred photovoltaic module position from the sample image according to the blurred photo, and then superimposing the remaining part of the sample image to obtain a superimposed image;

[0073] If the superimposed image is damaged, the unmanned aerial vehicle is commanded to obtain a plurality of sample images by real-time shooting to improve the clarity of image shooting and reduce the influence of blurred images on data processing.

[0074] The process of obtaining the pollutant image data includes: obtaining a plurality of sample images by real-time shooting for the X1, Y1 and H1 coordinate axes;

[0075] Obtaining the position of the pollutant in each sample image, removing the blurred pollutant position from the sample image according to the blurred photo, and then superimposing the remaining part of the sample image to obtain a superimposed image;

[0076] If the superimposed image is damaged, the unmanned aerial vehicle is commanded to obtain a plurality of sample images by real-time shooting to improve the clarity of image shooting and reduce the influence of blurred images on data processing.

[0077] The method further includes meteorological monitoring data, which includes meteorological observation monitoring data, meteorological satellite monitoring data and meteorological radar monitoring data, wherein the meteorological observation monitoring data includes temperature data, humidity data, wind speed data and precipitation data, the meteorological satellite monitoring data includes infrared cloud image data and visible light cloud image data, and the meteorological radar monitoring data includes basic reflectivity data and precipitation data. Considering the influence factors of extreme weather, timely operation and maintenance are performed on the photovoltaic module to prolong the service life of the photovoltaic module and reduce costs.

[0078] The extreme data includes:

[0079] When the wind speed is greater than 30 meters per second, the ambient temperature is between -30 degrees Celsius and 0 degrees Celsius, or the ambient temperature is above 35 degrees Celsius.

[0080] When the air humidity is greater than 80%, the precipitation is 50 millimeters or more in 24 hours, it is called "heavy rain". According to the intensity of precipitation, it is divided into three grades, i.e. 50-99.9 millimeters in 24 hours is called "heavy rain"; 100-250 millimeters is called "heavy rain"; and 250 millimeters or more is called "heavy rain".

[0081] The method further includes photovoltaic module placement monitoring, which is used to monitor whether the photovoltaic module has a position deviation; if so, the unmanned aerial vehicle is commanded to obtain new position information of the photovoltaic module by real-time shooting to update the data in time and improve the data accuracy.

[0082] Before generating the inspection report according to the image data, the method further comprises: according to the weather monitoring data, if the weather monitoring data is greater than the adverse weather value, instructing the unmanned aerial vehicle to capture second image data by using the real-time shooting method, and comparing the second image data with the initial image data; the unmanned aerial vehicle timely captures the second image data after the extreme weather, monitors the pollution degree of the photovoltaic module, and improves the accuracy of the detection data.

[0083] The method further comprises: after the operation and maintenance order is completed, instructing the unmanned aerial vehicle to capture third image data by using the real-time shooting method, and comparing the third image data with the initial image data; the unmanned aerial vehicle timely captures the third image data after the staff completes the operation and maintenance, monitors the pollution degree of the photovoltaic module, improves the accuracy of the detection data, and evaluates the operation and maintenance completion degree.

[0084] The method further comprises constructing a data preprocessing model, training the data preprocessing model by using a core algorithm, and the core algorithm comprises: researching a basic sparse coding model of image information, researching an image sparse coding algorithm based on an optimized dictionary, and researching a convolutional neural network algorithm.

[0085] The above steps are mainly for large pollutants such as leaves and bird droppings, and the following steps are mainly for small pollutants such as dust.

[0086] S1035: convert the soot accumulation index problem into a 4-classification problem, for the model framework of the classification problem, the most commonly used classification network structure in deep learning is adopted, including a convolutional neural network (CNN), a recurrent neural network (RNN), and the like.

[0087] The soot accumulation index finally outputs a probability distribution, and the item with the maximum probability value is the category belonging to the soot accumulation degree. The soot accumulation index degree is converted into the following through the defined category: 0: no soot accumulation, 1: slight soot accumulation, 2: more soot accumulation, and 3: serious soot accumulation.

[0088] According to the soot accumulation index degree value from large to small, the operation and maintenance order is dispatched, the cleaning efficiency is improved, and the priority of the operation and maintenance order is more reasonable.

[0089] A computer-readable storage medium comprises a stored program, wherein when the program is running, the computer-readable storage medium controls the device where the computer-readable storage medium is located to execute the unmanned aerial vehicle inspection method for monitoring the pollution state of the photovoltaic module as described above.

[0090] A processor is used to run a program, wherein when the program is running, the processor executes the unmanned aerial vehicle inspection method for monitoring the pollution state of the photovoltaic module as described above.

[0091] Those skilled in the art can understand that the units (or modules, steps, etc., the same below) of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components of each example have been described in the above description in general terms. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0092] In the embodiments provided by the present application, it should be understood that the division of units is only a logical functional division, and there can be another division manner when actually implemented, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.

[0093] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0094] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0095] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the description of the present application.

Claims

1. A method for monitoring the pollution state of a photovoltaic assembly by means of a drone, characterized in that, Comprising S101: read the photovoltaic module position setting information from the power grid data platform and satellite GPS to generate an inspection target template, and generate initial image data according to the inspection target template and the inclination parameter of the photovoltaic module setting; S102: establish at least one photovoltaic module three-dimensional rectangular coordinate system, generate a patrol map according to the photovoltaic module three-dimensional rectangular coordinate system, the photovoltaic module three-dimensional rectangular coordinate system includes X axis, Y axis and H axis, use several division standard areas to tile the photovoltaic module three-dimensional rectangular coordinate system, keep the long side of each division standard area parallel to the X axis, the width of each division standard area is parallel to the Y axis, and the height of each division standard area is parallel to the H axis; S103: divide the patrol map to obtain a plurality of sub-patrol areas, and set a patrol path according to the plurality of sub-patrol areas after division, further comprising: at least one pollutant three-dimensional rectangular coordinate system, generating a sub-patrol map according to the pollutant three-dimensional rectangular coordinate system, the pollutant three-dimensional rectangular coordinate system includes X1 axis, Y1 axis and H1 axis, using a plurality of division standard areas to tile the pollutant three-dimensional rectangular coordinate system, keeping the long side of each division standard area parallel to the X1 axis, the width of each division standard area is parallel to the Y1 axis, and the height of each division standard area is parallel to the H1 axis; S104: sort a plurality of groups of sub-patrol areas according to the H axis coordinate from small to large, and / or sort a plurality of groups of sub-patrol areas according to the H1 axis coordinate from small to large to obtain a patrol group sorting; S105: complete the unmanned aerial vehicle patrol operation preparation and unmanned aerial vehicle patrol operation analysis according to the patrol group sorting, and perform the unmanned aerial vehicle patrol operation; collect image data and perform data preprocessing; construct a corresponding patrol report according to the patrol data to perform operation and maintenance order.

2. A UAV inspection method for monitoring the state of contamination of a photovoltaic assembly according to claim 1, characterized in that, The acquisition process of the photovoltaic module image data includes: acquiring a plurality of sample images by using a real-time shooting method for the X axis, Y axis and H axis coordinate axes; Obtain the position of the photovoltaic module in each sample image, remove the blurred photovoltaic module position from the sample image according to the blurred photo, and then superimpose the remaining part of the sample image to obtain a superimposed image; Determine whether the superimposed image is damaged, if so, command the unmanned aerial vehicle to acquire a plurality of sample images by using a real-time shooting method.

3. A UAV inspection method for monitoring the state of contamination of a photovoltaic assembly according to claim 1, characterized in that, The acquisition process of the pollutant image data includes: acquiring a plurality of sample images by using a real-time shooting method for the X1 axis, Y1 axis and H1 axis coordinate axes; Obtain the position of the photovoltaic module in each sample image, remove the blurred photovoltaic module position from the sample image according to the blurred photo, and then superimpose the remaining part of the sample image to obtain a superimposed image; Determine whether the superimposed image is damaged, if so, command the unmanned aerial vehicle to acquire a plurality of sample images by using a real-time shooting method.

4. The UAV inspection method for monitoring the pollution state of a photovoltaic module according to claim 1, wherein, The method further comprises meteorological monitoring data, the meteorological monitoring data comprising meteorological observation monitoring data, meteorological satellite monitoring data and meteorological radar monitoring data, wherein the meteorological observation monitoring data comprises temperature data, humidity data, wind speed data and precipitation data, the meteorological satellite monitoring data comprises infrared cloud image data and visible light cloud image data, and the meteorological radar monitoring data comprises basic reflectivity data and precipitation data.

5. The UAV inspection method for monitoring the pollution state of a photovoltaic module according to claim 1, wherein, The method further comprises photovoltaic module placement monitoring, which is used to monitor whether the photovoltaic module has a positional deviation; if yes, the UAV uses real-time shooting to obtain new position information of the photovoltaic module.

6. A UAV inspection method for monitoring the state of contamination of a photovoltaic assembly according to claim 4, characterized in that, Before generating the inspection report according to the image data, the method further comprises: According to the meteorological monitoring data, if the meteorological monitoring data is greater than a severe weather value, the UAV uses real-time shooting to shoot second image data, and compares the second image data with the initial image data.

7. A UAV inspection method for monitoring the state of contamination of a photovoltaic assembly according to claim 1, characterized in that, After the operation and maintenance order is completed, the method further comprises that the UAV uses real-time shooting to shoot third image data, and compares the third image data with the initial image data.

8. A UAV inspection method for monitoring the state of contamination of a photovoltaic assembly according to claim 1, characterized in that, The method further comprises constructing a data preprocessing model, training the data preprocessing model using a core algorithm, and the core algorithm comprises researching a basic sparse coding model of image information, researching an image sparse coding algorithm based on an optimized dictionary, and researching a convolutional neural network algorithm.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored program, wherein the program, when executed, controls a device in which the computer readable storage medium is located to perform the UAV inspection method for monitoring a pollution state of a photovoltaic module according to any one of claims 1 to 8.

10. A processor, comprising: The processor is configured to execute a program, wherein the program, when executed, performs the UAV inspection method for monitoring a pollution state of a photovoltaic module according to any one of claims 1 to 8.

Images