Photovoltaic power station operation state monitoring method based on unmanned aerial vehicle inspection

Abstract

The invention provides a photovoltaic power station operation state monitoring method based on unmanned aerial vehicle inspection, and the method comprises the steps: carrying out the surveying and mapping of the geographic position, landform and photovoltaic array layout of a photovoltaic power station through employing a surveying and mapping technology and a geographic information system, so as to build a digital photovoltaic power station model; acquiring technical parameters of each component of the photovoltaic power station, and mapping the technical parameters to the digital photovoltaic power station model; based on a preset inspection plan and the digital photovoltaic power station model, automatically generating an unmanned aerial vehicle inspection path; obtaining monitoring data obtained by the unmanned aerial vehicle according to the unmanned aerial vehicle inspection path, wherein the monitoring data comprises an appearance image of a photovoltaic array, temperature distribution of a photovoltaic module and illumination intensity data of different areas of a photovoltaic power station; vectorizing the monitoring data to obtain an input vector; and the operation state monitoring model performs feature extraction on the input vector so as to analyze the operation state of the photovoltaic power station.

Description

Technical Field

[0001] The present application relates to the field of photovoltaic technology, and in particular to a method for monitoring the operating status of a photovoltaic power station based on drone inspection. Background Art

[0002] As the global demand for clean energy grows, photovoltaic power stations have been widely built and applied as an important renewable energy utilization facility. The efficient and stable operation of photovoltaic power stations is extremely critical to ensuring energy supply and reducing carbon emissions. However, photovoltaic power stations usually occupy a large area, contain numerous photovoltaic components, complex electrical connection systems, and are affected by a variety of environmental factors, which brings great challenges to their operating status monitoring.

[0003] Traditional inspections of photovoltaic power stations mainly rely on manual methods. Operation and maintenance personnel regularly inspect the power station, observe the appearance of the photovoltaic array with the naked eye, and use simple measuring instruments to detect some electrical parameters. This manual inspection method has many disadvantages. First, manual inspections are inefficient. For large-scale photovoltaic power stations, completing a comprehensive inspection often requires a lot of manpower and time, which is difficult to meet the needs of high-frequency inspections. Secondly, the accuracy and reliability of manual inspections are limited by factors such as personnel experience, skill level, and working status. Some minor faults or potential problems may be missed, resulting in the failure to discover and deal with the faults in time, which in turn affects the power generation efficiency and service life of the power station. For example, slight dirt on the surface of the photovoltaic panel or early hot spot phenomenon may not be easily detected by manual inspections, and these problems will gradually worsen over time, reducing the power generation performance of the photovoltaic panel. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a method for monitoring the operating status of a photovoltaic power station based on drone inspection.

[0005] A method for monitoring the operating status of a photovoltaic power station based on drone inspection, comprising:

[0006] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0007] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0008] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0009] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0010] Vectorizing the monitoring data to obtain an input vector;

[0011] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0012] Optionally, the use of surveying and mapping technology and geographic information system to survey and map the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model includes:

[0013] Use surveying and mapping technology and geographic information systems to survey the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to generate geographic location coordinate data, digital elevation model (DEM), digital orthophoto map (DOM) and measurement data of photovoltaic array layout;

[0014] Generate a basic data layer with spatial location attributes based on the geographic location coordinate data, digital elevation model (DEM), digital orthophoto map (DOM) and measurement data of photovoltaic array layout;

[0015] Imported PV array layout data, creating vector layers to represent different PV module entities;

[0016] The vector layer is superimposed on the basic data layer to establish a digital photovoltaic power station model.

[0017] Optionally, the acquiring technical parameters of each component of the photovoltaic power station to map them to the digital photovoltaic power station model includes:

[0018] Classify the components of the photovoltaic power station, including photovoltaic panels, inverters, combiner boxes, brackets, etc. For each type of component, determine the technical parameters that need to be collected;

[0019] Parsing the technical parameter description document to collect the technical parameters;

[0020] The spatial entity corresponding to each type of component is determined, and the spatial entity is mapped to the digital photovoltaic power station model through a unique identifier.

[0021] Optionally, the automatically generating a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model includes:

[0022] Analyze the preset inspection plan, and plan inspection points and inspection accuracy on the digital photovoltaic power station model;

[0023] Based on the inspection points and the progress of the inspection, determine the dwell time of the drone at each inspection point and the frequency of data collection;

[0024] Based on the dwell time of the drone at each inspection point and the frequency of data collection, the drone inspection path is automatically generated.

[0025] Optionally, the acquisition of monitoring data acquired by the drone along the inspection path of the drone, the monitoring data including the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station, includes:

[0026] Acquire monitoring data acquired by the drone along the inspection route of the drone, and separate the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station;

[0027] The appearance images, temperature distribution images and light intensity data collected in the same area and at the same time are associated and integrated according to the time sequence and location information to form a comprehensive detection data set.

[0028] Optionally, vectorizing the monitoring data to obtain an input vector includes:

[0029] According to the different types of data contained in the monitoring data, a vectorization model is matched to input the appearance image and the temperature distribution image into the feature extractor for vectorization to obtain an image data vector, and the light intensity data and the temperature value are input into the statistical feature extraction model for vectorization to obtain a numerical data vector;

[0030] The image data vector and the numerical data vector are combined to form the input vector.

[0031] Optionally, the operation status monitoring model performs feature extraction on the input vector to analyze the operation status of the photovoltaic power station, including:

[0032] Inputting the input vector into the operation status monitoring model for forward propagation, gradually extracting features at different levels in each layer, and obtaining feature representations of valuable information about the operation status of the photovoltaic power station;

[0033] The operating status of the photovoltaic power station is analyzed according to the characteristic representation.

[0034] A photovoltaic power station operation status monitoring device based on drone inspection, comprising:

[0035] The first program unit is used to map the geographical location, topography and photovoltaic array layout of the photovoltaic power station by using surveying and mapping technology and geographic information system to establish a digital photovoltaic power station model;

[0036] A second program unit is used to obtain technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0037] A third program unit is used to automatically generate a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model;

[0038] A fourth program unit is used to obtain monitoring data obtained by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic module, and light intensity data of different areas of the photovoltaic power station;

[0039] A fifth program unit, used for vectorizing the monitoring data to obtain an input vector;

[0040] The sixth program unit is used to run the status monitoring model to extract features from the input vector to analyze the operating status of the photovoltaic power station.

[0041] A computer program product having computer executable instructions stored thereon, wherein the computer executable instructions are executed to perform the following steps:

[0042] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0043] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0044] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0045] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0046] Vectorizing the monitoring data to obtain an input vector;

[0047] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0048] An electronic device comprises a memory and a processor, wherein the memory stores computer executable instructions, and the processor runs the computer executable instructions to perform the following steps:

[0049] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0050] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0051] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0052] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0053] Vectorizing the monitoring data to obtain an input vector;

[0054] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0055] In the embodiment of the present application, the following technical advantages are achieved:

[0056] (1) Based on the drone inspection method, a digital photovoltaic power station model is established through surveying and mapping technology and geographic information system, and an inspection path is automatically generated based on it. The drone can quickly perform the inspection task according to the path. The flight speed and maneuverability of the drone are much higher than manual inspection. It can cover a large area of ​​photovoltaic power stations in a short time, greatly improving the inspection efficiency, meeting the requirements of high-frequency inspections, and grasping the operation status of the power station in a more timely manner.

[0057] (2) UAVs can obtain multi-dimensional monitoring data such as the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic modules, and the light intensity data of different areas. These data can more comprehensively and accurately reflect the operating status of the photovoltaic power station, avoiding the omission of problems due to the subjectivity and limitations of manual inspections. For example, the thermal imaging equipment carried by drones can accurately detect the temperature distribution of photovoltaic modules, and even early and minor hot spot phenomena can be discovered in time, while manual inspections may be difficult to detect such subtle problems.

[0058] (3) Vectorize the acquired monitoring data to obtain the input vector, and then extract and analyze the features through the operation status monitoring model. This data and model-based analysis method is more scientific, accurate and reliable than manual judgment based on experience. It can more accurately judge the operation status of the photovoltaic power station and discover potential faults and problems in time, thereby ensuring the power generation efficiency and service life of the power station.

[0059] (4) Using surveying and mapping technology and geographic information systems, a digital photovoltaic power station model is established, and the technical parameters of each component are mapped to the model, realizing the digital presentation of the photovoltaic power station. This provides an intuitive and efficient platform for subsequent inspection route planning, data management and analysis, making it easier for operation and maintenance personnel to fully understand the structure and operation of the power station.

[0060] (5) Based on the preset inspection plan and digital model, the drone inspection route is automatically generated, which reduces the workload and subjectivity of manual route planning and improves the scientificity and rationality of the inspection. At the same time, the monitoring data is analyzed through the operation status monitoring model to realize the intelligent evaluation of the operation status of the photovoltaic power station, which can timely and accurately identify problems and provide corresponding decision support, thus improving the operation and maintenance management level of the photovoltaic power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0062] Figure 1 It is a flow chart of a method for monitoring the operating status of a photovoltaic power station based on drone inspection according to an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to enable those skilled in the art to better understand the solution of the present application, the following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0064] Figure 1 FIG. 1 is a flow chart of a method for monitoring the operation status of a photovoltaic power station based on drone inspection according to an embodiment of the present application. Figure 1 As shown, a method for monitoring the operation status of a photovoltaic power station based on drone inspection includes:

[0065] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0066] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0067] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0068] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0069] Vectorizing the monitoring data to obtain an input vector;

[0070] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0071] Optionally, the use of surveying and mapping technology and geographic information system to survey and map the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model includes:

[0072] Use surveying and mapping technology and geographic information systems to survey the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to generate geographic location coordinate data, digital elevation model (DEM), digital orthophoto map (DOM) and measurement data of photovoltaic array layout;

[0073] Generate a basic data layer with spatial location attributes based on the geographic location coordinate data, digital elevation model (DEM), digital orthophoto map (DOM) and measurement data of photovoltaic array layout;

[0074] Imported PV array layout data, creating vector layers to represent different PV module entities;

[0075] The vector layer is superimposed on the basic data layer to establish a digital photovoltaic power station model.

[0076] Specifically, when generating surveying and mapping data, for geographic location coordinate data:

[0077] The global positioning system (GPS) is used for measurement, and the obtained geographic location coordinates are expressed in longitude and latitude. In geodesy, points on the earth's surface can be expressed in geographic coordinates (B, L, H), where B is latitude, L is longitude, and H is geodetic height. In practical applications, in order to facilitate subsequent processing, they are often converted into plane rectangular coordinates. The Gauss-Krüger projection is used here, and its forward calculation formula (from longitude and latitude to plane rectangular coordinates) is as follows:

[0078]

[0079] in:

[0080] - a is the semi-major axis of the Earth's ellipsoid, and e is the first eccentricity.

[0081] - is the radius of curvature of the Maoyou circle.

[0082] -ρ = 180 / π (a constant for converting radians to degrees)

[0083] -η 2 =e′ 2 ·cos 2 B, e′ is the second eccentricity, and L0 is the central meridian longitude.

[0084] To this end, in the photovoltaic power station mapping scenario of this application, the Gauss-Krüger projection is used to convert the longitude and latitude coordinates into plane rectangular coordinates, which has the following technical advantages in spatial matching and processing with other data, as follows:

[0085] 1. Unified coordinate system for easy data integration

[0086] The data sources of photovoltaic power plants are wide-ranging, including terrain data, meteorological data, photovoltaic module layout data, etc. These data may use different coordinate systems. Latitude and longitude coordinates are a kind of geographic coordinate system, which is universal when describing the location on the earth, but it is inconvenient when performing spatial analysis and data integration.

[0087] By converting longitude and latitude coordinates into plane rectangular coordinates through Gauss-Krüger projection, all data related to photovoltaic power plants can be unified into the same plane rectangular coordinate system. For example, terrain data may exist in the form of digital elevation model (DEM). By converting it into the same plane rectangular coordinates, it can be overlaid and analyzed with the location information of photovoltaic modules in the same coordinate system, making it easier to determine the terrain slope, slope direction and other information at the location of the photovoltaic modules, providing more accurate data support for the site selection, design and layout optimization of photovoltaic power plants.

[0088] 2. Simplify spatial calculations and improve processing efficiency

[0089] When performing spatial analysis and processing, calculations in the plane rectangular coordinate system are relatively simple and intuitive. Compared with the spherical geometry calculations involved in the longitude and latitude coordinate system, calculations of distance, angle, area, etc. in the plane rectangular coordinate system can use simpler mathematical formulas, greatly reducing the complexity and amount of calculations.

[0090] For example, when calculating the distance between two photovoltaic modules, if longitude and latitude coordinates are used, the curvature of the earth needs to be considered and complex geodetic formulas need to be used for calculation; however, after conversion to plane rectangular coordinates, the Euclidean distance formula can be used directly. The results can be obtained quickly. This can significantly improve the computing efficiency and save the computing resources and time of the processor for tasks such as photovoltaic power station inspection path planning and shadow analysis that require frequent spatial calculations.

[0091] 3. Accurate spatial matching to improve analysis accuracy

[0092] Plane rectangular coordinates can provide more accurate position representation, making the spatial matching of different types of data more accurate. In photovoltaic power plants, it is necessary to accurately match the appearance image, temperature distribution data, light intensity data, etc. of photovoltaic modules with the actual position of the modules in order to conduct effective operation status monitoring and analysis.

[0093] The plane rectangular coordinates obtained by Gauss-Krüger projection can accurately correspond these data to the actual position of the photovoltaic module. For example, when analyzing the relationship between the temperature distribution and light intensity of the photovoltaic module, the photovoltaic module corresponding to each temperature measurement point and light measurement point can be accurately determined, so as to more accurately analyze the impact of light on the module temperature and promptly discover potential problems such as hot spot effect.

[0094] 4. It is conducive to data storage and transmission, reducing resource consumption

[0095] The data representation of plane rectangular coordinates is relatively simple, and compared with longitude and latitude coordinates, it occupies less space and bandwidth during data storage and transmission. This can reduce data storage costs and transmission burdens and improve data management and transmission efficiency when processing large amounts of photovoltaic power station monitoring data.

[0096] For example, when transmitting monitoring data of a photovoltaic power station to a remote server for analysis and storage, a smaller data volume can reduce transmission time and the possibility of network congestion, ensuring timely and stable data transmission. At the same time, when storing data locally, it can also save storage space and improve the utilization of storage devices.

[0097] For digital elevation models (DEM), terrain elevation data is obtained through technologies such as airborne laser radar (LiDAR). There are n measurement points in the measurement area, and their coordinates are (x i ,y i ), the corresponding elevation is z i In order to generate a DEM with a regular grid, the Kriging interpolation method is preferably used.

[0098] Assume z(s) is the elevation value at spatial location s, and the Kriging interpolation formula is:

[0099] in is the estimated elevation value at the interpolation point s0, λ iis the weight coefficient, obtained by solving the following system of equations:

[0100] Among them, γ(s i ,s j ) is the semivariogram function, which is used to describe the degree of variation between two points in space. Its calculation formula is:

[0101]

[0102] In the photovoltaic power station scenario, DEM reflects the terrain undulations in the area where the power station is located, which is of great significance for evaluating the installation angle and drainage conditions of photovoltaic modules.

[0103] Therefore, in the photovoltaic power station scenario, the use of Kriging interpolation to generate a digital elevation model (DEM) has the following technical benefits:

[0104] 1. High-precision terrain simulation

[0105] The Kriging interpolation method is based on the regionalized variable theory and fully considers the autocorrelation of data in spatial locations. In the photovoltaic power station area, the terrain elevation data is not randomly distributed, and the elevations of adjacent locations are often similar. The semivariogram γ(s i ,s j ) quantifies this spatial autocorrelation by solving the weight coefficient λ i The set of equations enables the estimated elevation of the interpolation point to integrate the information of the surrounding measurement points, so that the generated DEM can more accurately reflect the actual ups and downs of the terrain and provide high-precision terrain data support for the planning and design of photovoltaic power stations.

[0106] Photovoltaic power stations may be built in various terrain conditions, such as mountains, hills and other complex terrain areas. Kriging interpolation method can adaptively adjust the interpolation results according to the distribution of measurement points and terrain characteristics, and generate a smooth DEM that conforms to the actual terrain even when the measurement points are unevenly distributed. This helps to accurately evaluate parameters such as terrain slope and slope direction at different locations, which is crucial for the design of the optimal installation angle of photovoltaic modules, ensuring that photovoltaic modules can receive sunlight to the greatest extent and improve power generation efficiency.

[0107] 2. Optimization of the installation angle of photovoltaic modules

[0108] An accurate DEM can provide precise terrain information for each PV module installation location. By analyzing DEM data, the terrain slope and slope direction of each location can be accurately calculated, so that the best installation angle for the PV module can be designed according to the direction and angle of solar radiation. For example, in a mountain photovoltaic power station, the terrain slope and slope direction at different locations vary greatly. The DEM generated by the Kriging interpolation method can determine a personalized installation angle for each PV module, so that the PV module can obtain more light time and radiation in a year, thereby improving the overall power generation efficiency.

[0109] The appropriate installation angle of photovoltaic modules can not only improve power generation efficiency, but also enhance the stability of energy output. By optimizing the installation angle based on high-precision DEM, the problem of uneven illumination caused by terrain shading can be reduced, the power generation differences between photovoltaic modules can be reduced, and the energy output of the entire photovoltaic power station can be made more stable and reliable, which is conducive to the stable operation of the power system and power dispatching.

[0110] Specifically, for digital orthophotos (DOM), image data is obtained through aerial photogrammetry, and then orthorectification is performed to obtain DOM. In photogrammetry, the relationship between the image point coordinates (x, y) and the ground point coordinates (X, Y, Z) can be expressed by the collinear equation:

[0111]

[0112] Where: (x0, y0) is the coordinate of the image principal point, f is the focal length of the camera, (X S , Y S , Z S ) is the coordinate of the photography center, a i , b i , c i (i=1, 2, 3) is the rotation matrix element, which is related to the camera's posture.

[0113] By solving the above equations and combining the coordinates of the ground control points, the original image can be corrected to generate an orthophoto. DOM provides intuitive image information of the photovoltaic power station, which helps to identify the appearance and layout of photovoltaic modules.

[0114] Therefore, in the scenario of photovoltaic power station, aerial photogrammetry is used to obtain image data and orthorectification is performed through collinear equations to obtain digital orthophoto map (DOM), which has the following technical benefits:

[0115] (1) DOM uses orthographic projection to visually display the entire PV power station. Through DOM, operation and maintenance personnel, engineers, and managers can clearly see the layout of the entire PV power station, including the arrangement of PV modules, spacing, and the relationship with the surrounding geographical environment. This helps to make reasonable layout design in the project planning stage, and quickly locate and identify the location of each PV module in the later operation and maintenance.

[0116] (2) DOM can provide high-resolution image information, which clearly presents the appearance of photovoltaic modules. It can be observed whether there are stains, cracks, damage and other problems on the surface of photovoltaic modules. These details are crucial for timely detection of potential failures and damages of photovoltaic modules. Early detection and treatment of these problems can prevent the failure from further deteriorating, thereby ensuring the normal operation of photovoltaic power stations and improving power generation efficiency.

[0117] (3) Since DOM is an orthorectified image, the distance and area on the image have actual geographic coordinate meaning and can be accurately measured. DOM can be used to measure the size and spacing of photovoltaic modules, as well as the area occupied by photovoltaic power stations. These measurement data provide important references for the design, construction and expansion of photovoltaic power stations, and help rational planning and resource utilization.

[0118] (4) Combining DOM and related geographic information data, the installation angle of PV modules and the terrain slope of the area can be analyzed. The accurate installation angle is crucial to the power generation efficiency of PV modules. Through DOM angle analysis, it can be evaluated whether the installation of PV modules meets the design requirements and whether adjustments are needed. At the same time, the analysis of terrain slope helps to understand the drainage situation and avoid water accumulation from damaging PV modules.

[0119] (5) During the operation and maintenance of a PV power station, when a fault or abnormality occurs, DOM can help maintenance personnel quickly locate the problem. By comparing with historical DOM data, abnormal phenomena such as changes in the position and damage of PV modules can be discovered in a timely manner, shortening the troubleshooting time and improving maintenance efficiency.

[0120] DOM can be used as important data for regular inspection and long-term monitoring of photovoltaic power plants. By regularly acquiring and analyzing DOM data, the operating status of photovoltaic power plants can be monitored, the aging and damage trends of photovoltaic components can be observed, and maintenance plans and replacement strategies can be formulated in a timely manner to ensure the long-term stable operation of photovoltaic power plants.

[0121] Specifically, for the measurement data of the photovoltaic array layout, a total station or other measuring equipment is used to obtain the position and size information of each photovoltaic module. The coordinates of the four vertices of the i-th photovoltaic module in the plane rectangular coordinate system are (x i1 ,yi1 ), (x i2 ,y i2 ), (x i3 ,y i3 ), (x i4 ,y i4 ), these coordinate information constitutes the measurement data of the photovoltaic array layout. In practical applications, these data are used to accurately represent the position and shape of photovoltaic modules in the power station.

[0122] Specifically, when generating the basic data layer, the above-obtained geographic location coordinate data, DEM, DOM and the measurement data of the photovoltaic array layout are integrated, and the process is as follows:

[0123] First, the converted plane rectangular coordinates of the geographic location coordinate data are aligned with the DEM and DOM. There is already a reference coordinate system. For each pixel in the DEM and DOM, it is mapped to the same coordinate system as the geographic location coordinate data through coordinate transformation.

[0124] Assume (x d ,y d ) is the pixel coordinate in DEM or DOM, (x r ,y r ) is the coordinate in the reference coordinate system. The coordinate transformation can be achieved through affine transformation:

[0125]

[0126] where a ij are the affine transformation coefficients, which are solved using the known control point coordinates.

[0127] Then, the aligned DEM, DOM and geographic location coordinate data are combined into a multi-channel basic data layer. For example, DEM is used as one channel, and the red, green and blue channels of DOM are used as three channels respectively to form a four-channel basic data layer.

[0128] Specifically, when creating a vector layer, a vector layer is created to represent different photovoltaic component entities according to measurement data of the photovoltaic array layout.

[0129] A polygon (rectangular shape of a PV module) is represented by a sequence of vertex coordinates. For the ith PV module, its vertex coordinates are (x i1 ,y i1 ), (x i2 ,y i2 ), (x i3 ,y i3 ), (x i4 ,y i4 ), which can be represented as a Vector object.

[0130] For each photovoltaic module, a polygon geometry object is created and given corresponding attributes, such as module number, model, etc. These polygon geometry objects constitute the vector layer.

[0131] Specifically, when establishing a digital photovoltaic power station model, the vector layer is superimposed on the basic data layer to establish a digital photovoltaic power station model. The detailed process is as follows:

[0132] In geographic information systems, overlay operations are implemented through spatial indexing and geometric calculations. First, spatial indexes are created for vector layers and basic data layers to quickly find overlapping areas.

[0133] The spatial indexes of the vector layer V and the basic data layer B have been established. For each polygon object P in the vector layer, the area R overlapping with the polygon object in the basic data layer is found through the spatial index.

[0134] Then, the attribute information of the polygon object P and the overlapping area R are combined to generate the final digital photovoltaic power station model. For example, the number and attribute information of the photovoltaic module can be marked at the corresponding position of the basic data layer, so as to realize the superposition of the vector layer and the basic data layer.

[0135] Specifically, in this embodiment, the construction process is described by taking the quadtree index as an example. A quadtree is a tree-like data structure based on space division, which recursively divides the space area into four sub-areas (quadrants). A quadtree index is established for the basic data layer B and the vector layer V. For a rectangular area R in two-dimensional space (representing the entire photovoltaic power station area), the coordinates of its lower left corner are (x min ,y min ), the coordinate of the upper right corner is (x max ,y max ).

[0136] Each node N of the quadtree contains the following information: the rectangular area R represented by the node N , the coordinate of its lower left corner is (x Nmin ,y Nmin ), the coordinate of the upper right corner is (x Nmax ,y Nmax ).

[0137] -Four child node pointers c1, c2, c3, c4, pointing to the nodes of the four sub-quadrants respectively.

[0138] - A list of features contained in the area of ​​this node (for base data layers and vector layers, that is, the corresponding data features overlapping with it).

[0139] The construction process of the quadtree is as follows:

[0140] 1. Start from the root node, which represents the entire PV power station area R.

[0141] 2. For each node N, if the number of elements it contains exceeds a certain threshold T (for example, T = 10, which can be adjusted according to actual conditions), the area R represented by the node N Divided into four equal sub-areas (quadrants):

[0142] Upper left sub-region R1:

[0143] Upper right sub-region R2:

[0144] Lower right sub-region R3:

[0145] Lower left sub-region R4:

[0146] 3. Allocate the elements in node N to the corresponding child nodes, and determine which sub-region the element belongs to based on its spatial position.

[0147] 4. Recursively repeat steps 2 and 3 for each child node until the number of elements contained in all nodes is less than or equal to the threshold T.

[0148] After the quadtree index is established, for each polygon object P in the vector layer, find the area R that overlaps with it in the basic data layer. The specific process is as follows:

[0149] For a polygon object P, its boundary can be represented by a series of vertex coordinates (x i ,y i ), i = 1,…,n represents (in the photovoltaic power plant scenario, these vertex coordinates define the shape of the photovoltaic module).

[0150] First, starting from the root node of the quadtree, determine whether polygon P is consistent with the region R represented by the root node. N There is overlap. The judgment method is to check whether the boundary of polygon P overlaps with the rectangular area R N The boundaries of the two polygons intersect. The Separating Axis Theorem (SAT) can be used to determine whether two polygons (here one is a rectangle and the other is a PV panel polygon) intersect.

[0151] The separating axis theorem states that two convex polygons do not intersect if and only if there exists an axis such that the projections of the two polygons on that axis do not overlap. For the two-dimensional case, we only need to check the normal vectors of the polygon's edges as the separating axis.

[0152] Rectangular area R NThe normal vector of the edge is The normal vector of the edge of polygon P is For each normal vector (include and ), calculate polygon P and rectangular area R N The projection onto this normal vector.

[0153] set up is a normal vector, and the vertex coordinates of polygon P are (x i ,y i ), the minimum and maximum values ​​of the projection are:

[0154]

[0155] For the rectangular area R N , and similarly calculate its normal vector The minimum and maximum values ​​of the projection on r min and r max If for all normal vectors All have p max ≥r min And r max ≥p min , then the polygon P and the rectangular area R N There is overlap.

[0156] If polygon P overlaps with the root node area, the child nodes of the root node are recursively checked until all nodes overlapping with polygon P are found. The areas represented by these nodes are the areas R overlapping with polygon P in the base data layer.

[0157] After the overlapping region R is found, the attribute information of the polygonal object P (such as the number, model, power, etc. of the photovoltaic module) is associated with the data of the overlapping region R.

[0158] The data in the basic data layer B is stored in the form of grids, and each grid cell has certain attributes (such as elevation, image pixel value, etc.). For grid cells overlapping with polygon P, new fields can be added to their attribute tables to store the attribute information of polygon P.

[0159] For example, a "component number" field is added to the raster attribute table of the basic data layer, and for the raster cells overlapping with polygon P, the component number of polygon P is filled into the field. In this way, through this attribute association, the vector layer and the basic data layer are superimposed, and a digital photovoltaic power station model is generated.

[0160] Optionally, the acquiring technical parameters of each component of the photovoltaic power station to map them to the digital photovoltaic power station model includes:

[0161] Classify the components of the photovoltaic power station, including photovoltaic panels, inverters, combiner boxes, brackets, etc. For each type of component, determine the technical parameters that need to be collected;

[0162] Parsing the technical parameter description document to collect the technical parameters;

[0163] The spatial entity corresponding to each type of component is determined, and the spatial entity is mapped to the digital photovoltaic power station model through a unique identifier.

[0164] Specifically, the steps of obtaining the technical parameters of each component of the photovoltaic power station and mapping them to the digital photovoltaic power station model are exemplified as follows:

[0165] In photovoltaic power stations, there are various types of components, and the functions and characteristics of different types of components determine the need to collect different technical parameters.

[0166] For photovoltaic panels, their main function is to convert solar energy into electrical energy. The key technical parameters include:

[0167] Rated power P rated :Indicates that under standard test conditions (STC, usually the light intensity is 1000W / m 2 , battery temperature 25℃, air mass AM=1.5) The maximum power that a photovoltaic panel can output. The unit is watt (W).

[0168] Open circuit voltage V oc :The voltage across the photovoltaic panel when it is not connected to a load under sunlight conditions. The unit is volt (V).

[0169] Short circuit current I sc :The current when both ends of the photovoltaic panel are short-circuited under light conditions. The unit is ampere (A).

[0170] Conversion efficiency η: The efficiency of photovoltaic panels in converting solar energy into electrical energy, calculated as Where A is the effective area of ​​the photovoltaic panel (in m 2 ), G is the light intensity (unit W / m 2 ).

[0171] For the inverter, which converts the direct current generated by the photovoltaic panel into alternating current, the technical parameters that need to be collected are:

[0172] Rated capacity S rated : The rated apparent power that the inverter can output, in volt-amperes (VA) or kilovolt-amperes (kVA).

[0173] Conversion efficiency η inv: The efficiency of the inverter in converting DC power into AC power, calculated by the formula Where P ac is the AC power output by the inverter, P dc is the input DC power.

[0174] Maximum input voltage V max-in and the maximum input current I max-in : The maximum input DC voltage and current that the inverter can withstand.

[0175] The combiner box is used to collect the current of multiple photovoltaic panels and output them in aggregate. The relevant technical parameters include:

[0176] Rated input current I rated-in : The rated current that each input port can withstand.

[0177] Rated output current I rated-out : Rated current of the combiner box output port.

[0178] Number of input channels N in : The number of input ports of the combiner box.

[0179] The bracket is mainly used to support photovoltaic panels. The technical parameters are:

[0180] Load-bearing capacity W load : The maximum weight that the bracket can bear, in kilograms (kg).

[0181] Wind resistance level grade : Indicates the maximum wind force level that the bracket can withstand, which is determined by relevant indicators such as wind speed v$ (unit: m / s). Different standards have different corresponding relationships. For example, under a certain standard, wind resistance level 10 corresponds to a wind speed of 24.5-28.4m / s.

[0182] The above technical parameters are provided in the form of documents, such as product manuals, technical specifications, etc. Parsing these documents to collect technical parameters requires natural language processing and information extraction technology.

[0183] Document D is a text containing technical parameter information, which can be represented as a word sequence D = {w1,w2,…,w n}, where w i is the i-th word in the document.

[0184] In order to extract the technical parameters of a specific component, you first need to classify the document to determine which component the technical document belongs to. You can use a text classification algorithm such as a support vector machine (SVM). (can be extracted from the text through the bag-of-words model), the classification decision function of SVM is:

[0185]

[0186] Where: α i is the Lagrange multiplier. i is the category label of the training sample (for photovoltaic panel document y i =1, other component documents set different labels according to the situation). It is a kernel function, commonly used is radial basis function (RBF) γ is the kernel function parameter. b is the bias term.

[0187] After determining the component category to which the document belongs, it is necessary to extract specific technical parameters from the document. Taking the extraction of the rated power of photovoltaic panels as an example, regular expression matching technology can be used. The expression pattern of rated power in the document is "rated power: [number] [unit]", and the regular expression rated power: (\d+)(W|kW) can be defined, where \d+ means matching one or more numbers, and (W|kW) means matching the power unit watt (W) or kilowatt (kW). By applying this regular expression in the document, you can find the string containing the rated power value and then extract the specific value.

[0188] Specifically, when performing spatial entity mapping, the spatial entity corresponding to each type of component is determined and mapped to the digital photovoltaic power station model through a unique identifier. The specific process is as follows:

[0189] In the digital photovoltaic power station model, each spatial entity has its position and shape information in three-dimensional space. Taking the photovoltaic panel as an example, its spatial entity can be represented by a three-dimensional rectangle, and its position can be represented by the centroid coordinates (x c ,y c ,z c ) indicates that the length, width and height are l, w and h respectively.

[0190] For each type of component, a unique ID is assigned to it. For example, the identification of a photovoltaic panel can be its number in the photovoltaic array, and the inverter, combiner box, etc. can generate a unique ID based on its installation location and device number.

[0191] In the digital photovoltaic power station model, a mapping table M can be established to associate the unique identification ID of the component with the geometric information and attribute information of the corresponding spatial entity. The mapping table M is a key-value pair collection Where ID i Is the unique identifier of the component. It is the corresponding spatial entity information, including location coordinates, geometric shape, and technical parameters.

[0192] For component classification and technical parameter determination, it is necessary to store the parameter definitions of various components in the database for query and management. The document parsing process can implement text classification and information extraction algorithms by writing program codes. Spatial entity mapping can implement the storage and query operations of the mapping table M through data structures (such as hash tables). In order to improve efficiency, large-scale document data and mapping tables can be processed and stored in a distributed manner, and multi-threading or parallel computing technology can be used to accelerate the text classification and information extraction process.

[0193] Optionally, the automatically generating a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model includes:

[0194] Analyze the preset inspection plan, and plan inspection points and inspection accuracy on the digital photovoltaic power station model;

[0195] Based on the inspection points and the progress of the inspection, determine the dwell time of the drone at each inspection point and the frequency of data collection;

[0196] Based on the dwell time of the drone at each inspection point and the frequency of data collection, the drone inspection path is automatically generated.

[0197] Specifically, when parsing the inspection plan and planning the inspection points and accuracy, the preset inspection plan usually includes the inspection requirements for different areas and components of the photovoltaic power station. The inspection plan is represented as a set S = {s1, s2, ..., s m}, where s i Represents the i-th inspection task. Each task can contain information such as inspection area and inspection focus.

[0198] In the digital photovoltaic power station model, the space of the photovoltaic power station is divided into a discrete point set P = {p1, p2, ..., p n}, these points represent possible inspection points. For each inspection task s i , it is necessary to select appropriate inspection points from the point set P according to the task requirements.

[0199] A method based on importance weights is used to select inspection points. For each point p j , define an importance weight w ij , which represents the point p j For inspection tasks i This weight can be determined based on factors such as the type, location, and fault history of the PV panels. For example, PV panels close to the inverter may have a higher weight because they have a greater impact on the power generation efficiency of the entire power station.

[0200] Inspection points are selected by solving the following optimization problem:

[0201]

[0202] Where P i For inspection tasks i The selected inspection point set, N i It is an inspection task i The maximum number of inspection points allowed.

[0203] Inspection accuracy can be expressed by defining the coverage of each inspection point. j Has a coverage radius r j , this radius depends on the sensor performance of the drone and the inspection requirements. For different types of inspection tasks, the coverage radius may be different. For example, for the task of inspecting surface defects of photovoltaic panels, a smaller coverage radius may be required to achieve higher accuracy; while for inspecting the light intensity of a large area, the coverage radius can be larger.

[0204] Specifically, when determining the dwell time of the UAV at each inspection point and the frequency of data collection, the dwell time t of the UAV at each inspection point is j and data acquisition frequency f j Depends on the inspection accuracy and data collection requirements.

[0205] Residence time t j It can be calculated by the following formula:

[0206] Among them A j It is the inspection point p j The area to be covered depends on the number and layout of the PV panels to be inspected around the point; v is the flight speed of the drone in m / s; r j It is the inspection point p j The coverage radius is in m$; k is an empirical coefficient used to consider the stability and redundancy of data collection, usually k>1.

[0207] Data acquisition frequency f j It can be determined based on the amount of data to be collected and the performance of the sensor. j Collection N data data samples, the time interval between each data collection by the sensor is Δt, then the data collection frequency f j for:

[0208] At the same time, in order to ensure the accuracy and completeness of the data, the data collection frequency f j It is also necessary to satisfy the sampling theorem of the sensor, that is, f j ≥2f max , where f maxIt is the highest frequency of the collected signal. In the inspection of photovoltaic power plants, the collected signals may be the temperature change of photovoltaic modules, the change of light intensity, etc. The highest frequency of these signals can be determined through historical data or theoretical analysis.

[0209] Specifically, when automatically generating the drone inspection path, a variant of the traveling salesman problem (TSP) can be used to generate the drone inspection path based on the drone's residence time and data collection frequency at each inspection point.

[0210] Assume that the set of inspection points is Q = {q1,q2,…,q k}, where q i is the actual inspection point selected. Define the drone from the inspection point q i to q j The flight time T ij for:

[0211] Where d(q i ,q j ) is the inspection point q i and q j The Euclidean distance between them, v is the flight speed of the drone.

[0212] The optimization problem to be solved is: The constraint is that each inspection point q i Visited once.

[0213] To this end, the embodiment of the present application adopts a genetic algorithm to solve the above-mentioned optimal problem, as follows:

[0214] The inspection path is represented as a chromosome, and each gene of the chromosome represents the number of a inspection point. Through genetic operations such as selection, crossover and mutation, the chromosome is continuously optimized to make the total flight time of the corresponding path the shortest.

[0215] For the optimization problem of inspection plan analysis and inspection point selection, linear programming or integer programming solvers can be used. The calculation of residence time and data collection frequency can be realized by writing simple program code. For the TSP problem of generating inspection paths, approximate algorithms such as genetic algorithms can also be implemented by writing code. In order to improve computing efficiency, large-scale inspection point data can be processed in parallel, and multi-threading or distributed computing technology can be used to accelerate the solution process.

[0216] Specifically, the detailed processing steps of automatically generating the inspection path of the UAV using the genetic algorithm can be as follows:

[0217] In the photovoltaic power station inspection scenario, the inspection point set is Q = {q1, q2, ..., q k}, where qi These inspection points are located at different locations of the PV power station, corresponding to the locations of PV modules, combiner boxes and other equipment that need to be inspected.

[0218] Define the drone from the inspection point q i to q j The flight time T ij for:

[0219]

[0220] Among them, d(q i ,q j ) is the inspection point q i and q j The Euclidean distance between them is calculated as (x i ,y i ) and (x j ,y j ) are inspection points q i and q j The coordinates in the plane coordinate system (available in the digital model of the photovoltaic power station); v is the flight speed of the drone in meters per second (m / s), which depends on the performance of the drone and the inspection mission requirements.

[0221] The goal of the optimization is to find a path with the shortest total flight time that passes through each inspection point exactly once and returns to the starting point, that is, to solve the following optimization problem:

[0222] The constraint condition is that for each inspection point q i Visited once.

[0223] The specific process of solving the above genetic algorithm is as follows:

[0224] (1) The inspection path is represented as a chromosome, and each gene of the chromosome represents the number of a inspection point. For example, if there are five inspection points q1, q2, q3, q4, q5, a possible chromosome code is [3, 1, 5, 2, 4], indicating that the inspection path is q3→q1→q5→q2→q4→q3 (back to the starting point).

[0225] (2) Randomly generate a certain number of chromosomes as the initial population. Let the population size be N, and each chromosome is a legal inspection path code. For example, in the case of k inspection points, each chromosome is generated by randomly permuting the numbers from 1 to k.

[0226] (3) Define a fitness function to evaluate the quality of each chromosome (i.e., each inspection path). The fitness function F can be the inverse of the total flight time, that is: The shorter the total flight time, the higher the fitness value, which means the path is better.

[0227] (4) Use the roulette wheel selection method to select excellent chromosomes from the population to enter the next generation. Calculate the proportion of each chromosome's fitness to the total fitness p i : Among them, F i is the fitness of the i-th chromosome, and N is the population size. Based on these ratios, chromosomes are selected by roulette, and chromosomes with high fitness are more likely to be selected.

[0228] (5) Crossover operation is one of the core operations of genetic algorithm, which simulates the gene exchange process in biological evolution. Commonly used crossover methods include Order Crossover (OX). Select two parent chromosomes P1 = [p 11 ,p 12 ,…,p 1k ] and P2 = [p 21 ,p 22 ,…,p 2k ], to perform sequential crossover:

[0229] 1. Randomly select two intersection points c1 and c2 (1≤c1 <c2≤k)。

[0230] 2. Extract the gene fragment between c1 and c2 from P1 and obtain

[0231] 3. Starting from position c2 of P2, fill in the genes in S that are not in the c1 to c2 segment of P2 in order to obtain the daughter chromosome C1.

[0232] 4. Similarly, cut the gene segment between c1 and c2 from P2, and start from the c2 position of P1, fill in the genes in the segment that are not in the c1 to c2 segment of P1 in order to obtain the daughter chromosome C2.

[0233] (5) The mutation operation randomly changes the genes of the chromosome with a certain probability to prevent the algorithm from falling into a local optimum. The mutation probability is usually set to a small value, such as 0.01. The mutation method can be to randomly exchange the positions of two genes. For example, for chromosome [3,1,5,2,4], if the two gene positions randomly selected for mutation are 2 and 4, the mutated chromosome is [3,2,5,1,4].

[0234] (6) Iteration and termination conditions Repeat the selection, crossover and mutation operations to generate a new population. After each generation of population is generated, the fitness of each chromosome is recalculated. When a certain termination condition is met, the algorithm stops. The termination condition can be reaching the maximum number of iterations G. max, or the average fitness of the population has not increased significantly in several consecutive generations.

[0235] To this end, a genetic algorithm is used to automatically generate UAV inspection paths based on the inspection point dwell time and data collection frequency, which has the following technical benefits:

[0236] In the photovoltaic power station inspection scenario, the use of genetic algorithms to automatically generate drone inspection paths has many significant technical benefits:

[0237] (1) Traditional path planning methods may easily fall into local optimal solutions, while genetic algorithms search the entire solution space by simulating the biological evolution process. It starts with a set of randomly generated initial paths (chromosomes) and continuously explores and optimizes the paths through operations such as selection, crossover, and mutation. In the inspection of photovoltaic power plants, this means being able to find the path with the shortest total flight time in theory among many possible inspection path combinations, ensuring that the drone completes the inspection task in the most efficient way, thereby reducing inspection time and energy consumption.

[0238] (2) The layout of a PV power station may be complex, with irregular distribution of inspection points in different areas and various restrictions. Genetic algorithms can handle this complex situation by automatically adapting to different inspection environments and task requirements through random encoding and continuous evolution of inspection points. For example, even if some new equipment or areas that need to be inspected are added to a PV power station, the genetic algorithm can quickly adjust and generate a new optimal inspection path.

[0239] (3) Although the genetic algorithm is a heuristic algorithm and cannot guarantee to find the absolutely optimal solution every time, in practical applications, it can usually find a solution close to the optimal solution in a relatively short time. For large-scale photovoltaic power plants, there are a large number of inspection points. Using traditional exact algorithms to solve the traveling salesman problem (TSP) may take a lot of time and computing resources, or even fail to produce results within a reasonable time. By searching multiple solution spaces in parallel, the genetic algorithm can quickly converge to a better path solution, meeting the time requirements of actual inspection tasks.

[0240] The genetic algorithm is less dependent on the initial solution, and different initial populations are likely to find a better solution through evolution. This allows the algorithm to stably generate effective inspection paths when faced with different photovoltaic power plant layouts and inspection task settings. Moreover, since the algorithm itself has a certain degree of randomness, different results may be obtained each time it is run, but these results can usually meet actual needs and will not cause a significant drop in performance due to slight changes in initial conditions.

[0241] (3) Each operation step of the genetic algorithm (such as selection, crossover, and mutation) has multiple implementation methods, and the performance and behavior of the algorithm can be controlled by adjusting parameters (such as population size, crossover probability, mutation probability, etc.). In the inspection of photovoltaic power plants, these parameters can be flexibly adjusted according to actual conditions (such as inspection time limit, equipment importance, etc.) to generate the most suitable inspection path. For example, if the inspection time requirement is high, the mutation probability can be appropriately increased to speed up the convergence of the algorithm; if a more accurate optimal solution is desired, the population size can be increased and the number of iterations can be extended.

[0242] (5) As the scale of photovoltaic power plants expands and the demand for inspections increases, more factors may need to be considered in path planning, such as the power limit of drones and the inspection priorities of different areas. Genetic algorithms have good scalability and can easily incorporate these additional constraints into the fitness function or algorithm implementation process. For example, a penalty term for power consumption can be added to the fitness function, or different weights can be set for inspection points of different priorities, so that the generated inspection path is more in line with actual needs.

[0243] (6) In the process of generating inspection paths, the genetic algorithm makes full use of the inspection point coordinate information provided in the digital model of the photovoltaic power station. Through the effective processing and analysis of these data, the optimal inspection sequence can be discovered and the inspection efficiency can be improved. At the same time, the operation process of the algorithm can also provide a reference for subsequent data analysis. For example, by observing the evolution of different generations of populations, the optimization trend of the inspection path can be understood, providing a basis for further improving the algorithm or adjusting the inspection strategy.

[0244] Optionally, the acquisition of monitoring data acquired by the drone along the inspection path of the drone, the monitoring data including the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station, includes:

[0245] Acquire monitoring data acquired by the drone along the inspection route of the drone, and separate the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station;

[0246] The appearance images, temperature distribution images and light intensity data collected in the same area and at the same time are associated and integrated according to the time sequence and location information to form a comprehensive detection data set.

[0247] Specifically, when acquiring and separating monitoring data, considering that the drone flies along the preset inspection route, the various sensors carried by it will simultaneously acquire various monitoring data. These data are usually packaged in a unified format during transmission and need to be separated.

[0248] The original monitoring data obtained by the drone is a data stream D$ containing a variety of information, which can be represented as a high-dimensional vector set D={d1,d2,…,d n}, where d i It is a data unit in a data stream.

[0249] For the separation of photovoltaic array appearance images: the appearance image data has a specific format and identification in the data stream. Assume that the image data format follows a certain image coding standard (such as the JPEG standard) and is represented in the data stream with a specific header information H img By searching for H in data stream D img To locate the starting position of the appearance image data. Find the starting position as s img , according to the length information L of the image data img (can be obtained from header information or other related metadata), the appearance image data I can be separated array ,Right now

[0250] Separation of PV module temperature distribution data: Temperature distribution data is usually obtained by thermal imaging sensors, which also have specific identifiers in the data stream. Let the temperature data be identified as H temp , find its starting position s in the data stream temp The length of the temperature data is L temp It can also be obtained from the relevant metadata. Then the separated temperature distribution data T is In practical applications, temperature distribution data may be represented in matrix form, where each matrix element corresponds to a temperature value of a photovoltaic module or a certain area.

[0251] Separation of light intensity data in different areas of the photovoltaic power station: Light intensity data is obtained by light sensors, marked as H light . Find its starting position s in the data stream light and length L light , thereby separating the light intensity data E$, that is, Light intensity data may be represented as a discrete sequence of values, each of which corresponds to a light intensity measurement at a specific location.

[0252] Furthermore, the separated data need to be associated and integrated according to time sequence and location information to form a comprehensive detection data set.

[0253] For time and location information extraction: When the drone acquires data, it will simultaneously record the acquisition time t and location information P = (x, y, z) of each data point. array, temperature distribution data T and light intensity data E$, the acquisition time t can be extracted from their corresponding metadata img ,t temp ,t light and location information P img , P temp , P light .

[0254] In order to associate the data collected in the same area and at the same time, a similarity measurement function S is defined to measure the similarity of different data points in time and location.

[0255] For temporal similarity, define:

[0256]

[0257] Where t1 and t2 are the acquisition times of two data points, σ t It is a temporal similarity parameter, which is determined by factors such as the flight speed of the UAV and the frequency of data collection. t The closer the value is to 1, the more similar the two time points are.

[0258] For location similarity, define:

[0259]

[0260] Where P1 = (x1, y1, z1) and P2 = (x2, y2, z2) are the position coordinates of the two data points. is the Euclidean distance, σ p is a location similarity parameter, which is determined based on factors such as the scale of the photovoltaic power station and the positioning accuracy of the drone. p The closer the value is to 1, the more similar the two locations are.

[0261] Combining the similarity of time and location, we define the overall similarity function: S = S t ·S p .

[0262] For the appearance image I array For each data point in the temperature distribution data T and the light intensity data E, we traverse all the data points and calculate the similarity S$ between them. When S is greater than a certain threshold θ (set according to the actual situation, for example, θ = 0.8), we consider that the three data points are collected in the same area and at the same time, and associate them.

[0263] Specifically, when forming a comprehensive detection data set, the associated data points are combined into a comprehensive detection data unit D unit =(I array,T,E). All such integrated detection data units constitute the integrated detection data set D = {D unit1 ,D unit2 ,…,D unitm}.

[0264] Therefore, in the photovoltaic power station monitoring scenario, the above method for acquiring and separating monitoring data and integrating them has the following technical advantages:

[0265] (1) Assuming that the original monitoring data is packaged in a unified format, the data obtained by different types of sensors can be transmitted in the same data stream. This method has strong compatibility and versatility. No matter what new sensor is installed or different monitoring technologies are adopted, as long as the unified data format standard is followed, it can be easily integrated into the existing data acquisition system. This facilitates the expansion and upgrade of the photovoltaic power station monitoring system and reduces the complexity and maintenance cost of the system.

[0266] (2) By locating and separating different types of data through specific formats, identifiers, and header information, the required monitoring data can be accurately extracted from complex data streams. For example, by separating the appearance image data of photovoltaic arrays based on image coding standards and header information, image information can be accurately identified and extracted, avoiding interference from other irrelevant data. This precise data extraction method improves the efficiency and accuracy of data processing and provides a reliable data basis for subsequent data analysis and decision-making.

[0267] (3) Separate and process different types of data so that each type of data has its own independent structure and representation. For example, temperature distribution data is represented in matrix form, and light intensity data is represented in discrete numerical sequences. This clear data structure facilitates targeted processing and analysis of different types of data. For example, the matrix form of temperature distribution data facilitates the application of image processing and analysis algorithms to identify abnormal temperature areas; discrete light intensity numerical sequences are suitable for statistical analysis and trend prediction.

[0268] (4) By correlating and integrating different types of data collected in the same area and at the same time, the operating status of the photovoltaic power station can be fully reflected from multiple dimensions. For example, by correlating appearance images, temperature distribution and light intensity data, the impact of lighting conditions on the temperature and appearance of photovoltaic modules can be comprehensively analyzed. This multi-dimensional analysis helps to discover potential problems, such as local overheating caused by uneven lighting, so that measures can be taken in advance to ensure the stable operation of the photovoltaic power station.

[0269] (5) Use the similarity metric function to measure the similarity of data points in time and location, so as to achieve accurate data association. This association method based on time and space fully considers the actual data collection background, making the association results more accurate and reliable. Compared with simple time or location matching, comprehensive consideration of time and location similarity can avoid erroneous associations caused by time errors or location deviations, and improve the accuracy and credibility of data analysis.

[0270] (6) The data association and integration process helps to explore the potential relationships between different types of data. For example, by analyzing the integrated detection data after association, the intrinsic relationship between the changes in the appearance of photovoltaic modules and temperature and light can be found, providing valuable information for the performance optimization and fault prediction of photovoltaic power plants. This kind of mining of potential relationships in data can help operation and maintenance personnel better understand the operating mechanism of photovoltaic power plants and formulate more scientific operation and maintenance strategies.

[0271] Optionally, vectorizing the monitoring data to obtain an input vector includes:

[0272] According to the different types of data contained in the monitoring data, a vectorization model is matched to input the appearance image and the temperature distribution image into the feature extractor for vectorization to obtain an image data vector, and the light intensity data and the temperature value are input into the statistical feature extraction model for vectorization to obtain a numerical data vector;

[0273] The image data vector and the numerical data vector are combined to form the input vector.

[0274] Specifically, for appearance images and temperature distribution images, we need to use a feature extractor to convert them into image data vectors. Here we use a convolutional neural network (CNN) as a feature extractor, which has a strong ability in image feature extraction.

[0275] Let the input appearance image be I array , whose size is H×W×C, where H is the height of the image, W is the width of the image, and C is the number of channels of the image (for example, C=3 for RGB image); the temperature distribution image is T, and its size is also H T ×W T ×C T (Thermal imaging may be C T =1).

[0276] Construct a CNN model, which consists of multiple convolutional layers, pooling layers and fully connected layers. The convolution kernel size of the convolutional layer is k×k, the step size is s, and the padding is p. For a convolutional layer l, the input feature map X l-1 The size is H l-1 ×W l-1 ×Cl-1 , after the convolution operation, the output feature map X l The dimensions are:

[0277]

[0278] C l =F

[0279] Where F is the number of convolution kernels.

[0280] The pooling layer is usually used to reduce the size of the feature map. Taking the maximum pooling as an example, the pooling kernel size is m×m and the step size is t. For the input feature map X l , after the pooling operation, the output feature map Y l The dimensions are:

[0281]

[0282] After multiple convolutional and pooling layers, the feature map is flattened into a one-dimensional vector. The feature map output by the last pooling layer has a size of H f ×W f ×C f , the length of the flattened vector is L img =H f ×W f ×C f , this vector is the feature vector obtained by the CNN feature extractor of the appearance image

[0283] Similarly, for the temperature distribution image T$, after being processed by the same CNN feature extractor, the feature vector

[0284] Concatenate these two image feature vectors to get the image data vector The “;” indicates the vector concatenation operation.

[0285] Specifically, when vectorizing numerical data, for the light intensity data E and the temperature value T value , vectorized using a statistical feature extraction model.

[0286] Assume that the light intensity data E = {e1, e2, ..., e n}, temperature value T value ={t1,t2,…,t m}.

[0287] A variety of statistical features can be extracted, such as mean, variance, maximum, minimum, etc.

[0288] The mean μ of the light intensity data E for: variance for:

[0289] Maximum value E =max{e1,e2,…,e n}, minimum value min E =min{e1,e2,…,e n}.

[0290] Similarly, for the temperature value T value , whose mean value μ T ,variance Maximum value T and minimum value min T It can also be calculated according to the above formula.

[0291] Combine these statistical features into a numerical data vector

[0292] Specifically, when the input vector is formed, the image data vector and a vector of numerical data Combine to get the input vector

[0293] Therefore, in the scenario of photovoltaic power station operation status monitoring, the above method of extracting features from appearance images and temperature distribution images and vectorizing numerical data to finally form an input vector has many significant technical benefits as follows:

[0294] 1. Convolutional neural networks (CNNs) have powerful local feature extraction capabilities and can automatically learn features such as texture, shape, and structure in images. In photovoltaic power plants, appearance images contain information such as the surface condition and connection status of photovoltaic modules, and temperature distribution images reflect the thermal state of the modules. Through the convolution layer of CNN, features closely related to the operating status of photovoltaic power plants can be extracted from these images, such as abnormal features such as cracks and hot spots in modules, providing key information for subsequent fault diagnosis and performance evaluation.

[0295] 2. The use of the pooling layer allows the feature map to retain important feature information while reducing its size, achieving size invariance of image features. This means that no matter how the position and size of the photovoltaic module in the image change, CNN can extract similar features, improving the robustness of the model. At the same time, through multiple operations of the convolution layer and the pooling layer, the high-dimensional image data is gradually converted into a low-dimensional feature vector, reducing the dimension of the data, reducing the computational complexity of subsequent processing, and improving computational efficiency.

[0296] 3. CNN is an end-to-end learning model that does not require manual design of feature extraction methods. In photovoltaic power station monitoring, different photovoltaic power stations may have different environmental conditions, component layouts, and operating conditions. CNN can automatically learn effective feature representations from a large number of appearance images and temperature distribution images to adapt to different application scenarios, reduce the workload of manual intervention and feature engineering, and improve the adaptability and generalization ability of the model.

[0297] 4. After the appearance image and the temperature distribution image are processed by the same CNN feature extractor, the obtained feature vectors are spliced ​​to achieve the fusion of different types of image features. This fusion can comprehensively consider the appearance and thermal status information of the photovoltaic module, provide a more comprehensive feature representation, and help to more accurately analyze the operating status of the photovoltaic power station. For example, combining the physical damage characteristics of the module in the appearance image and the thermal anomaly characteristics in the temperature distribution image can more accurately determine whether the module has potential faults.

[0298] 5. For light intensity data and temperature values, extracting statistical features such as mean, variance, maximum, and minimum can simplify a large amount of raw numerical data into several representative statistics. These statistics summarize the main characteristics and distribution of the data, extract key information, and make subsequent analysis and processing more efficient. For example, the mean of light intensity can be used to understand the overall light level of a photovoltaic power station, and the variance can reflect the stability of light. This information is of great significance for evaluating the power generation efficiency and stability of a photovoltaic power station.

[0299] 6. The statistical characteristics of numerical data and image features complement each other, providing a more comprehensive perspective for the analysis of the operating status of photovoltaic power plants. Image features mainly reflect the spatial distribution information of the appearance and thermal status of photovoltaic modules, while the statistical characteristics of numerical data describe the changes in light intensity and temperature as a whole. Combining the two can provide a deeper understanding of the operating mechanism of photovoltaic power plants and improve the accuracy of fault diagnosis and performance evaluation.

[0300] 7. Combine the image data vector and the numerical data vector to form an input vector containing multiple information, thus realizing the fusion of multi-source data. This fusion can make full use of the appearance image, temperature distribution image, light intensity and temperature value and other data information, providing a richer and more comprehensive input for the photovoltaic power station operation status monitoring model, which helps to improve the accuracy and reliability of the model.

[0301] Optionally, the operation status monitoring model performs feature extraction on the input vector to analyze the operation status of the photovoltaic power station, including:

[0302] Inputting the input vector into the operation status monitoring model for forward propagation, gradually extracting features at different levels in each layer, and obtaining feature representations of valuable information about the operation status of the photovoltaic power station;

[0303] The operating status of the photovoltaic power station is analyzed according to the characteristic representation.

[0304] Specifically, during the forward propagation and feature extraction of the running state monitoring model, the running state monitoring model is a deep neural network (DNN) containing multiple hidden layers L1, L2, ..., L N , the input vector is

[0305] For the first layer L1, let its weight matrix be W1 and its bias vector be b1. Input vector The linear transformation after the first layer is: in is the result vector after linear transformation, the dimension of W1 is m1×d, m1 is the number of neurons in the first layer, and d is the input vector The dimension of b1 is m1×1.

[0306] In order to introduce nonlinearity, an activation function σ(·) is used, such as the ReLU (Rectified Linear Unit) function:

[0307] σ(x)=max(0,x), after the activation function, the output of the first layer for:

[0308] For the subsequent lth layer (2≤l≤N), the same linear transformation and activation function processing are performed. Let the weight matrix of the lth layer be W l , the bias vector is b l ,but: in is the output of layer l-1, W l The dimension is m l ×m l-1 , m l is the number of neurons in layer l, b l The dimension is m l ×1.

[0309] After N layers of forward propagation, the final output is It is a feature representation of valuable information about the operating status of a photovoltaic power station. In the scenario of a photovoltaic power station, features at different levels represent different aspects of the operating status of the photovoltaic power station. For example, shallower features may represent local appearance features of photovoltaic modules, such as whether there are stains or cracks; deeper features may represent the overall operating mode of the photovoltaic power station, such as the comprehensive impact of light intensity and temperature distribution in different areas on power generation efficiency.

[0310] Specifically, when analyzing the operating status of a photovoltaic power station according to the characteristic representation, the characteristic representation is obtained. After that, further analysis is needed to determine the operating status of the photovoltaic power station. The operating status is divided into normal state and abnormal state, and a classifier is constructed to make judgments.

[0311] Here we use a logistic regression classifier, and assume that the weight vector of the classifier is The bias is b$. Then the predicted probability p is: in yes The transpose of .

[0312] If p≥0.5, the PV power station is judged to be in normal operation; if p<0.5, it is judged to be in abnormal operation.

[0313] In practical applications, in order to analyze the operating status more accurately, support vector machine (SVM) may be used to achieve this. Taking SVM as an example, a linear kernel function is used for feature representation. The decision function is:

[0314] where sgn(·) is the sign function. hour, Indicates normal operating status; when hour, Indicates an abnormal operating status.

[0315] Therefore, in the operation status monitoring of photovoltaic power plants, the above-mentioned deep neural network (DNN) is used for forward propagation and feature extraction, and logistic regression or support vector machine (SVM) is used for operation status analysis, which has the following technical benefits:

[0316] 1. Deep neural networks can automatically learn complex feature representations from input vectors. In the photovoltaic power station scenario, the operating status is affected by a combination of factors, such as the appearance of photovoltaic components, temperature distribution, light intensity, etc. The multiple hidden layers of DNN can extract features at different levels layer by layer, from the shallower local appearance features of photovoltaic components to the deeper overall operating mode features of the power station. This ability to automatically learn complex features enables the model to capture the laws and patterns hidden in the data without the need for manual design of a large number of features, greatly improving the efficiency and accuracy of feature extraction.

[0317] 2. By using activation functions (such as ReLU functions), DNN can introduce nonlinear transformations, thereby mapping linearly inseparable data to high-dimensional space, making it linearly separable. In the operation status monitoring of photovoltaic power plants, the boundaries between different operating states are often nonlinear, and traditional linear models are difficult to accurately distinguish. The nonlinear mapping capability of DNN can effectively handle this complex nonlinear relationship and improve the model's ability to classify and predict the operation status of photovoltaic power plants.

[0318] 3. The multi-layer structure of DNN makes the feature representation hierarchical. The shallower features focus more on local and detailed information, such as stains and cracks on the surface of photovoltaic modules, while the deeper features can better reflect the overall and abstract information, such as the comprehensive impact of light intensity and temperature distribution in different regions on power generation efficiency. This hierarchical feature representation can comprehensively describe the operating status of photovoltaic power stations and provide rich information for subsequent analysis and decision-making.

[0319] 4. In the operation status monitoring of photovoltaic power stations, it is necessary to quickly analyze and judge a large amount of monitoring data. The simplicity and efficiency of logistic regression can meet the real-time requirements and provide timely feedback on the operation status. The parameters of the logistic regression model have clear physical meanings. By analyzing the weight vector and bias, we can intuitively understand the degree of influence of each feature on the operation status. In the operation and maintenance management of photovoltaic power stations, this easy-to-understand and easy-to-interpret feature enables operation and maintenance personnel to quickly grasp the decision-making basis of the model and accurately analyze and handle abnormal situations. The output of the logistic regression classifier is the probability that the sample belongs to a certain category, which has great advantages in practical applications. For example, when the prediction result is close to 0.5, it means that there is a certain uncertainty in the operation status of the photovoltaic power station, which requires further attention and analysis. This probability output can provide operation and maintenance personnel with richer information and help them make more reasonable decisions.

[0320] 5: SVM is good at classification in high-dimensional space and can effectively handle complex nonlinear classification problems. In the operation status monitoring of photovoltaic power plants, feature representation is often high-dimensional. SVM can map data to high-dimensional space through kernel functions and find the optimal classification hyperplane, thereby accurately distinguishing normal operation from abnormal operation. Compared with other classification algorithms, SVM has better classification performance on high-dimensional data.

[0321] SVM finds the optimal classification hyperplane by maximizing the classification interval, which makes it highly resistant to noise and outliers. In the monitoring data of photovoltaic power plants, there may be some noise data caused by sensor errors, environmental interference and other factors. SVM can ignore these noises to a certain extent and maintain the accuracy and stability of classification.

[0322] In actual photovoltaic power station operation status monitoring, it may be impossible to obtain a large amount of labeled data due to limitations in data collection costs, time, and other factors. SVM performs well in small sample learning. It can use limited sample data to find the optimal classification model, avoid overfitting problems, and improve the generalization ability of the model. This allows SVM to effectively analyze and judge the operation status of photovoltaic power stations even when the amount of data is relatively small.

[0323] The embodiment of the present application also provides a photovoltaic power station operation status monitoring device based on drone inspection, which includes:

[0324] The first program unit is used to map the geographical location, topography and photovoltaic array layout of the photovoltaic power station by using surveying and mapping technology and geographic information system to establish a digital photovoltaic power station model;

[0325] A second program unit is used to obtain technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0326] A third program unit is used to automatically generate a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model;

[0327] A fourth program unit is used to obtain monitoring data obtained by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic module, and light intensity data of different areas of the photovoltaic power station;

[0328] A fifth program unit, used for vectorizing the monitoring data to obtain an input vector;

[0329] The sixth program unit is used to run the status monitoring model to extract features from the input vector to analyze the operating status of the photovoltaic power station.

[0330] For an exemplary explanation of the functions performed by each unit in the above device embodiment, see the above Figure 1 Records of.

[0331] The present application also provides a computer program product having computer executable instructions stored thereon, and the computer executable instructions are executed to perform the following steps:

[0332] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0333] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0334] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0335] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0336] Vectorizing the monitoring data to obtain an input vector;

[0337] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0338] For an exemplary explanation of the functions performed by each unit in the above computer program product embodiment, see the above Figure 1 Records of.

[0339] The embodiment of the present application further provides an electronic device, which includes a memory and a processor, wherein the memory stores computer executable instructions, and the processor runs the computer executable instructions to perform the following steps:

[0340] Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model;

[0341] Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model;

[0342] Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model;

[0343] Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station;

[0344] Vectorizing the monitoring data to obtain an input vector;

[0345] The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

[0346] For an exemplary explanation of the functions performed by each unit in the above electronic device embodiment, see the above Figure 1 Records of.

[0347] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A method for monitoring the operating status of a photovoltaic power station based on drone inspection, characterized in that: include: Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model; Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model; Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model; Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station; Vectorizing the monitoring data to obtain an input vector; The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

2. A method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1, characterized in that: The use of surveying and mapping technology and geographic information system to survey and map the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model includes: Use surveying and mapping technology and geographic information systems to survey the geographical location, topography, and photovoltaic array layout of the photovoltaic power station to generate geographical location coordinate data, digital elevation models, digital orthophotos, and measurement data of the photovoltaic array layout; Generate a basic data layer with spatial location attributes based on the geographic location coordinate data, digital elevation model, digital orthophoto map and measurement data of photovoltaic array layout; Imported PV array layout data, creating vector layers to represent different PV module entities; The vector layer is superimposed on the basic data layer to establish a digital photovoltaic power station model.

3. The method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1 is characterized in that: The obtaining of technical parameters of each component of the photovoltaic power station to map them to the digital photovoltaic power station model includes: Classify the components of the photovoltaic power station, including photovoltaic panels, inverters, combiner boxes, and brackets, and determine the technical parameters that need to be collected for each type of component; Parsing the technical parameter description document to collect the technical parameters; The spatial entity corresponding to each type of component is determined, and the spatial entity is mapped to the digital photovoltaic power station model through a unique identifier.

4. The method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1 is characterized in that: The method of automatically generating a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model includes: Analyze the preset inspection plan, and plan inspection points and inspection accuracy on the digital photovoltaic power station model; Based on the inspection points and the progress of the inspection, determine the dwell time of the drone at each inspection point and the frequency of data collection; Based on the dwell time of the drone at each inspection point and the frequency of data collection, the drone inspection path is automatically generated.

5. The method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1 is characterized in that: The monitoring data acquired by the drone along the inspection path of the drone includes the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station, including: Acquire monitoring data acquired by the drone along the inspection route of the drone, and separate the appearance image of the photovoltaic array, the temperature distribution of the photovoltaic components, and the light intensity data of different areas of the photovoltaic power station; The appearance images, temperature distribution images and light intensity data collected in the same area and at the same time are associated and integrated according to the time sequence and location information to form a comprehensive detection data set.

6. The method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1, characterized in that: The step of vectorizing the monitoring data to obtain an input vector includes: According to the different types of data contained in the monitoring data, a vectorization model is matched to input the appearance image and the temperature distribution image into the feature extractor for vectorization to obtain an image data vector, and the light intensity data and the temperature value are input into the statistical feature extraction model for vectorization to obtain a numerical data vector; The image data vector and the numerical data vector are combined to form the input vector.

7. A method for monitoring the operation status of a photovoltaic power station based on drone inspection according to claim 1, characterized in that: The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station, including: Inputting the input vector into the operation status monitoring model for forward propagation, gradually extracting features at different levels at each layer, and obtaining feature representations of valuable information about the operation status of the photovoltaic power station; The operating status of the photovoltaic power station is analyzed according to the characteristic representation.

8. A photovoltaic power station operation status monitoring device based on drone inspection, characterized in that: include: The first program unit is used to map the geographical location, topography and photovoltaic array layout of the photovoltaic power station by using surveying and mapping technology and geographic information system to establish a digital photovoltaic power station model; A second program unit is used to obtain technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model; A third program unit is used to automatically generate a drone inspection path based on a preset inspection plan and the digital photovoltaic power station model; A fourth program unit is used to obtain monitoring data obtained by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, a temperature distribution of the photovoltaic module, and light intensity data of different areas of the photovoltaic power station; A fifth program unit, used for vectorizing the monitoring data to obtain an input vector; The sixth program unit is used to run the status monitoring model to extract features from the input vector to analyze the operating status of the photovoltaic power station.

9. A computer program product having computer executable instructions stored thereon, wherein the computer executable instructions are executed to perform the following steps: Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model; Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model; Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model; Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station; Vectorizing the monitoring data to obtain an input vector; The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

10. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores computer executable instructions, and the processor runs the computer executable instructions to perform the following steps: Use surveying and mapping technology and geographic information systems to survey the location, topography, and photovoltaic array layout of the photovoltaic power station to establish a digital photovoltaic power station model; Acquiring technical parameters of each component of the photovoltaic power station to map them onto the digital photovoltaic power station model; Automatically generate a drone inspection route based on a preset inspection plan and the digital photovoltaic power station model; Acquire monitoring data acquired by the drone along the inspection path of the drone, wherein the monitoring data includes an appearance image of the photovoltaic array, temperature distribution of the photovoltaic components, and light intensity data of different areas of the photovoltaic power station; Vectorizing the monitoring data to obtain an input vector; The operation status monitoring model extracts features from the input vector to analyze the operation status of the photovoltaic power station.

Images