Photovoltaic power station inspection method and system based on unmanned aerial vehicle and GIS

Abstract

The invention belongs to the technical field of unmanned aerial vehicle inspection, and particularly provides a photovoltaic power station inspection method and system based on an unmanned aerial vehicle and a GIS, and the method comprises the steps: obtaining initial geographic information data of a photovoltaic power station, enabling the unmanned aerial vehicle to execute initial surveying and mapping flight, collecting high-resolution image data, and constructing a three-dimensional model of the photovoltaic power station; fusing the high-resolution image data with the initial geographic information data to generate a multi-factor initial intelligent inspection path; the unmanned aerial vehicle flies along the initial intelligent inspection path, collects multi-modal data of the photovoltaic module, and transmits the multi-modal data to a GIS for data fusion and updating to obtain fused data; generating a fault distribution thermodynamic diagram and a maintenance priority map; the inspection path is dynamically adjusted, and the three-dimensional model of the photovoltaic power station is updated. According to the method and the system, comprehensive inspection can be carried out on the photovoltaic power station, the inspection accuracy and efficiency are improved, the inspection path can be dynamically adjusted, unnecessary repeated inspection is reduced, and the inspection cost is effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle inspection, and in particular, relates to a photovoltaic power station inspection method and system based on unmanned aerial vehicle and GIS. Background Art

[0002] With the transformation of the global energy structure and the rapid development of renewable energy, photovoltaic power generation, as an important component of clean energy, is experiencing rapid growth. The scale of large-scale photovoltaic power stations continues to expand, and the installed capacity of a single power station can reach hundreds of megawatts or even gigawatts. However, as the scale of photovoltaic power stations expands, the operation and maintenance management of photovoltaic power stations faces increasing challenges, among which inspection is a key link.

[0003] Using drones for inspections can quickly cover large areas and collect rich data through thermal imagers, high-definition cameras and other equipment. However, in existing technologies, most drone inspection systems operate independently, making it difficult to fully utilize the spatial analysis and visualization capabilities of GIS. Inspection data is difficult to associate with the spatial information of power plants, reducing the value of the data. At the same time, inspection routes often use preset fixed routes, resulting in some areas being repeatedly inspected while other areas being ignored, resulting in low inspection accuracy and fault recognition rates.

[0004] Therefore, finding a method that can effectively combine the data collected by drones with the GIS system for analysis and improve the accuracy of inspections and the fault identification rate is a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a photovoltaic power station inspection method and system based on drones and GIS, which can perform comprehensive, efficient and intelligent inspections of photovoltaic power stations, which not only improves the accuracy and efficiency of the inspections, but also can dynamically adjust the inspection paths, reduce unnecessary repeated inspections, and effectively reduce inspection costs.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a photovoltaic power station inspection method based on drone and GIS, comprising the following steps: S1. Obtaining initial geographic information data of the photovoltaic power station, importing the initial geographic information into GIS, controlling the UAV to perform initial surveying and mapping flight, collecting high-resolution image data, and constructing a three-dimensional model of the photovoltaic power station in GIS based on the high-resolution image data; S2, integrate high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and use GIS spatial analysis to generate a multi-factor initial intelligent inspection path; S3, controlling the UAV to fly along the initial intelligent inspection path generated in step S2, collecting multimodal data of the photovoltaic components, transmitting the multimodal data to the GIS in real time, and fusing and updating the multimodal data with the data in the basic spatial database in real time to obtain fused data; S4. Perform spatiotemporal correlation analysis on the fused data to identify fault characteristics and their spatial distribution patterns, and generate fault distribution heat maps and maintenance priority maps in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; S5. According to the fault distribution heat map and the maintenance priority map, the inspection path is dynamically adjusted, and the equipment status of the undetected area is predicted, and based on the predicted equipment status of the undetected area and the actual inspection data, the three-dimensional model of the photovoltaic power station is updated.

[0007] In a preferred solution, the initial geographic information data in step S1 includes terrain data, equipment distribution data and historical inspection records.

[0008] In a preferred solution, the multiple factors in step S2 include terrain complexity, equipment density, failure history and environmentally sensitive areas.

[0009] In a preferred embodiment, step S2 comprises the following steps: S201, spatially registering the high-resolution image data with the initial geographic information data, extracting the location and orientation information of the photovoltaic modules from the high-resolution image, and comparing and updating the location and orientation information of the photovoltaic modules with the initial equipment distribution data to obtain a basic spatial database of GIS, wherein the basic spatial database includes the registered terrain data, the updated equipment distribution data and the historical inspection records; S202, calculating the terrain complexity of the site according to the registered terrain data, wherein the terrain complexity of the site includes slope, slope aspect and complexity index; S203, analyzing the device distribution density according to the updated device distribution data obtained in step S201, and identifying the device distribution high-density area; S204, performing spatial cluster analysis on the historical inspection records obtained in step S201 to identify high-prone fault areas and environmentally sensitive areas; S205, gridding the photovoltaic power station area according to the terrain complexity of the site, the high-density equipment distribution area, the high-prone fault area, and the environmentally sensitive area, and scoring each grid unit according to each factor to calculate the comprehensive score of each grid unit; S206. Based on the comprehensive scores of the grid units, construct an initial intelligent inspection path with multiple factors.

[0010] In a preferred solution, in step S206, constructing a multi-factor initial intelligent inspection path includes the following steps: 1) Create a multi-level grid structure, where the grid structure includes the top layer, the middle layer, and the bottom layer; 2) Perform rough planning at the top level of the grid structure to obtain a rough path, and determine the initial inspection area and sequence based on the rough path; 3) Connect each segment of the rough path in the middle layer of the grid structure to obtain the middle layer path; 4) Detailed planning of the middle-layer path is carried out at the bottom layer of the grid structure to obtain the initial intelligent inspection path with multiple factors.

[0011] In the preferred solution, an improved ant colony algorithm is used to roughly plan the top layer of the grid structure, specifically including: Initialize the pheromone matrix of the ant colony algorithm ; Select the next node according to the state transition probability. The formula for the transition probability is: ; ; in, Indicates that the kth ant is from node i Move to Node j The probability of Represents a slave node i Move to Node j Heuristic information, Represents a slave node i Move to Node j The distance Represents a slave node i Move to Node j Additional heuristic information for represents the pheromone weight, represents the weight of the heuristic information, represents the weight of the additional heuristic information, represents the kth ant, represents the pheromone matrix, Represents a slave node i Move to Node l The pheromone concentration, Represents a slave node i Move to Node l Heuristic information, Represents a slave node i Move to Node l Additional heuristic information for l Represents all nodes in the top level of the grid structure, i expressl Any node in j express l Remove Node i Any node other than When all ants complete a search, they update the pheromone. The formula is: ; in, represents the pheromone evaporation rate, Indicates pheromone increment.

[0012] In a preferred solution, an optimized RRT* algorithm is used for fine planning, and the fine planning specifically includes: Step 1: Initialize the path tree T and take the starting point as the root node; Step 2: Randomly sample a point X in the grid cell rand , find the distance X in the path tree T rand The nearest node X nearest , using the steering function from X nearest To X rand Direction generates new node X new ; If X new If there is no collision with an obstacle, then: Find the neighboring node set X within the radius r near ; From X near Select the best parent node X min , the selection criteria are: X min = argmin(X∈X near ) {cost(X) + d(X,X new ) - μ * I(X new )}; Among them, cost(X) represents the current minimum cost to reach node X, d(X, X new ) means X to X new The distance, I(X new ) indicates X new The importance score of is, μ represents the importance weight factor; Step 3: X new Add to the path tree T and compare with X min Connect to X near If other nodes in the new The cost of reaching these nodes is lower, so a new path tree T is generated; Step 4: Repeat steps 2 to 3 until the iteration requirement is met. Based on the new path tree T, the optimal path from the starting point to the target area is extracted to obtain the initial intelligent inspection path with multiple factors.

[0013] In a preferred solution, in step S3, the multimodal data is integrated and updated in real time with the data in the basic spatial database, specifically including: S301, normalizing the multimodal data and the data in the basic spatial database to obtain first multimodal data and first GIS data; S302, resampling the processed first multimodal data and the processed first GIS data to the same spatial resolution using an interpolation method to obtain second multimodal data and second GIS data; S303, determining corresponding points between images of different modalities according to the second multimodal data, aligning the corresponding points, and projecting the second GIS data into the same coordinate system; S304, performing multi-scale decomposition on the second multimodal data and the second GIS data on the coordinate system, and performing coefficient fusion, and performing inverse discrete wavelet transform on the coefficient fused data to obtain first fused data; S305, extracting features of the first multimodal data and the first GIS data respectively to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain second fused data; S306, extracting features of the first fused data and constructing a feature vector to obtain a first feature vector, and constructing a feature vector for the second fused data to obtain a second feature vector; S307 , fuse the first eigenvector and the second eigenvector to obtain fused data.

[0014] In a preferred solution, step S4 specifically includes the following steps: S401, performing time series analysis and spatial autocorrelation analysis based on the fused data, and performing time-space interaction analysis to obtain time-space correlation analysis results; S402, extracting the time features, space features and time-space composite features of the fused data, performing fault pattern recognition and fault classification according to the time features, space features and time-space composite features, and obtaining the fault category; S403, performing point mode analysis and hot spot mode analysis on the fault category to obtain fault distribution characteristics and fault concentration, and using the spatial difference method to perform prediction to obtain the fault change trend; S404. Build a scenario in GIS, generate a fault distribution heat map based on the spatiotemporal correlation analysis results, fault categories, fault distribution characteristics, and fault concentration, and create a maintenance priority map based on the fault change trend.

[0015] The present invention also provides a photovoltaic power station inspection system based on a drone and GIS, which uses the above-mentioned photovoltaic power station inspection method to perform inspections, including: A model building module configured to obtain initial geographic information data of the photovoltaic power station, import the initial geographic information into the GIS, control the UAV to perform an initial surveying and mapping flight, collect high-resolution image data, and build a three-dimensional model of the photovoltaic power station in the GIS based on the high-resolution image data; An initial path planning module is configured to fuse high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and to generate a multi-factor initial intelligent inspection path using spatial analysis of GIS; A data fusion module is configured to control the UAV to fly along the initial intelligent inspection path, collect multimodal data of the photovoltaic components, transmit the multimodal data to the GIS in real time, and fuse and update the multimodal data with the data in the basic spatial database in real time to obtain fused data; a data analysis module configured to perform spatiotemporal correlation analysis on the fused data, identify fault characteristics and their spatial distribution patterns, and generate a fault distribution heat map and a maintenance priority map in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; A model update module is configured to dynamically adjust the inspection path according to the fault distribution heat map and the maintenance priority map, and predict the equipment status of the undetected area, and update the three-dimensional model of the photovoltaic power station based on the predicted equipment status of the undetected area and the actual inspection data.

[0016] The present invention provides a photovoltaic power station inspection method and system based on drone and GIS, which has the following beneficial effects: 1. Through the data collected by drones and GIS data, and intelligent inspection route planning, comprehensive, efficient and intelligent inspection of photovoltaic power stations is achieved, which not only improves the accuracy and efficiency of inspections, but also can dynamically adjust the inspection routes, reduce unnecessary repeated inspections, and effectively reduce inspection costs.

[0017] 2. By adopting a multi-level grid structure and an improved ant colony algorithm for inspection route planning, not only can the inspection sequence be optimized, but also refined route planning can be achieved, which greatly improves the inspection coverage and efficiency, while reducing unnecessary repeated inspections and effectively reducing inspection costs.

[0018] 3. By integrating multimodal data and GIS data, not only the integrity and accuracy of the data are improved, but also the basic spatial database can be updated in real time. By performing spatiotemporal correlation analysis on the fused data, combined with point pattern analysis and hotspot analysis, the accuracy of fault diagnosis is improved, the operation and maintenance costs are reduced, and the operation and maintenance management level of the power station is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] Embodiment 1: like Figure 1 As shown, the present invention provides a photovoltaic power station inspection method based on drone and GIS, comprising the following steps: S1. Obtain initial geographic information data of the photovoltaic power station, import the initial geographic information into GIS, control the drone to perform initial surveying and mapping flight, collect high-resolution image data, and build a three-dimensional model of the photovoltaic power station in GIS based on the high-resolution image data; wherein the initial geographic information data includes terrain data, equipment distribution data and historical inspection records.

[0022] Specifically, step S1 is described with a specific embodiment: For a 100MW PV power station, a fixed-wing drone equipped with a 40MP camera was used with a flight altitude of 120 meters, 80% front-to-back overlap, and 60% lateral overlap to complete aerial photography of the entire PV power station.

[0023] Use PiX4Dmapper software to process images and generate orthophotos and point cloud data.

[0024] Use ConteXtCapture software to build a 3D model based on point cloud data and import it into ArcGIS Pro for integration with existing equipment distribution data.

[0025] S2. Fuse high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and use GIS spatial analysis to generate an initial intelligent inspection path based on multiple factors, including terrain complexity, equipment density, fault history, and environmentally sensitive areas.

[0026] Specifically, step S2 specifically includes: S201. Spatial registration of the high-resolution image data with the initial geographic information data, extraction of the location and orientation information of the photovoltaic modules from the high-resolution image, and comparison and update of the location and orientation information of the photovoltaic modules with the equipment distribution data to obtain a basic spatial database of GIS; wherein the basic spatial database includes the registered terrain data, the updated equipment distribution data and the historical inspection records.

[0027] S202: Calculate the terrain complexity of the site according to the registered terrain data, wherein the terrain complexity includes slope, slope aspect and complexity index.

[0028] S203: Analyze device distribution density according to the updated device distribution data, and identify areas with high device distribution density.

[0029] S204. Perform spatial cluster analysis on historical inspection records to identify high-failure areas and environmentally sensitive areas.

[0030] S205. Grid-divide the photovoltaic power station area according to terrain complexity, equipment high-density distribution areas, high-failure areas, and environmentally sensitive areas, and score each grid unit according to each factor to calculate the comprehensive score of each grid unit.

[0031] S206. Based on the comprehensive scores of the grid units, construct an initial intelligent inspection path with multiple factors.

[0032] As can be understood by those skilled in the art, the entire photovoltaic power station area is divided into uniform grids, and each grid represents a path point.

[0033] In one embodiment of the present invention, the grid is divided into 10m*10m. The grid size can be set according to actual conditions, and this embodiment does not make any specific restrictions on this.

[0034] The slope is calculated as follows: The photovoltaic power station area is divided into multiple grid cells, and the registered terrain data is preprocessed, including removing outliers and filling holes; A 3×3 moving window is used to calculate the slope of the central pixel e0 in each grid cell, where the gradient calculation formula in the east-west direction is: G x =((e3+2e6+e9)-(e1+2e4+e7)) / (8×d x ) Among them, G xrepresents the elevation gradient of the grid unit in the east-west direction, e1 represents the elevation value of the upper left pixel of the grid unit, e3 represents the elevation value of the upper right pixel of the grid unit, e4 represents the elevation value of the left pixel of the grid unit, e6 represents the elevation value of the right pixel of the grid unit, e7 represents the elevation value of the lower left pixel of the grid unit, e9 represents the elevation value of the lower right pixel of the grid unit, and d x Represents the horizontal distance between adjacent pixels in the east-west direction of the grid cell; The gradient calculation formula in the north-south direction is: G y =((e7+2e8+e9)-(e1+2e2+e3)) / (8×d y ) Among them, G y represents the elevation gradient of the grid cell in the north-south direction, e2 represents the elevation value of the pixel above the grid cell, e8 represents the elevation value of the pixel below the grid cell, and d y Represents the vertical distance between adjacent pixels in the north-south direction of the grid cell; The gradient is calculated as:

[0035] in, represents the gradient; The aspect is calculated as follows: The slope angle is calculated using a 3×3 moving window. The calculation formula is: Aspect = 180 / π×arctan2(G y , -G x ) Among them, Aspect represents the slope angle; Convert the calculated result to 0-360 degrees. If Aspect<0, then Aspect = 360 + Aspect. The slope aspect of flat land with a slope of 0 is assigned a value of 1.

[0036] The calculation method of complexity index is: Calculate the terrain relief, that is, calculate the square root of the sum of the squares of the elevation differences between the central pixel and the surrounding eight pixels; Calculate the terrain position index, that is, calculate the difference between the average elevation of the central pixel and the surrounding pixels; Terrain roughness was calculated using a 3×3 moving window.

[0037] Furthermore, the spatial cluster analysis in step S204 specifically includes: Standardize historical inspection records and convert fault point data into spatial coordinates; Use DBSCAN (density-based spatial clustering algorithm) to determine the spatial distribution characteristics of fault-prone areas; Combined with geographical environmental factors, identify high-failure areas and environmentally sensitive areas.

[0038] Furthermore, a multi-factor initial intelligent inspection path is constructed, including: 1) Create a multi-level grid structure, where the grid structure includes the top layer, the middle layer, and the bottom layer; 2) Perform rough planning at the top level of the grid structure to obtain a rough path, and determine the initial inspection area and sequence based on the rough path; 3) Connect each segment of the rough path in the middle layer of the grid structure to obtain the middle layer path; 4) Detailed planning of the middle-layer path is carried out at the bottom layer of the grid structure to obtain the initial intelligent inspection path with multiple factors.

[0039] In one embodiment of the present invention, the top layer is the coarsest-grained grid division, the middle layer is the medium-grained grid division, and the bottom layer is the finest-grained grid division.

[0040] In one embodiment of the present invention, the top layer is a 100m*100m grid, the middle layer is a 20m*20m grid, and the bottom layer is a 5m*5m grid.

[0041] Furthermore, the improved ant colony algorithm is used to perform rough planning on the top layer of the grid structure, including: Initialize the pheromone matrix of the ant colony algorithm ; Select the next node according to the state transition probability. The formula for the transition probability is: ; ; in, Indicates that the kth ant is from node i Move to Node j The probability of Represents a slave node i Move to Node j Heuristic information, Represents a slave node i Move to Node j The distance Represents a slave node i Move to Node j Additional heuristic information for represents the pheromone weight, represents the weight of the heuristic information, represents the weight of the additional heuristic information, represents the kth ant, represents the pheromone matrix, Represents a slave nodei Move to Node l The pheromone concentration, Represents a slave node i Move to Node l Heuristic information, Represents a slave node i Move to Node l Additional heuristic information for l Represents all nodes in the top level of the grid structure, i express l Any node in j express l Remove Node i Any node other than When all ants complete a search, they update the pheromone. The formula is: ; in, represents the pheromone evaporation rate, Indicates pheromone increment.

[0042] The present invention adopts a multi-level grid structure and an improved ant colony algorithm for inspection path planning, which can not only optimize the inspection sequence, but also realize refined path planning, greatly improving the inspection coverage and efficiency, while reducing unnecessary repeated inspections and effectively reducing inspection costs.

[0043] Furthermore, each segment of the rough path is connected in the middle layer of the grid structure, including: Find the shortest path from the starting point of the grid unit to each node in the grid unit, update the rough path according to the shortest path, and get the middle-level path.

[0044] Specifically, a specific embodiment is used to illustrate: For each pair of adjacent nodes (s, t) in the coarse path, perform the following steps: a. Initialize the distance array d[v]=∞, where d[v] represents the current shortest path length from the starting point s to the node v; b. Judge each edge (u, v) of the grid unit: u represents the node of the edge in the grid unit, If d[v]>d[u]+w(u,v)+λ*(1 / I(v)), then d[v]=d[u]+w(u,v)+λ*(1 / I(v)), record the predecessor node of v: predecessor[v]=u; where d[u] represents the current known shortest path length from the starting point s to the node u, w(u,v) represents the distance from the node u to the node v, λ represents the trade-off parameter, and I(v) represents the importance index of the node v; c. Repeat step b |V-1| times, using the predecessor array to trace back from the end point t to the starting point s to obtain the middle-level path; where V represents the number of nodes.

[0045] Furthermore, the optimized RRT* algorithm is used for detailed planning, which includes: Step 1: Initialize the path tree T and take the starting point as the root node; Step 2: Randomly sample a point X in the grid cell rand , find the distance X in the path tree T rand The nearest node X nearest , using the steering function from X nearest To X rand Direction generates new node X new ; If X new If there is no collision with an obstacle, then: Find the neighboring node set X within the radius r near ; From X near Select the best parent node X min , the selection criteria are: X min = argmin(X∈X near ) {cost(X) + d(X,X new ) - μ * I(X new )}; Among them, cost(X) represents the current minimum cost to reach node X, d(X, X new ) means X to X new The distance, I(X new ) indicates X new The importance score of is, μ represents the importance weight factor; Step 3: X new Add to the path tree T and compare with X min Connect to X near If other nodes in the new The cost of reaching these nodes is lower, so a new path tree T is generated; Step 4: Repeat steps 2 to 3 until the iteration requirement is met. Based on the new path tree T, the optimal path from the starting point to the target area is extracted to obtain the initial intelligent inspection path with multiple factors.

[0046] Specifically, a 100MW photovoltaic power station covering an area of ​​about 2 square kilometers is used as an example for description: The photovoltaic power station area is divided into a top grid of 20x20 (100m*100m per grid), a middle grid of 100x100 (20m*20m per grid), and a bottom grid of 400x400 (5m*5m per grid).

[0047] The average slope and aspect diversity of each grid were calculated based on the DEM data.

[0048] Calculate the number of PV panels in each grid; Analyze the failure records of the past 6 months and calculate the failure frequency of each grid.

[0049] Mark grids near water sources or wildlife habitats.

[0050] In the top grid, initialize the parameters: α=1, β=2, γ=1.5, ρ=0.1, set the number of ants to 20, the maximum number of iterations to 100, and set additional heuristic information based on the comprehensive score of the grid. , perform rough planning in the top-level grid and obtain a rough path.

[0051] In the middle-level grid, each pair of adjacent top-level grids in the coarse path is connected to obtain the middle-level path.

[0052] In the bottom grid, the number of sampling points is set to 1000, the connection radius r = 10m, and μ = 0.3. The path length and importance are balanced, and the middle-level path is carefully planned to obtain the initial intelligent inspection path with multiple factors.

[0053] S3. Control the UAV to fly along the initial intelligent inspection path, collect multimodal data of the photovoltaic components, transmit the multimodal data to GIS in real time, and fuse and update the multimodal data with the data in the basic spatial database in real time to obtain fused data; wherein the multimodal data includes visible light images, thermal imaging data and spectral data.

[0054] Specifically, the multimodal data is integrated and updated in real time with the data in the basic spatial database, including: S301, normalizing the multimodal data and the data in the basic spatial database to obtain first multimodal data and first GIS data; S302, resampling the processed multimodal data and the processed GIS data to the same spatial resolution using an interpolation method to obtain second multimodal data and second GIS data; S303, determining corresponding points between images of different modalities according to the second multimodal data, aligning the corresponding points, and projecting the second GIS data into the same coordinate system; S304, performing multi-scale decomposition on the second multimodal data and the second GIS data on the coordinate system, and performing coefficient fusion, and performing inverse discrete wavelet transform on the coefficient fused data to obtain first fused data; S305, extracting features of the first multimodal data and the first GIS data respectively to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain second fused data; S306, extracting features of the first fused data and constructing a feature vector to obtain a first feature vector, and constructing a feature vector for the second fused data to obtain a second feature vector; S307 , fuse the first eigenvector and the second eigenvector to obtain fused data.

[0055] By fusing multimodal data with GIS data, the present invention not only improves the integrity and accuracy of the data, but also can update the basic spatial database in real time. By performing spatiotemporal correlation analysis on the fused data, combined with point pattern analysis and hotspot analysis, the accuracy of fault diagnosis is improved, the operation and maintenance costs are reduced, and the operation and maintenance management level of the power station is enhanced.

[0056] In one embodiment of the present invention, the first eigenvector and the second eigenvector may be fused by weighted fusion, by random forest algorithm, or by deep learning neural network. The present invention does not impose any specific limitation on the fusion method.

[0057] S4. Perform spatiotemporal correlation analysis on the fused data to identify fault characteristics and their spatial distribution patterns, and generate fault distribution heat maps and maintenance priority maps in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; Specifically, step S4 specifically includes: S401, performing time series analysis and spatial autocorrelation analysis based on the fused data, and performing time-space interaction analysis to obtain time-space correlation analysis results; S402, extracting the time features, space features and time-space composite features of the fused data, performing fault pattern recognition and anomaly detection according to the time features, space features and time-space composite features, and obtaining the fault category; S403, performing point mode analysis and hot spot mode analysis on the fault category to obtain fault distribution characteristics and fault concentration, and using the spatial difference method to perform prediction to obtain the fault change trend; S404. Build a scenario in GIS, generate a fault distribution heat map based on the spatiotemporal correlation analysis results, fault categories, fault distribution characteristics, and fault concentration, and create a maintenance priority map based on the fault change trend.

[0058] Among them, in step S402, fault mode recognition and anomaly detection are performed according to time characteristics, space characteristics and time-space composite characteristics, specifically including: Determine the fault frequency and duration based on time characteristics, determine the fault location based on spatial characteristics, and determine the fault diffusion speed based on time-space composite characteristics; Faults are classified according to the frequency of occurrence, duration of the fault, location of the fault, and speed of fault spread. For example, Class A faults are serious faults that require immediate processing, Class B faults are potential faults that require intensive monitoring, and Class C faults are minor faults that require regular inspection.

[0059] It can be understood that the method of determining the fault frequency, fault duration, fault location and fault diffusion speed in the present application can select different methods according to the actual use environment, and the present application does not make specific limitations on the comparison.

[0060] Furthermore, the point mode analysis and hotspot mode analysis in step S403 specifically include: Calculate the distance matrix between different faults, count the nearest neighbor distances, and evaluate the spatial distribution characteristics of faults; Set the bandwidth and search radius, calculate the fault density surface based on the bandwidth and search radius, and identify the high-value fault density area; The global Moran's I index is calculated by constructing a spatial weight matrix based on the inverse distance weight, and the correlation between faults is evaluated according to the Moran's I index. Calculate LISA statistics and identify local hot spots.

[0061] In one embodiment of the present invention, an autoregressive integrated moving average model and wavelet transform are used to perform time series analysis.

[0062] In one embodiment of the present invention, Moran's I index is calculated and spatial autocorrelation analysis is performed using LISA statistics.

[0063] In one embodiment of the present invention, a spatiotemporal interaction analysis is performed using a spatiotemporal K function and spatiotemporal scanning statistics to obtain a spatiotemporal correlation analysis result.

[0064] S5. According to the fault distribution heat map and the maintenance priority map, the inspection path is dynamically adjusted, and the equipment status of the undetected area is predicted, and based on the predicted equipment status of the undetected area and the actual inspection data, the three-dimensional model of the photovoltaic power station is updated.

[0065] The present invention realizes comprehensive, efficient and intelligent inspection of photovoltaic power stations through data collected by drones and GIS data, and performs intelligent inspection route planning. It not only improves the accuracy and efficiency of inspections, but also can dynamically adjust inspection routes, reduce unnecessary repeated inspections, and effectively reduce inspection costs.

[0066] Embodiment 2: The present invention also provides a photovoltaic power station inspection system based on drones and GIS, which adopts the photovoltaic power station inspection method as described above, including: A model building module configured to obtain initial geographic information data of the photovoltaic power station, import the initial geographic information into the GIS, control the UAV to perform an initial surveying and mapping flight, collect high-resolution image data, and build a three-dimensional model of the photovoltaic power station in the GIS based on the high-resolution image data; An initial path planning module is configured to fuse high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and to generate a multi-factor initial intelligent inspection path using spatial analysis of GIS; A data fusion module is configured to control the UAV to fly along the initial intelligent inspection path, collect multimodal data of the photovoltaic components, transmit the multimodal data to the GIS in real time, and fuse and update the multimodal data with the data in the basic spatial database in real time to obtain fused data; a data analysis module configured to perform spatiotemporal correlation analysis on the fused data, identify fault characteristics and their spatial distribution patterns, and generate a fault distribution heat map and a maintenance priority map in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; A model update module is configured to dynamically adjust the inspection path according to the fault distribution heat map and the maintenance priority map, and predict the equipment status of the undetected area, and update the three-dimensional model of the photovoltaic power station based on the predicted equipment status of the undetected area and the actual inspection data.

[0067] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic power station inspection method based on drones and GIS, characterized in that: The following steps are involved: S1. Obtaining initial geographic information data of the photovoltaic power station, importing the initial geographic information into GIS, controlling the UAV to perform initial surveying and mapping flight, collecting high-resolution image data, and constructing a three-dimensional model of the photovoltaic power station in GIS based on the high-resolution image data; S2, integrate high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and use GIS spatial analysis to generate a multi-factor initial intelligent inspection path; S3, controlling the UAV to fly along the initial intelligent inspection path generated in step S2, collecting multimodal data of the photovoltaic components, transmitting the multimodal data to the GIS in real time, and fusing and updating the multimodal data with the data in the basic spatial database in real time to obtain fused data; S4. Perform spatiotemporal correlation analysis on the fused data to identify fault characteristics and their spatial distribution patterns, and generate fault distribution heat maps and maintenance priority maps in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; S5. According to the fault distribution heat map and the maintenance priority map, the inspection path is dynamically adjusted, and the equipment status of the undetected area is predicted, and based on the predicted equipment status of the undetected area and the actual inspection data, the three-dimensional model of the photovoltaic power station is updated.

2. A photovoltaic power station inspection method based on drone and GIS according to claim 1, characterized in that: The initial geographic information data in step S1 includes terrain data, equipment distribution data and historical inspection records.

3. The photovoltaic power station inspection method based on drone and GIS according to claim 1 is characterized in that: The multiple factors in step S2 include terrain complexity, equipment density, failure history, and environmentally sensitive areas.

4. The photovoltaic power station inspection method based on drone and GIS according to claim 1 is characterized in that: The step S2 comprises the following steps: S201, spatially registering the high-resolution image data with the initial geographic information data, extracting the location and orientation information of the photovoltaic modules from the high-resolution image, and comparing and updating the location and orientation information of the photovoltaic modules with the initial equipment distribution data to obtain a basic spatial database of GIS, wherein the basic spatial database includes the registered terrain data, the updated equipment distribution data and the historical inspection records; S202, calculating the terrain complexity of the site according to the registered terrain data, wherein the terrain complexity of the site includes slope, slope aspect and complexity index; S203, analyzing the device distribution density according to the updated device distribution data obtained in step S201, and identifying the device distribution high-density area; S204, performing spatial cluster analysis on the historical inspection records obtained in step S201 to identify high-prone fault areas and environmentally sensitive areas; S205, gridding the photovoltaic power station area according to the terrain complexity of the site, the high-density equipment distribution area, the high-prone fault area, and the environmentally sensitive area, and scoring each grid unit according to each factor to calculate the comprehensive score of each grid unit; S206. Based on the comprehensive scores of the grid units, construct an initial intelligent inspection path with multiple factors.

5. A photovoltaic power station inspection method based on drone and GIS according to claim 4, characterized in that: In step S206, constructing a multi-factor initial intelligent inspection path includes the following steps: 1) Create a multi-level grid structure, where the grid structure includes the top layer, the middle layer, and the bottom layer; 2) Perform rough planning at the top level of the grid structure to obtain a rough path, and determine the initial inspection area and sequence based on the rough path; 3) Connect each segment of the rough path in the middle layer of the grid structure to obtain the middle layer path; 4) Detailed planning of the middle-layer path is carried out at the bottom layer of the grid structure to obtain the initial intelligent inspection path with multiple factors.

6. A photovoltaic power station inspection method based on drone and GIS according to claim 5, characterized in that: The improved ant colony algorithm is used to roughly plan the top layer of the grid structure, including: Initialize the pheromone matrix of the ant colony algorithm ; Select the next node according to the state transition probability. The formula for the transition probability is: ； ； in, Indicates that the kth ant is from node i Move to Node j The probability of Represents a slave node i Move to Node j Heuristic information, Represents a slave node i Move to Node j The distance Represents a slave node i Move to Node j Additional heuristic information for represents the pheromone weight, represents the weight of the heuristic information, represents the weight of the additional heuristic information, represents the kth ant, represents the pheromone matrix, Represents a slave node i Move to Node l The pheromone concentration, Represents a slave node i Move to Node l Heuristic information, Represents a slave node i Move to Node l Additional heuristic information for l Represents all nodes in the top level of the grid structure, i express l Any node in j express l Remove Node i Any node other than When all ants complete a search, they update the pheromone. The formula is: ； in, represents the pheromone evaporation rate, Indicates pheromone increment.

7. A photovoltaic power station inspection method based on drone and GIS according to claim 5, characterized in that: Use the optimized RRT* algorithm for detailed planning, which specifically includes: Step 1: Initialize the path tree T and take the starting point as the root node; Step 2: Randomly sample a point X in the grid cell rand , find the distance X in the path tree T rand The nearest node X nearest , using the steering function from X nearest To X rand Direction generates new node X new ; If X new If there is no collision with an obstacle, then: Find the neighboring node set X within the radius r near ; From X near Select the best parent node X min , the selection criteria are: X min = argmin(X∈X near ) {cost(X) + d(X,X new ) - μ * I(X new )}; Among them, cost(X) represents the current minimum cost to reach node X, d(X, X new ) means X to X new The distance, I(X new ) indicates X new The importance score of is, μ represents the importance weight factor; Step 3: X new Add to the path tree T and compare with X min Connect to X near If other nodes in the new The cost of reaching these nodes is lower, so a new path tree T is generated; Step 4: Repeat steps 2 to 3 until the iteration requirement is met. Based on the new path tree T, the optimal path from the starting point to the target area is extracted to obtain the initial intelligent inspection path with multiple factors.

8. The photovoltaic power station inspection method based on drone and GIS according to claim 1 is characterized in that: In step S3, the multimodal data is integrated and updated in real time with the data in the basic spatial database, which specifically includes: S301, normalizing the multimodal data and the data in the basic spatial database to obtain first multimodal data and first GIS data; S302, resampling the processed first multimodal data and the processed first GIS data to the same spatial resolution using an interpolation method to obtain second multimodal data and second GIS data; S303, determining corresponding points between images of different modalities according to the second multimodal data, aligning the corresponding points, and projecting the second GIS data into the same coordinate system; S304, performing multi-scale decomposition on the second multimodal data and the second GIS data on the coordinate system, and performing coefficient fusion, and performing inverse discrete wavelet transform on the coefficient fused data to obtain first fused data; S305, extracting features of the first multimodal data and the first GIS data respectively to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain second fused data; S306, extracting features of the first fused data and constructing a feature vector to obtain a first feature vector, and constructing a feature vector for the second fused data to obtain a second feature vector; S307 , fuse the first eigenvector and the second eigenvector to obtain fused data.

9. The photovoltaic power station inspection method based on drone and GIS according to claim 1 is characterized in that: The step S4 specifically comprises the following steps: S401, performing time series analysis and spatial autocorrelation analysis based on the fused data, and performing time-space interaction analysis to obtain time-space correlation analysis results; S402, extracting the time features, space features and time-space composite features of the fused data, performing fault mode recognition and fault classification according to the time features, space features and time-space composite features, and obtaining the fault category; S403, performing point mode analysis and hot spot mode analysis on the fault category to obtain fault distribution characteristics and fault concentration, and using the spatial difference method to perform prediction to obtain the fault change trend; S404. Build a scenario in GIS, generate a fault distribution heat map based on the spatiotemporal correlation analysis results, fault categories, fault distribution characteristics, and fault concentration, and create a maintenance priority map based on the fault change trend.

10. A photovoltaic power station inspection system based on drones and GIS, characterized in that: The photovoltaic power station inspection method according to any one of claims 1 to 9 is used for inspection, comprising: A model building module configured to obtain initial geographic information data of the photovoltaic power station, import the initial geographic information into the GIS, control the UAV to perform an initial surveying and mapping flight, collect high-resolution image data, and build a three-dimensional model of the photovoltaic power station in the GIS based on the high-resolution image data; An initial path planning module is configured to fuse high-resolution image data with initial geographic information data to form a basic spatial database of GIS, and to generate a multi-factor initial intelligent inspection path using spatial analysis of GIS; A data fusion module is configured to control the UAV to fly along the initial intelligent inspection path, collect multimodal data of the photovoltaic components, transmit the multimodal data to the GIS in real time, and fuse and update the multimodal data with the data in the basic spatial database in real time to obtain fused data; a data analysis module configured to perform spatiotemporal correlation analysis on the fused data, identify fault characteristics and their spatial distribution patterns, and generate a fault distribution heat map and a maintenance priority map in a three-dimensional geographic information system environment based on the identified fault characteristics and spatial distribution patterns; A model update module is configured to dynamically adjust the inspection path according to the fault distribution heat map and the maintenance priority map, and predict the equipment status of the undetected area, and update the three-dimensional model of the photovoltaic power station based on the predicted equipment status of the undetected area and the actual inspection data.

Images