Method for identifying installation quality of photovoltaic array based on remote sensing data

Abstract

The invention mainly relates to the technical field of remote sensing detection. In order to solve the problems that the installation efficiency of a photovoltaic array is low, the installation quality detection cannot cover the whole area and the random error is large, the invention provides a method for identifying the installation quality of a photovoltaic array based on remote sensing data, and the core idea of the method comprises the steps: collecting first point cloud data before the installation of the photovoltaic array to generate a DEM image; during the construction period of the photovoltaic array, an orthoimage and second point cloud data are collected, a DOM image is generated based on the orthoimage, and a DSM digital earth surface model is generated based on the DOM image and the spatial position of the second point cloud data; establishing a photovoltaic array installation quality identification model to judge the construction stage of the photovoltaic array, performing fitting comparison on the DEM image and the second point cloud data, identifying strings and installation parameters in the photovoltaic array, and comparing the strings and the installation parameters with the installation parameters in the established photovoltaic array installation quality standard library; and marking an unqualified photovoltaic array string in the DSM digital earth surface model.

Description

Technical Field

[0001] The present invention mainly relates to the field of remote sensing detection technology, and in particular to a method for identifying the installation quality of a photovoltaic array based on remote sensing data. Background Art

[0002] Most of the centralized photovoltaic power station sites in the southwest region are located in remote mountainous areas with complex terrain and large altitude differences. Therefore, the construction and installation of the array area are difficult, which indirectly affects the construction quality. In addition, the photovoltaic site occupies a large area, and the working interfaces are numerous and scattered during the peak construction period. For project managers, it is impossible to control the construction quality at all times. The traditional method generally adopts manual spot checks, and uses measuring tools to go to the site to detect the pile height, the height of the bracket from the ground, the installation position and angle of the inclined beam, the installation position of the purlin, and the complete assembly position. The above methods are extremely inefficient, have limited coverage, and have large random errors. They are not suitable for large-scale centralized photovoltaic power stations with huge installed capacity.

[0003] In order to improve the quality control capabilities of the construction process and help project managers improve work efficiency, the demand for automation and intelligence to replace manual quality inspection is becoming increasingly obvious, and it is also an inevitable trend. Summary of the invention

[0004] Technical Problems to be Solved by the Invention

[0005] The present invention provides a photovoltaic array installation method based on remote sensing data, aiming to solve the problems of low installation efficiency of the current photovoltaic array and the inability of installation quality detection to cover the entire area and large random errors.

[0006] The technical solution adopted by the present invention to solve the above technical problems

[0007] The method for identifying the installation quality of photovoltaic arrays based on remote sensing data includes:

[0008] Step S1: Establishing a photovoltaic array installation quality standard library, wherein the photovoltaic array installation quality standard library stores installation parameters of various components of the photovoltaic array under standard installation conditions at different construction stages;

[0009] Step S2: collecting point cloud data of the front area of ​​the photovoltaic array installation as first point cloud data, performing three-dimensional reconstruction on the first point cloud data to generate a DEM image;

[0010] Step S3: during the construction of the photovoltaic array, the orthophoto image and the second point cloud data of the field area are collected according to the set collection frequency, a DOM image is generated based on the orthophoto image collected once, and a DSM digital surface model is generated based on the spatial positions of the generated multiple DOM images and the second point cloud data;

[0011] Step S4: Establish a photovoltaic array installation quality recognition model, the photovoltaic array installation quality recognition model recognizes the DOM image, determines the construction stage of a single photovoltaic array string, and verifies the construction stage judgment result with the second point cloud data of the corresponding acquisition time; fits and compares the DEM image and the second point cloud data of the corresponding acquisition time, screens and segments the point cloud data above the surface in the second point cloud data, identifies the photovoltaic array strings and their installation parameters, and compares the identified installation parameters with the installation parameters under the standard installation state of the corresponding construction stage in the standard installation library, and marks unqualified photovoltaic array strings in the DSM digital surface model based on the comparison results.

[0012] Furthermore, after collecting the orthophoto images and point cloud data, the orthophoto images and point cloud data are preprocessed. The preprocessing of the orthophoto images includes: denoising, geometric correction, contrast enhancement and distortion correction; the preprocessing of the point cloud data includes: filtering redundant point clouds, removing local areas, checking terrain and optimizing.

[0013] Furthermore, in step S3, the collected point cloud data are divided according to the sub-array boundary of the photovoltaic array general plan layout diagram with the sub-array of the photovoltaic array as the basic unit, and the second point cloud data are generated based on the set division result.

[0014] Furthermore, when generating the DOM image in step S3, the collected orthophoto images are grouped according to the sub-array boundaries of the photovoltaic array general plan layout diagram, taking the sub-array as the basic unit, and generating the sub-array DOM image containing geographic coordinates based on the group division result.

[0015] Furthermore, the construction stages specifically include a pile foundation construction stage, a column and inclined beam installation stage, a purlin installation stage, and a component installation stage.

[0016] Furthermore, a photovoltaic array installation quality recognition model is established based on the target detection algorithm.

[0017] Beneficial Effects of the Invention

[0018] The photovoltaic array installation method based on remote sensing data described in the present invention solves the low efficiency problem of traditional manual inspection of photovoltaic power station construction and installation quality. The photovoltaic array installation recognition model performs fast and full-coverage quality inspection of photovoltaic array installation quality based on orthophoto images and point cloud data collected by remote sensing, which greatly improves work efficiency. In addition, a photovoltaic array installation recognition model is established to automatically analyze orthophoto images and point cloud data collected by remote sensing, which reduces the influence of human factors and improves the accuracy and reliability of the inspection results. It is particularly suitable for large-scale centralized photovoltaic power stations with huge installed capacity, improves the comprehensiveness and accuracy of photovoltaic array quality inspection, and helps to improve the construction quality of photovoltaic power stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of the method for identifying the installation quality of a photovoltaic array based on remote sensing data according to the present invention;

[0020] Figure 2 This is an image of the photovoltaic array foundation pile construction stage;

[0021] Figure 3 This is an image of the photovoltaic array column diagonal beam installation stage;

[0022] Figure 4 Images of the photovoltaic array purlin installation phase;

[0023] Figure 5 Images of the photovoltaic array component installation stages. DETAILED DESCRIPTION

[0024] like Figure 1 As shown, the method for identifying the installation quality of a photovoltaic array based on remote sensing data described in the present invention is as follows.

[0025] Step S1: Establish a photovoltaic array installation quality standard library, which stores the installation parameters of each component of the photovoltaic array under standard installation conditions at different construction stages. Before the photovoltaic power station enters civil construction, the height and verticality of the piles above the ground, the height and verticality of the columns, the installation angle and flatness of the diagonal beams and braces, the bracket spacing, the height difference of the component edges and the relative height to the ground are standardized according to the photovoltaic power station acceptance specifications and design drawings, as a data comparison library for the identification results.

[0026] Step S2: collecting point cloud data of the photovoltaic array installation front area as first point cloud data, performing three-dimensional reconstruction on the first point cloud data, and generating a DEM image.

[0027] Since fixed-wing drones are expensive to purchase and use, and are not suitable for the complex terrain and changeable climate in the high-altitude areas of the southwest, they are less flexible than multi-rotors and cannot collect high-precision images. Therefore, a multi-rotor drone with a ground control station is used to collect orthophotos and laser point cloud data of the site. The drone is equipped with real-time dynamic differential positioning (RTK), laser radar (LiDAR) and high-resolution image sensors to collect data for the photovoltaic power station under construction. Since the communication between the drone control station and the drone is linear, mountain obstruction will affect signal transmission. The reasonable layout of the control station should be based on field survey and site selection, considering factors such as mountain obstruction, altitude difference, and cost, and multiple control stations should be dispersed to ensure that the communication signal can fully cover the site. The flight path planning should be based on the latest version of the photovoltaic site general plan provided by the design unit, and the sub-arrays should be individually planned according to the sub-array range. In order to ensure flight safety and avoid path obstacles, the route planning should be carried out and verified on site.

[0028] During data collection, the clarity of the orthophoto image and the integrity of the point cloud data need to be ensured. Therefore, the lateral overlap of the UAV flight path is set to be no less than 70%, the heading overlap is no less than 80%, the shooting frequency is no more than 1s / photo, the laser radar frequency is no less than 240KHz, and the ground sampling distance (GSD) is no more than 2cm. In addition, the formal images and point cloud data collected need to cover the entire range of the area within the red line of the land before the construction of the photovoltaic project, and the entire range of the area within the boundary of all sub-arrays during the array construction.

[0029] According to the above-mentioned UAV flight route and the planned collection area and collection parameters, the orthophoto images and point cloud data are collected before the construction of the photovoltaic array and during the construction of the photovoltaic array, and the collected orthophoto images and point cloud data are preprocessed respectively; wherein, the preprocessing of the orthophoto images includes but is not limited to image denoising, geometric correction, contrast enhancement and distortion correction, and the preprocessing of the point cloud data includes but is not limited to filtering redundant point clouds, removing local areas, checking terrain and optimization.

[0030] The point cloud data collected before the construction of the photovoltaic array in the site is used as the first point cloud data, and three-dimensional reconstruction is performed based on the first point cloud data to generate a DEM image containing topographic information such as elevation, slope and slope direction.

[0031] Step S3: during the construction of the photovoltaic array, according to the same remote sensing data collection route and other remote sensing data collection requirements, the orthophoto image and the second point cloud data of the field area are collected at a set collection frequency, a DOM image is generated based on the orthophoto image collected once, and a DSM digital surface model is generated based on the spatial positions of the generated multiple DOM images and the second point cloud data;

[0032] After preprocessing the orthophoto images and the second point cloud data, based on the orthophoto images of different construction stages, the collected orthophoto images are divided into sets with a single sub-array as the basic unit according to the sub-array boundaries of the general plan layout. DOM images with geographic coordinates of the sub-arrays are synthesized based on each set, and the DOM images of different sub-arrays are named according to the sub-array numbers.

[0033] The spatial XY positions of DOM and the second point cloud data are superimposed to further synthesize the DSM digital surface model, which contains 1) landform information: elevation, slope, aspect, slope morphology and more complex landform factors describing the surface undulations; 2) columns, inclined beams, complete brackets, components, etc. containing position, height, angle and color information. The DSM digital surface model will serve as the visualization data base. The data in the construction phase will be updated in real time according to the construction progress, and the array construction and installation quality will be monitored in real time through AI algorithms, and the results will be visualized on the DSM digital surface model.

[0034] Step S4: Establish a photovoltaic array installation quality recognition model, the photovoltaic array installation quality recognition model recognizes the DOM image, determines the construction stage of a single photovoltaic array string, and verifies the construction stage judgment result with the second point cloud data of the corresponding acquisition time; fits and compares the DEM image and the second point cloud data of the corresponding acquisition time, screens and segments the point cloud data above the surface in the second point cloud data, identifies the installation parameters of a single photovoltaic array string, and compares the identified installation parameters with the installation parameters under the standard installation state of the corresponding construction stage in the standard installation library, and marks unqualified photovoltaic array strings in the DSM digital surface model based on the comparison results.

[0035] The process of establishing the photovoltaic array installation quality identification model includes:

[0036] 1. Data annotation: The annotation method uses 2D / 3D fusion annotation. The synthesized DOM image contains various stages of the construction period, among which the stages that need to be detected are divided into the following Figure 2 The pile foundation construction stages shown in Figure 3 The installation stage of the column and inclined beam shown in the figure is as follows: Figure 4 The purlin installation stage shown in Figure 5 The component installation stages shown are marked according to the above four stages.

[0037] (1) 2D annotation: In the DOM image, the pile foundation, column beam, purlin, complete bracket, PV component, and complete string are annotated separately, including the type, ID attribute and other features of the above six types of objects. The annotated orthophoto images are divided into a test data set and a validation data set.

[0038] (2) 3D annotation: Annotate the second point cloud data, including pile foundations, column diagonal beams, purlins, complete brackets, PV components, and complete strings, and divide the annotated second point cloud data into a test data set and a verification data set.

[0039] 2. Model training: Train the target detection algorithm based on the test data set, verify the accuracy through the verification data set, and repeatedly optimize the parameters.

[0040] 3. Identify and return results

[0041] 2D / 3D data are mapped to each other through spatial geographic coordinates. First, the photovoltaic array installation quality recognition model identifies the sub-array DOM image to determine the construction stage to which it belongs, clarifies the bracket structure form of this stage and the spatial geographic location corresponding to the sub-array in the point cloud data, clarifies the sub-array range and the corresponding construction stage identified by the point cloud data, and calls the pre-construction DEM site digital elevation model to compare with the sub-array point cloud data to verify the installation stage. Then, based on the target detection algorithm, the photovoltaic array installation parameters are analyzed, and the installation parameters such as pile foundation positioning, surface verticality of the column and inclined beam installation stage, flatness of the purlin installation stage, inclination angle and mutual spacing of the complete bracket, and height difference of the component edge are calculated. The calculated installation parameters are compared with the installation parameters under the standard installation state of the corresponding construction stage in the standard installation library. The photovoltaic array installation parameters that do not meet the standard installation state are returned and mapped to the sub-array DSM digital surface model, and the strings that do not meet the standards are marked.

Claims

1. A method for identifying the installation quality of a photovoltaic array based on remote sensing data, characterized in that: include: Step S1: Establishing a photovoltaic array installation quality standard library, wherein the photovoltaic array installation quality standard library stores installation parameters of various components of the photovoltaic array under standard installation conditions at different construction stages; Step S2: collecting point cloud data of the front area of ​​the photovoltaic array installation as first point cloud data, performing three-dimensional reconstruction on the first point cloud data to generate a DEM image; Step S3: during the construction of the photovoltaic array, the orthophoto image and the second point cloud data of the field area are collected according to the set collection frequency, a DOM image is generated based on the orthophoto image collected once, and a DSM digital surface model is generated based on the spatial positions of the generated multiple DOM images and the second point cloud data; Step S4: Establish a photovoltaic array installation quality recognition model, the photovoltaic array installation quality recognition model recognizes the DOM image, determines the construction stage of a single photovoltaic array string, and verifies the construction stage judgment result with the second point cloud data of the corresponding acquisition time; fits and compares the DEM image and the second point cloud data of the corresponding acquisition time, screens and segments the point cloud data above the surface in the second point cloud data, identifies the photovoltaic array strings and their installation parameters, and compares the identified installation parameters with the installation parameters under the standard installation state of the corresponding construction stage in the standard installation library, and marks unqualified photovoltaic array strings in the DSM digital surface model based on the comparison results.

2. The method for identifying the installation quality of a photovoltaic array based on remote sensing data according to claim 1, characterized in that: After collecting orthophoto images and point cloud data, they are preprocessed. The preprocessing of orthophoto images includes: denoising, geometric correction, contrast enhancement and distortion correction; the preprocessing of point cloud data includes: filtering redundant point clouds, removing local areas, checking terrain and optimizing.

3. The method for identifying the installation quality of a photovoltaic array based on remote sensing data according to claim 1, characterized in that: In step S3, the collected point cloud data are divided according to the sub-array boundary of the photovoltaic array general plan layout diagram, taking the sub-array of the photovoltaic array as the basic unit, and the second point cloud data are generated based on the set division result.

4. The method for identifying the installation quality of a photovoltaic array based on remote sensing data according to claim 1, characterized in that: When generating the DOM image in step S3, the sub-array is used as the basic unit, and the collected orthophoto images are grouped according to the sub-array boundaries of the photovoltaic array general plan layout diagram, and the sub-array DOM image containing geographic coordinates is generated based on the group division result.

5. The method for identifying the installation quality of a photovoltaic array based on remote sensing data according to claim 1, characterized in that: The construction stages specifically include a pile foundation construction stage, a column and inclined beam installation stage, a purlin installation stage, and a component installation stage.

6. The method for identifying the installation quality of a photovoltaic array based on remote sensing data according to claim 1, characterized in that: A photovoltaic array installation quality recognition model is established based on target detection algorithm.

Images