Photovoltaic inspection data processing method and photovoltaic inspection management system

Abstract

The invention provides a photovoltaic inspection data processing method and a photovoltaic inspection management system, and relates to the technical field of photovoltaic inspection. According to the invention, after a local data processing platform obtains a power station aerial image uploaded by an unmanned aerial vehicle cruise system for a target photovoltaic power station, aerial information recording and storage are carried out on the obtained power station aerial image at a local database; the method comprises the following steps: acquiring an infrared image of a power station, calling a self-deployed photovoltaic defect detection model to perform photovoltaic defect detection on the aerial image of the power station belonging to the infrared image to obtain a photovoltaic defect report file of the aerial image of the power station, and then associatively storing actual aerial information of the aerial image of the power station and the photovoltaic defect report file in a local database. Therefore, through organic cooperation between a platform local deployment mechanism and a visible light / infrared image differentiation processing mechanism, a photovoltaic defect rapid and accurate positioning effect is realized near the target photovoltaic power station at low cost, and power station operation and maintenance personnel can conveniently and timely know the overall condition of the photovoltaic power station.

Description

Technical Field

[0001] This application relates to the field of photovoltaic inspection technology, and more specifically, to a photovoltaic inspection data processing method and a photovoltaic inspection management system. Background Technology

[0002] With the continuous development of science and technology, photovoltaic (PV) power generation technology, which utilizes the photovoltaic effect of semiconductor materials to convert solar radiation into electrical energy, has become a development direction for clean energy technologies. However, in the practical application of PV power generation technology, PV devices deployed outdoors are often subject to numerous external interference factors (such as bird droppings, uneven lighting, etc.) leading to severe hot spot effects, physical damage, and other PV defects, which can easily affect the power generation and safety of PV power plants. Therefore, regular inspections of outdoor PV power plants (e.g., rooftop PV power plants) are particularly important during operation and maintenance.

[0003] Currently, drone inspection technology is being gradually applied to the operation and maintenance of photovoltaic power plants due to its high efficiency and flexibility. However, it is worth noting that there are still many problems to be solved in the process of photovoltaic power plant inspection based on drones. Among them, how to achieve rapid and accurate location of photovoltaic defects at low cost is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a photovoltaic inspection data processing method and a photovoltaic inspection management system. This system enables the local deployment of a local data processing platform at the target photovoltaic power station (i.e., an outdoor photovoltaic power station requiring inspection). The platform automatically captures and records aerial images (whether visible light or infrared) of the power station collected by a drone patrol system. It also automatically performs local photovoltaic defect detection and defect report generation and storage separately for the infrared aerial images. Through the organic combination of the platform's local deployment mechanism and the differentiated processing mechanism for visible light / infrared images, the system effectively reduces the implementation cost and computational resource consumption of photovoltaic inspection data processing operations. This facilitates low-cost, rapid, and accurate location of photovoltaic defects near the target photovoltaic power station, allowing power station maintenance personnel to promptly understand the overall status of the photovoltaic power station.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, this application provides a photovoltaic inspection data processing method, applied to a local data processing platform deployed at a target photovoltaic power station, included in a photovoltaic inspection management system, wherein the local data processing platform is equipped with a photovoltaic defect detection model and a local database, and the method includes: The system acquires aerial images of the target photovoltaic power station uploaded by the drone patrol system during the photovoltaic inspection process, and records and stores the aerial information of the acquired aerial images in the local database. The photovoltaic defect detection model is called to perform photovoltaic defect detection on the aerial images of the power plant that belong to infrared images, and a photovoltaic defect report file of the aerial images of the power plant is obtained; The actual aerial photography information of the power plant and the photovoltaic defect report file are associated and stored in the local database.

[0006] In an optional implementation, the step of recording and storing aerial information of the acquired power plant aerial images in the local database includes: Aerial information is extracted based on the image file name of the aerial image of the power station to obtain the image type information, aerial timestamp information, shooting waypoint identification information, shooting route identification information and power station building identification information of the aerial image of the power station at the target photovoltaic power station; Search the local database for the target database tables corresponding to the waypoint identification information, the flight route identification information, and the power plant building identification information, respectively. Based on the image type information and aerial timestamp information of the aerial images of the power station, image identification information of the aerial images of the power station in the local database is constructed, and a first data table to be associated corresponding to the image identification information is created in the local database; Based on the image identification information, shooting waypoint identification information, shooting route identification information, and power plant building identification information of the aerial images of the power plant, a foreign key constraint relationship is established between the first data table to be associated and all the target database tables found, and the aerial images of the power plant are stored as key-value pairs in the local database according to the image identification information.

[0007] In an optional implementation, the step of associating and storing the actual aerial photography information of the power plant aerial images and the photovoltaic defect report file in the local database includes: In the local database, adaptive report identification information is created for the photovoltaic defect report file of the aerial image of the power station, and a second data table to be associated is created corresponding to the report identification information; Based on the image identification information and report identification information corresponding to the aerial images of the power station, a foreign key constraint relationship is established between the first data table to be associated and the second data table to be associated for the aerial images of the power station. The photovoltaic defect report file of the aerial images of the power station is stored as a key-value pair in the local database according to the report identification information.

[0008] In an optional implementation, the method further includes: Obtain model training sample sets at different drone aerial photography altitudes, where each model training sample set includes multiple aerial infrared sample images of power plants with photovoltaic defects labeled at the corresponding drone aerial photography altitude; The initial object detection model is trained based on the obtained model training sample set to obtain the corresponding object detection model; The target detection model is sequentially pruned and quantized to obtain the photovoltaic defect detection model. The photovoltaic defect detection model is then loaded and cached to complete the local deployment of the photovoltaic defect detection model.

[0009] In an optional implementation, before the step of calling the photovoltaic defect detection model to perform photovoltaic defect detection on the aerial images of the power plant belonging to infrared images, the method further includes: The actual system load level of the local data processing platform is detected, and the target image batch processing volume that matches the actual system load level is determined based on the pre-configured inverse correlation mapping relationship between different system load levels and different image batch processing volumes. The aerial images of power plants belonging to infrared images are divided into batches according to the target image batch processing volume. For each batch of aerial images of power plants, the step of calling the photovoltaic defect detection model to perform photovoltaic defect detection is continued.

[0010] In an optional implementation, the method further includes: For any aerial image of a power plant, check whether the actual photovoltaic defect risk level recorded in the photovoltaic defect report file of that aerial image meets the preset alarm standard; When the actual photovoltaic defect risk level is detected to meet the preset alarm standard, an emergency repair alarm is issued for the photovoltaic power generation device corresponding to the target photovoltaic power station in the aerial image of the power station.

[0011] In an optional implementation, the photovoltaic inspection and management system further includes at least one platform access terminal, which is communicatively connected to the local data processing platform. The method further includes: In response to a power plant image detection request sent by any platform access terminal, the photovoltaic defect detection model is invoked to perform photovoltaic defect detection on the power plant image to be detected in the power plant image detection request, and the photovoltaic defect detection result information of the power plant image to be detected is obtained. The photovoltaic defect detection results are fed back to the platform access terminal, which then draws and annotates the photovoltaic defect detection results on the image of the power station to be inspected displayed on its own screen.

[0012] In an optional implementation, the method further includes: In response to a historical data analysis request sent by any platform access terminal, and according to the historical data filtering conditions included in the historical data analysis request, the system searches for target historical data in the local database that meets the historical data filtering conditions. Mathematical statistical analysis is performed on the target historical data to obtain the corresponding historical data analysis results, and the historical data analysis results are fed back to the platform access terminal for display.

[0013] In an optional implementation, the method further includes: In response to a report generation and export request sent by any platform access terminal, search for the expected report content that matches the report generation and export request; According to the expected report format for generating export request records in the report, a preset report template matching the expected report format is called to fill in the expected report content and obtain the corresponding expected export report file; The desired exported report file is sent to the platform access terminal.

[0014] Secondly, this application provides a photovoltaic inspection and management system, the system including at least one platform access terminal and a local data processing platform deployed at the target photovoltaic power station; wherein, the at least one platform access terminal is communicatively connected to the local data processing platform; the local data processing platform is equipped with a photovoltaic defect detection model and a local database, the local database can be used to record and store aerial images of the power station uploaded by the UAV patrol system for the target photovoltaic power station, and the photovoltaic defect detection model can be used to implement photovoltaic defect detection function; Each platform access terminal responds to user access operations through the provided platform access interface, generates a platform call request matching the user access operation, and sends the platform call request to the local data processing platform to drive the local data processing platform to run according to the platform call request; The local data processing platform stores computer programs and can implement the photovoltaic inspection data processing method described in any of the foregoing embodiments by running the computer programs in conjunction with the at least one platform access terminal.

[0015] In this case, the beneficial effects of the embodiments of this application may include the following: The local data processing platform deployed at the target photovoltaic power station in this application, after acquiring aerial images of the power station uploaded by the drone patrol system during photovoltaic inspection, records and stores the aerial information of the acquired aerial images (whether visible light or infrared) in its local database. Simultaneously, it calls its own deployed photovoltaic defect detection model to perform photovoltaic defect detection on the infrared aerial images, obtaining a photovoltaic defect report file for that image. Then, it associates and stores the actual aerial information of the aerial image with the photovoltaic defect report file in the local database. Thus, through the organic cooperation between the platform's local deployment mechanism and the differentiated processing mechanism for visible light / infrared images, the implementation cost and computational resource consumption of photovoltaic inspection data processing operations are effectively reduced, while simultaneously improving the efficiency of photovoltaic inspection data processing. This facilitates the rapid and accurate location of photovoltaic defects near the target photovoltaic power station at low cost, enabling power station maintenance personnel to promptly understand the overall status of the photovoltaic power station.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system composition of the photovoltaic inspection and management system provided in the embodiments of this application; Figure 2 This is one of the flowcharts illustrating the photovoltaic inspection data processing method provided in the embodiments of this application; Figure 3 This is a schematic diagram of a multi-level associated storage structure provided for a local database in an embodiment of this application; Figure 4 A second schematic flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment; Figure 5 The third flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment; Figure 6 The fourth flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment; Figure 7 The fifth flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment; Figure 8 A flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment is shown in Figure 6. Figure 9 This is the seventh flowchart illustrating the photovoltaic inspection data processing method provided in the embodiments of this application.

[0019] Icons: 10 - Photovoltaic Inspection and Management System; 100 - Local Data Processing Platform; 200 - Platform Access Terminal. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In the description of this application, it should be understood that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are used only for the convenience of describing this application and simplifying the description, and are not intended to indicate or imply that the equipment or component referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0024] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set up," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0025] Furthermore, it is understood in the description of this application that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0026] Through painstaking research, the applicant discovered that the existing drone inspection data processing mechanism relies on cloud servers, which requires costly deployment or subscription to cloud servers in advance. Furthermore, the inspection drones need to remotely transmit the photovoltaic device image data collected by high-definition camera equipment (such as high-definition visible light cameras, high-definition infrared cameras, etc.) to the cloud server before the cloud server can use pre-deployed image recognition algorithms to detect and analyze photovoltaic defects such as hot spot effects and physical damage, and finally generate a photovoltaic power station inspection report at the cloud server.

[0027] However, it is worth noting that this drone inspection data processing mechanism has problems such as high cloud server deployment and subscription costs, excessive consumption of computing resources and low overall detection efficiency due to the undifferentiated detection of photovoltaic defects in massive images, response delays caused by massive image data transmission, and difficulty for power plant operation and maintenance personnel to understand the overall status of photovoltaic power plants in a timely manner by reviewing reports. In fact, it is difficult to achieve the effect of rapid and accurate location of photovoltaic defects at low cost.

[0028] To address this, the applicant has developed a photovoltaic (PV) inspection data processing method and a PV inspection management system. This system involves deploying a local data processing platform at the target PV power station and automatically capturing and storing aerial images (both visible and infrared) of the power station taken by a drone patrol system. Specifically, it automatically performs local PV defect detection and defect report generation and storage for the infrared aerial images. Through the organic combination of the local platform deployment mechanism and the differentiated processing mechanism for visible / infrared images, the system effectively reduces the implementation cost and computational resource consumption of PV inspection data processing operations while simultaneously improving efficiency. This facilitates low-cost, rapid, and accurate PV defect location near the target PV power station, allowing power station maintenance personnel to promptly understand the overall status of the PV power station. This effectively solves the technical problems existing in the aforementioned drone inspection data processing mechanism.

[0029] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0030] Please refer to Figure 1 , Figure 1This is a schematic diagram of the system composition of the photovoltaic inspection and management system 10 provided in this application embodiment. In this application embodiment, the photovoltaic inspection and management system 10 may include a local data processing platform 100. The local data processing platform 100 is deployed at the target photovoltaic power station and can communicate with the drone inspection system responsible for the target photovoltaic power station (which may include multiple patrol drones equipped with high-definition visible light cameras and high-definition infrared cameras). This allows the system to directly and quickly acquire aerial images of the power station (including aerial visible light images of the power station captured by high-definition visible light cameras and aerial infrared images of the power station captured by high-definition infrared cameras) collected in real time during any photovoltaic inspection at the target photovoltaic power station, record and store the aerial information, and then call upon its own locally deployed... The photovoltaic (PV) defect detection model performs local PV defect detection on acquired aerial images of power plants (i.e., PV aerial infrared images of power plants), which are classified as infrared images. Simultaneously, it utilizes its own deployed local database to record and store PV defect report files for the acquired aerial images and PV aerial infrared images of power plants. These reports record information such as the type of PV defect detected in the PV aerial infrared images (e.g., hot spot effect, physical damage, etc.), the number of PV defects, the distribution range of PV defects, the confidence level of PV defects, and the risk level of PV defects (which positively describes the severity of the corresponding PV defects). This allows for low-cost, rapid, and accurate PV defect location near the target PV power plant, facilitating timely understanding of the overall status of the PV power plant by power plant maintenance personnel. The local data processing platform 100 can be implemented using computer equipment such as personal computers and web servers.

[0031] In this embodiment, the local data processing platform 100 can pre-allocate an aerial image cache space for the UAV inspection system to perform image caching processing on the power plant aerial images uploaded by the UAV inspection system during photovoltaic inspection. The local data processing platform 100 can detect in real-time or periodically whether uploaded power plant aerial images exist in the aerial image cache space. Upon detection, it automatically retrieves the uploaded power plant aerial images from the cache space for aerial information recording and storage. It also performs local photovoltaic defect detection separately for infrared images of power plant aerial images, and then uses a local database to complete the local recording and storage of the corresponding photovoltaic defect report files, facilitating subsequent historical inspection data analysis.

[0032] During this process, it is understood that when the local data processing platform 100 establishes a communication connection with the UAV patrol system, it agrees on the file naming conventions for power plant aerial images. This allows the UAV patrol system to name the image files according to the aforementioned file naming conventions when uploading the aerial images of the power plant it has captured. For each captured aerial image of the power plant, the local data processing platform 100 can also perform image duplicate upload detection or file naming information omission checks based on the actual file name of the image. This allows for differentiated processing of duplicate uploaded images or images with non-standard names (e.g., removing duplicate uploaded images, prompting the UAV patrol system to supplement the naming information for the corresponding non-standard images, marking and storing duplicate uploaded images or images with non-standard names, etc.).

[0033] In this embodiment of the application, the photovoltaic inspection and management system 10 may include at least one platform access terminal 200, each of which is communicatively connected to the local data processing platform 100, so that power plant operation and maintenance personnel can quickly access the local data processing platform 100 through the at least one platform access terminal 200 and promptly understand the overall status of the photovoltaic power plant.

[0034] In this embodiment, each platform access terminal 200 and the local data processing platform 100 can achieve front-end and back-end separation using a RESTful API interface, and WebSocket communication between the platform access terminal 200 and the local data processing platform 100 can be achieved using the Flask-SocketIO protocol. Each platform access terminal 200 is a terminal device (which can be, but is not limited to, smartphones, laptops, personal computers, etc.) that can access the local data processing platform 100 through a browser.

[0035] When power plant maintenance personnel need to access the local data processing platform 100 through a platform access terminal 200, the platform access terminal 200 will provide the power plant maintenance personnel with a platform access interface (which can use Bootstrap under the Flask framework). The system is built with 5 technologies and integrates three core functional modules—image upload and detection, historical inspection data analysis, and report generation and export—into a unified web interface, enabling the platform access interface to respond in less than 2 seconds. This allows power plant maintenance personnel to perform user access operations on the platform access interface (which may include, but are not limited to, power plant image upload and detection operations, data configuration operations, and report generation and export operations in a specified format). The platform access terminal responds to these user access operations and generates matching platform call requests (e.g., a power plant image detection request matching the power plant image upload and detection operation, a historical data analysis request matching the data configuration operation, and a report generation and export request matching the report generation and export operation in a specified format). These platform call requests are then sent to the local data processing platform 100, which in turn drives the platform to run according to the platform call requests, achieving a visualized real-time interactive effect between power plant maintenance personnel and the photovoltaic inspection management system 10.

[0036] In this embodiment, the local data processing platform 100 pre-stores a specific computer program related to photovoltaic inspection data processing. By running the specific computer program, it can cooperate with at least one platform access terminal 200 to adaptively manage the photovoltaic inspection data (including aerial images of the power station uploaded by the UAV cruise system and photovoltaic defect report files generated by the photovoltaic defect detection model based on the aerial images of the power station) at the target photovoltaic power station. This facilitates the rapid and accurate location of photovoltaic defects at low cost near the target photovoltaic power station, and allows power station operation and maintenance personnel to understand the overall status of the photovoltaic power station in a timely manner.

[0037] Understandable, Figure 1 The block diagram shown is only a schematic diagram of one component of the photovoltaic inspection and management system 10. The photovoltaic inspection and management system 10 may also include components such as... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0038] In this application, to ensure that the local data processing platform 100 in the aforementioned photovoltaic inspection management system 10 can achieve rapid and accurate location of photovoltaic defects at low cost near the target photovoltaic power station, and facilitate power station operation and maintenance personnel to understand the overall status of the photovoltaic power station in a timely manner, this application embodiment achieves the aforementioned objective by providing a photovoltaic inspection data processing method. The photovoltaic inspection data processing method provided in this application will be described in detail below.

[0039] Please refer to Figure 2 , Figure 2 This is one of the flowcharts illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, Figure 2 The photovoltaic inspection data processing method shown may include steps S210 to S230.

[0040] Step S210: Acquire aerial images of the target photovoltaic power station uploaded by the drone patrol system during the photovoltaic inspection process, and record and store the aerial information of the acquired aerial images in the local database.

[0041] In this embodiment, the local data processing platform 100 can detect in real time or periodically whether there are power plant aerial images uploaded by the UAV patrol system in the aerial image cache space. When it detects that there are uploaded power plant aerial images, it automatically retrieves all uploaded power plant aerial images from the aerial image cache space. Then, the local data processing platform 100 extracts aerial information based on the image file name of the retrieved power plant aerial image (which records at least part of the aerial information of the corresponding power plant aerial image) to obtain the actual aerial information of the power plant aerial image at the target photovoltaic power plant (including image type information, aerial timestamp information, shooting waypoint identification information, shooting route identification information, and power plant building identification information, etc. during a photovoltaic inspection operation). Then, based on the extracted actual aerial information, the power plant aerial image and its actual aerial information are recorded and stored in the local database so as to quickly realize the historical inspection data analysis function.

[0042] Optionally, in this embodiment, the step "recording and storing aerial information of power plant aerial images in the local database" in step S210 above may include sub-steps A to D, so as to improve the data storage integrity and data consistency of power plant aerial images and their actual aerial information through the effective combination of foreign key constraint technology and multi-level association storage technology, which facilitates the subsequent realization of fast data retrieval and search effect.

[0043] Sub-step A: Extract aerial information based on the image file name of the obtained aerial image of the power station to obtain the image type information, aerial timestamp information, shooting waypoint identification information, shooting route identification information, and power station building identification information of the aerial image of the power station at the target photovoltaic power station.

[0044] In this embodiment, the aerial images of power plants uploaded by the UAV patrol system typically need to follow the file naming convention of "UAV equipment code_image aerial timestamp_image capture sequence number of individual buildings_image type (e.g., 'T' represents infrared image, 'V' represents visible light image)_identifier of the photographed power plant building_UAV waypoint number". This allows the image file name of the corresponding power plant aerial image to be directly extracted using a regular expression analysis algorithm to extract the image capture sequence number, image type information, aerial timestamp information, and power plant building identification information. Then, the local data processing platform 100 can quickly find the target waypoint mapping rule that matches the aforementioned power plant building identification information by calling the pre-configured waypoint mapping rule query engine (which supports a rule caching mechanism to improve query performance and can support different waypoint mapping rules for different power plant buildings, as well as some power plant buildings using a general waypoint mapping rule). In order to use the target waypoint mapping rule to calculate the waypoint identification based on the aforementioned image shooting sequence number, the shooting waypoint identification information of the aerial image of the power plant can be obtained. At the same time, the local data processing platform 100 can query the shooting route identification information that matches the aforementioned power plant building identification information based on the pre-agreed aerial shooting order of all inspection routes (wherein, each inspection route can cover multiple power plant buildings, and the aerial shooting order of the multiple power plant buildings involved in the same inspection route is fixed).

[0045] Sub-step B: Search the local database for the target database tables corresponding to the waypoint identification information, the flight route identification information, and the power station building identification information, respectively.

[0046] In this embodiment, the local data processing platform 100 can search for the target database table whose corresponding actual primary key information is the shooting waypoint identification information, the shooting route identification information, and the power station building identification information, based on the actual primary key (PK) information of each database table recorded in the local database. Figure 3 For example, the target database table corresponding to the waypoint identification information is... Figure 3 The "shooting waypoint data table" (which records attribute information such as the UAV waypoint number, shooting waypoint name, and shooting waypoint description for each shooting waypoint) is the target database table corresponding to the shooting route identification information. Figure 3The "shooting route data table" (which records attribute information such as the name of the corresponding shooting route) is the target database table corresponding to the power station building identification information. Figure 3 The "Power Plant Building Data Table" contains attribute information such as the building name of the corresponding power plant building.

[0047] Sub-step C: Based on the image type information and aerial timestamp information of the aerial image of the power station, construct the image identification information of the aerial image of the power station in the local database, and create a first data table to be associated in the local database corresponding to the image identification information.

[0048] In this embodiment, the first data table to be associated is... Figure 3 The "Aerial Image Data Table" that corresponds to the actual primary key information of the image identification information can record attribute information such as image file name, image type, and aerial timestamp of the corresponding power station aerial image.

[0049] Sub-step D: Based on the image identification information, shooting waypoint identification information, shooting route identification information, and power plant building identification information of the aerial image of the power plant, establish foreign key constraint association relationships between the first data table to be associated and all the target database tables found, and store the aerial image of the power plant in the local database as key-value pairs according to the image identification information.

[0050] In this embodiment, the local data processing platform 100 can utilize database indexing technology (e.g., B+ tree indexing technology) and foreign key constraint technology in the local database to establish multi-level association index relationships (i.e., foreign key constraint association relationships) between multiple database tables by mutually inputting the primary key information of other database tables as foreign key (FK) information. This facilitates the effective combination of foreign key constraint technology and multi-level association storage technology in the local database, thereby improving the data storage integrity and consistency of power plant aerial images and their actual aerial information, and facilitating subsequent rapid data retrieval and search.

[0051] by Figure 3For example, the primary key information of the "Shooting Waypoint Data Table", "Power Plant Building Data Table" and "Shooting Route Data Table" are each used as a foreign key information of the "Aerial Image Data Table". The primary key information of the "Power Plant Building Data Table" is used as a foreign key information of the "Shooting Waypoint Data Table". The primary key information of the "Shooting Route Data Table" is used as a foreign key information of the "Power Plant Building Data Table" through a "Route-Building Sequence Table". The "Route-Building Sequence Table" can describe the aerial shooting order of all power plant buildings involved in a certain inspection route through a single sequence (each sequence corresponds to a separate sequence identifier, that is, the sequence identifier is the actual primary key information of the "Route-Building Sequence Table"). The shooting route identifier and all power plant building identifiers in the same route-building sequence table are each used as a foreign key information.

[0052] Furthermore, for any aerial image of a power plant, the image identification information of the aerial image can be used as the key, and the aerial image itself can be used as the value corresponding to the key. Then, the image storage path of the aerial image can be recorded in the corresponding first data table to be associated, so as to achieve the key-value pair storage effect of the aerial image of the power plant.

[0053] Step S220: Call the photovoltaic defect detection model to perform photovoltaic defect detection on the aerial image of the power station that belongs to the infrared image, and obtain the photovoltaic defect report file of the aerial image of the power station.

[0054] In this embodiment, for each aerial image of a power plant that is an infrared image captured by the local data processing platform 100, the local data processing platform 100 calls its locally deployed photovoltaic defect detection model to perform photovoltaic defect detection on the aerial image of the power plant, so as to obtain photovoltaic defect detection result information of the aerial image of the power plant (including the number of photovoltaic defects involved in the aerial image of the power plant, the type of photovoltaic defect, the distribution range of photovoltaic defects, and the confidence level of photovoltaic defects). Then, the local data processing platform 100 performs photovoltaic defect severity assessment based on the obtained photovoltaic defect detection result information to obtain the current photovoltaic defect risk level of the photovoltaic power generation device corresponding to the aerial image of the power plant (wherein, the higher the photovoltaic defect severity, the higher the corresponding photovoltaic defect risk level). Then, the photovoltaic defect detection result information and photovoltaic defect risk level of the aerial image of the power plant are integrated to generate a photovoltaic defect report file for the aerial image of the power plant. Among them, when the local data processing platform 100 captures a large number of aerial infrared images of the power plant at one time, it can realize the effect of concurrent detection of multiple images based on Celery asynchronous task queue to improve the efficiency of large-batch image detection.

[0055] Step S230: The actual aerial photography information of the power station aerial images and the photovoltaic defect report file are associated and stored in the local database.

[0056] In this embodiment, after the local data processing platform 100 generates a corresponding photovoltaic defect report file for any aerial infrared image of a power plant, it creates adapted report identification information for the photovoltaic defect report file of that aerial image in the local database, and simultaneously establishes a second data table to be associated (i.e., ...) corresponding to the report identification information. Figure 3 The corresponding actual primary key information in the data is the "defect report data table" of the report identification information. Then, based on the image identification information and report identification information corresponding to the aerial image of the power station, a foreign key constraint relationship is established between the first data table to be associated and the second data table to be associated for the aerial image of the power station (which is achieved by "using the actual primary key information 'image identification' of the first data table to be associated as a foreign key information of the second data table to be associated"). The photovoltaic defect report file of the aerial image of the power station is stored in the local database according to the report identification information (which is achieved by recording the various attribute contents involved in the photovoltaic defect report file into the corresponding second data table to be associated). By combining multi-level association storage technology and foreign key constraint technology, the data storage integrity and data consistency between the aerial infrared image of the power station and its actual aerial information and the photovoltaic defect report file are improved, which facilitates the subsequent rapid data retrieval and search effect.

[0057] Therefore, by executing the above steps S210 to S230, this application utilizes the local data processing platform 100 to automatically capture and record aerial images of the target photovoltaic power station collected by the UAV patrol system, and separately implements local photovoltaic defect detection and defect report file generation and storage functions for the power station's aerial infrared images. Through the organic cooperation between the platform's local deployment mechanism and the visible light / infrared image differential processing mechanism, the implementation cost and computing resource consumption of photovoltaic inspection data processing operations are effectively reduced, while improving the efficiency of photovoltaic inspection data processing. This facilitates the low-cost, rapid, and accurate location of photovoltaic defects near the target photovoltaic power station, enabling power station operation and maintenance personnel to understand the overall status of the photovoltaic power station in a timely manner.

[0058] Alternatively, please refer to Figure 4 , Figure 4 This is a second schematic flowchart of the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, with Figure 2 Compared to the photovoltaic inspection data processing method shown, Figure 4 The photovoltaic inspection data processing method shown may further include steps S240 and S250 (which combine) before performing step S220. Figure 2Step S220) ensures that the local data processing platform 100 can effectively meet the dynamic load balancing requirements of the system during the photovoltaic defect detection process, and ensures that the local data processing platform 100 can operate smoothly.

[0059] Step S240: Detect the actual system load level of the local data processing platform, and determine the target image batch processing volume that matches the actual system load level based on the pre-configured inverse correlation mapping relationship between different system load levels and different image batch processing volumes.

[0060] In this embodiment, the local data processing platform 100 can assess its system load level by monitoring system load parameters such as CPU usage, memory usage, and task queue length in real time. A higher system load level indicates a greater current system load on the local data processing platform 100. In this case, to ensure dynamic system load balancing, the corresponding image batch processing size will be smaller. For example, the image batch processing sizes for "high load level," "medium load level," and "low load level" can be set to 4, 8, and 16, respectively.

[0061] Step S250: Divide the aerial images of power plants belonging to infrared images into batches according to the target image batch processing volume, and for each batch of aerial images of power plants, call the photovoltaic defect detection model to perform photovoltaic defect detection, and obtain the photovoltaic defect report file of the corresponding aerial images of power plants.

[0062] In this embodiment, after the image batching operation of the captured aerial images of the power plant is completed according to the target image batch processing volume, the above step S220 can be executed sequentially for each batch of aerial images of the power plant to complete the photovoltaic defect detection operation of the aerial images of the power plant in batches, so as to ensure that the local data processing platform 100 can operate smoothly.

[0063] Therefore, by executing the above steps S210, S240, S250 and S230, this application can achieve rapid and accurate location of photovoltaic defects at low cost near the target photovoltaic power station, while effectively ensuring that the local data processing platform 100 can effectively meet the dynamic load balancing requirements of the system and operate smoothly.

[0064] Optionally, please follow Figure 5 , Figure 5 This is the third flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, [the method is related to...]. Figure 2 or Figure 4 Compared to the photovoltaic inspection data processing method shown, Figure 5The photovoltaic inspection data processing method shown may also include steps S260 to S270 to remind power plant operation and maintenance personnel to carry out emergency repairs on photovoltaic power generation devices with high-risk defects in a timely manner, so as to avoid serious power generation loss and associated fire damage.

[0065] Step S260: For any aerial image of a power plant, check whether the actual photovoltaic defect risk level recorded in the photovoltaic defect report file of the aerial image of the power plant meets the preset alarm standard.

[0066] In this embodiment, the "power station aerial image" in step S260 is essentially an infrared image, and the preset alarm standard is used to describe the risk level distribution range of high-risk photovoltaic defects.

[0067] Step S270: When the actual photovoltaic defect risk level is detected to meet the preset alarm standard, an emergency repair alarm is issued for the photovoltaic power generation device corresponding to the target photovoltaic power station in the aerial image of the power station.

[0068] In this embodiment, when the actual photovoltaic defect risk level corresponding to a certain aerial photograph of a power station belonging to infrared images meets the preset alarm standard, it indicates that the photovoltaic power generation device corresponding to the target photovoltaic power station in the aerial photograph of the power station actually has a high-risk defect. At this time, the local data processing platform 100 can further determine the actual distribution location of the photovoltaic power generation device at the target photovoltaic power station based on the actual aerial photograph information of the stored aerial photograph of the power station. Then, it will send an emergency maintenance alarm to the operation and maintenance management terminal or platform access terminal 200 held by the power station operation and maintenance personnel through information push, so as to instruct the power station operation and maintenance personnel to carry out emergency maintenance on the photovoltaic power generation device in a timely manner.

[0069] Therefore, by executing the above steps S260 to S270, this application can remind power plant operation and maintenance personnel to carry out emergency repairs on photovoltaic power generation devices with high-risk defects in a timely manner, so as to avoid serious power generation loss and associated fire damage.

[0070] Alternatively, please refer to Figure 6 , Figure 6 This is the fourth flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, [the method is related to...]. Figure 2 , Figure 4 or Figure 5 Compared to the photovoltaic inspection data processing method shown, Figure 6 The photovoltaic inspection data processing method shown may also include steps S310 to S330 to ensure that the locally deployed photovoltaic inspection model can achieve a lightweight deployment effect while ensuring high-precision photovoltaic defect detection at different cruise altitudes, thus avoiding excessive storage resource consumption.

[0071] Step S310: Obtain the model training sample set at different drone aerial photography altitudes.

[0072] In this embodiment, each model training sample set includes multiple aerial infrared sample images of power plants labeled with photovoltaic defects at the corresponding drone aerial photography altitude; the photovoltaic defect types involved in the multiple aerial infrared sample images of power plants in the same model training sample set can be completely the same, partially the same, or completely different, and the number of photovoltaic defects involved in the multiple aerial infrared sample images of power plants can be the same or different.

[0073] Step S320: Train the initial object detection model based on the obtained model training sample set to obtain the corresponding object detection model.

[0074] In this embodiment, knowledge distillation technology can be used in conjunction with model training sample sets at different drone aerial photography altitudes to jointly train an initialization network model with target detection capabilities (whose model architecture can be, but is not limited to, YOLOv8n, TensorFlow, PyTorch, OpenCV, scikit-image, etc.). This ensures that the trained target detection model possesses high-accuracy photovoltaic defect detection capabilities at different cruising altitudes. The target loss function that the initialization network model needs to minimize during model training can be implemented using a combination of Focal Loss and IoU Loss functions.

[0075] Step S330: Perform model pruning and model quantization on the target detection model in sequence to obtain the photovoltaic defect detection model, and load and cache the photovoltaic defect detection model to complete the local deployment of the photovoltaic defect detection model.

[0076] In this embodiment, redundant parameters of the target detection model can be removed by model pruning technology, and the model size of the target detection model can be further compressed by model quantization technology, so that the corresponding photovoltaic inspection model can maintain a small size (for example, the corresponding model size can be compressed to about 21MB) while having a high defect detection accuracy (for example, about 86.7%).

[0077] Therefore, by executing the above steps S310 to S330, this application can ensure that the locally deployed photovoltaic inspection model can achieve a lightweight deployment effect while ensuring high-precision photovoltaic defect detection at different cruise altitudes, thus avoiding excessive storage resource consumption.

[0078] Alternatively, please refer to Figure 7 , Figure 7This is the fifth flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, [the method is related to...]. Figure 2 , Figure 4 , Figure 5 and Figure 6 Compared to the photovoltaic inspection data processing method shown in any of the attached figures, Figure 7 The photovoltaic inspection data processing method shown may also include steps S410 to S420, which are used to detect photovoltaic defects in one or more images of the power station to be inspected uploaded by the power station operation and maintenance personnel, and to feed back the corresponding detection results to the power station operation and maintenance personnel for visualization.

[0079] Step S410: Respond to the power station image detection request sent by any platform access terminal, and call the photovoltaic defect detection model to perform photovoltaic defect detection on the power station image to be detected in the power station image detection request, and obtain the photovoltaic defect detection result information of the power station image to be detected.

[0080] In this embodiment, when power plant operation and maintenance personnel perform a power plant image upload and detection operation through the platform access interface provided by any platform access terminal 200, the platform access terminal 200 will generate a corresponding power plant image detection request (which includes at least one infrared image of the power plant to be detected, specified by the power plant image upload and detection operation), and send the generated power plant image detection request to the local data processing platform 100. The local data processing platform 100 will then perform photovoltaic defect detection on each uploaded power plant image to be detected, obtaining photovoltaic defect detection result information for each power plant image to be detected (which can use COCO format to achieve information recording). When the number of power plant images specified by the power plant image upload and detection operation is too large, the local data processing platform 100 can achieve multi-image concurrent detection based on the Celery asynchronous task queue to improve the efficiency of large-batch image detection.

[0081] Step S420: The photovoltaic defect detection result information is fed back to the platform access terminal, so that the platform access terminal can draw and mark the photovoltaic defect detection result information on the image of the power station to be inspected displayed on its own.

[0082] In this embodiment, when the platform access terminal 200 that sends the power plant image detection request receives the corresponding photovoltaic defect detection result information, it can call the Canvas 2D component to draw and annotate the photovoltaic defect type, photovoltaic defect distribution range, and photovoltaic defect confidence level involved in the photovoltaic defect detection result information based on the power plant image to be detected displayed on the platform access interface, so as to form and display the corresponding photovoltaic defect annotation image at the platform access terminal 200.

[0083] Therefore, this application can perform photovoltaic defect detection on one or more images of a power station to be inspected uploaded by the power station operation and maintenance personnel by executing the above steps S410 to S430, and feed back the corresponding detection results to the power station operation and maintenance personnel for visualization display.

[0084] Alternatively, please refer to Figure 8 , Figure 8 This is the sixth flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, [the method is related to...]. Figure 2 , Figures 4-7 Compared to the photovoltaic inspection data processing method shown in any of the attached figures, Figure 8 The photovoltaic inspection data processing method shown may also include steps S430 to S440, in response to the historical inspection data analysis instructions issued by the power plant operation and maintenance personnel, to perform defect distribution trend analysis on the relevant historical data stored in the local database, and to promptly inform the power plant operation and maintenance personnel of the corresponding analysis results for operation and maintenance reference.

[0085] Step S430: Respond to the historical data analysis request sent by any platform access terminal, and search for target historical data that meets the historical data filtering conditions in the local database according to the historical data filtering conditions included in the historical data analysis request.

[0086] In this embodiment, when the power plant operation and maintenance personnel perform a data configuration operation (i.e., configure the relevant filtering conditions for the historical inspection data to be analyzed (which may include, but are not limited to, specifying the aerial photography time period, specifying the aerial photography building, specifying the aerial photography route, etc.) through the platform access interface provided by any platform access terminal 200), the platform access terminal 200 will generate a historical data analysis request and send the generated historical data analysis request to the local data processing platform 100. The local data processing platform 100 will then search for the target historical data corresponding to the historical data filtering conditions in the local database according to the historical data filtering conditions recorded in the historical data analysis request (which may include the photovoltaic defect report files of all power plant aerial infrared images that meet the historical data filtering conditions).

[0087] Step S440: Perform mathematical statistical analysis on the target historical data to obtain the corresponding historical data analysis results, and then feed the historical data analysis results back to the platform access terminal for display.

[0088] In this embodiment, the distribution trend of photovoltaic defects matching the target historical data can be deduced by performing mathematical statistical analysis on the target historical data (i.e., the historical data analysis results mentioned above). At this time, the historical data analysis results can be fed back to the platform access terminal 200 for display, so as to inform the power plant operation and maintenance personnel in a timely manner for operation and maintenance reference and assist the power plant operation and maintenance personnel in making reasonable operation and maintenance decisions.

[0089] Therefore, this application can perform the above steps S430 to S440 to respond to the historical inspection data analysis instructions issued by the power plant operation and maintenance personnel, perform defect distribution trend analysis on the relevant historical data stored in the local database, and promptly inform the power plant operation and maintenance personnel of the corresponding analysis results for operation and maintenance reference, so as to assist the power plant operation and maintenance personnel in making reasonable operation and maintenance decisions.

[0090] Alternatively, please refer to Figure 9 , Figure 9 This is the seventh flowchart illustrating the photovoltaic inspection data processing method provided in this application embodiment. In this application embodiment, it is related to... Figure 2 , Figures 4-8 Compared to the photovoltaic inspection data processing method shown in any of the attached figures, Figure 9 The photovoltaic inspection data processing method shown may also include steps S450 to S470, in response to the power plant operation and maintenance personnel’s instruction to generate and export a report in a specified format, automatically generating a report in a specified format for the report content that the power plant operation and maintenance personnel are concerned about, so as to facilitate the power plant operation and maintenance personnel to archive and organize the power plant operation and maintenance data.

[0091] Step S450: In response to a report generation and export request sent by any platform access terminal, search for the expected report content that matches the report generation and export request.

[0092] In this embodiment, the expected report content pointed to by the report generation and export request may be, but is not limited to: photovoltaic defect detection result information of one or more images of the power plant to be inspected uploaded by the power plant operation and maintenance personnel, historical data analysis results of historical data of the target identified by the power plant operation and maintenance personnel, photovoltaic defect report files and actual aerial photography information of one or more aerial images of the power plant specified by the power plant operation and maintenance personnel, etc.

[0093] Step S460: According to the expected report format of the report generation export request record, call the preset report template that matches the expected report format to fill in the expected report content and obtain the corresponding expected export report file.

[0094] In this embodiment, the local data processing platform 100 is compatible with multiple report formats (e.g., HTML report format, PDF report format, Excel report format, etc.), and has an independent preset report template configured for each compatible report format (e.g., a preset report template for HTML report format can be generated using the Jinja2 template engine, a preset report template for PDF report format can be generated using the ReportLab database, and a preset report template for Excel report format can be generated using the Pandas analysis tool). The platform access interface provided by the platform access terminal 200 used by the power plant operation and maintenance personnel displays the format icons of all the report formats supported by the local data processing platform 100, so that the power plant operation and maintenance personnel can select the appropriate desired report format according to their own report display needs.

[0095] Step S470: Send the desired export report file to the platform access terminal.

[0096] Therefore, this application can automatically generate reports in a specified format based on the report content that the power plant operation and maintenance personnel focus on by executing the above steps S450 to S470, thus facilitating the power plant operation and maintenance personnel to archive and organize power plant operation and maintenance data.

[0097] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of the apparatus, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0098] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium and includes several instructions to cause a computer device (which may be a laptop, personal computer, web server, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application as the aforementioned local data processing platform 100. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0099] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A photovoltaic inspection data processing method, characterized in that, A local data processing platform deployed at a target photovoltaic power station, included in a photovoltaic inspection and management system, wherein the local data processing platform is equipped with a photovoltaic defect detection model and a local database, the method comprising: The system acquires aerial images of the target photovoltaic power station uploaded by the drone patrol system during the photovoltaic inspection process, and records and stores the aerial information of the acquired aerial images in the local database. The photovoltaic defect detection model is called to perform photovoltaic defect detection on the aerial images of the power station that belong to infrared images, and a photovoltaic defect report file of the aerial images of the power station is obtained; The actual aerial photography information of the power plant and the photovoltaic defect report file are associated and stored in the local database.

2. The method according to claim 1, characterized in that, The step of recording and storing aerial information of the acquired power plant aerial images in the local database includes: Aerial information is extracted based on the image file name of the aerial image of the power station to obtain the image type information, aerial timestamp information, shooting waypoint identification information, shooting route identification information and power station building identification information of the aerial image of the power station at the target photovoltaic power station; Search the local database for the target database tables corresponding to the waypoint identification information, the flight route identification information, and the power plant building identification information, respectively. Based on the image type information and aerial timestamp information of the aerial images of the power station, image identification information of the aerial images of the power station in the local database is constructed, and a first data table to be associated corresponding to the image identification information is created in the local database; Based on the image identification information, shooting waypoint identification information, shooting route identification information, and power plant building identification information of the aerial images of the power plant, a foreign key constraint relationship is established between the first data table to be associated and all the target database tables found, and the aerial images of the power plant are stored as key-value pairs in the local database according to the image identification information.

3. The method according to claim 2, characterized in that, The step of associating and storing the actual aerial photography information of the power plant and the photovoltaic defect report file in the local database includes: In the local database, adaptive report identification information is created for the photovoltaic defect report file of the aerial image of the power station, and a second data table to be associated is created corresponding to the report identification information; Based on the image identification information and report identification information corresponding to the aerial images of the power station, a foreign key constraint relationship is established between the first data table to be associated and the second data table to be associated for the aerial images of the power station. The photovoltaic defect report file of the aerial images of the power station is stored as a key-value pair in the local database according to the report identification information.

4. The method according to claim 1, characterized in that, The method further includes: Obtain model training sample sets at different drone aerial photography altitudes, where each model training sample set includes multiple aerial infrared sample images of power plants with photovoltaic defects labeled at the corresponding drone aerial photography altitude; The initial object detection model is trained based on the obtained model training sample set to obtain the corresponding object detection model; The target detection model is sequentially pruned and quantized to obtain the photovoltaic defect detection model. The photovoltaic defect detection model is then loaded and cached to complete the local deployment of the photovoltaic defect detection model.

5. The method according to claim 1, characterized in that, Before the step of calling the photovoltaic defect detection model to perform photovoltaic defect detection on the aerial images of the power plant that belong to infrared images, the method further includes: The actual system load level of the local data processing platform is detected, and the target image batch processing volume that matches the actual system load level is determined based on the pre-configured inverse correlation mapping relationship between different system load levels and different image batch processing volumes. The aerial images of power plants belonging to infrared images are divided into batches according to the target image batch processing volume. For each batch of aerial images of power plants, the step of calling the photovoltaic defect detection model to perform photovoltaic defect detection is continued.

6. The method according to claim 1, characterized in that, The method further includes: For any aerial image of a power plant, check whether the actual photovoltaic defect risk level recorded in the photovoltaic defect report file of that aerial image meets the preset alarm standard; When the actual photovoltaic defect risk level is detected to meet the preset alarm standard, an emergency repair alarm is issued for the photovoltaic power generation device corresponding to the target photovoltaic power station in the aerial image of the power station.

7. The method according to any one of claims 1-6, characterized in that, The photovoltaic inspection and management system further includes at least one platform access terminal, which is communicatively connected to the local data processing platform. The method further includes: In response to a power plant image detection request sent by any platform access terminal, the photovoltaic defect detection model is invoked to perform photovoltaic defect detection on the power plant image to be detected in the power plant image detection request, and the photovoltaic defect detection result information of the power plant image to be detected is obtained. The photovoltaic defect detection results are fed back to the platform access terminal, which then draws and annotates the photovoltaic defect detection results on the image of the power station to be inspected displayed on its own screen.

8. The method according to claim 7, characterized in that, The method further includes: In response to a historical data analysis request sent by any platform access terminal, and according to the historical data filtering conditions included in the historical data analysis request, the system searches for target historical data in the local database that meets the historical data filtering conditions. Mathematical statistical analysis is performed on the target historical data to obtain the corresponding historical data analysis results, and the historical data analysis results are fed back to the platform access terminal for display.

9. The method according to claim 8, characterized in that, The method further includes: In response to a report generation and export request sent by any platform access terminal, search for the expected report content that matches the report generation and export request; According to the expected report format for generating export request records in the report, a preset report template matching the expected report format is called to fill in the expected report content and obtain the corresponding expected export report file; The desired exported report file is sent to the platform access terminal.

10. A photovoltaic inspection and management system, characterized in that, The system includes at least one platform access terminal and a local data processing platform deployed at the target photovoltaic power station; wherein, the at least one platform access terminal is communicatively connected to the local data processing platform; the local data processing platform is equipped with a photovoltaic defect detection model and a local database, the local database can be used to record and store aerial images of the power station uploaded by the UAV patrol system for the target photovoltaic power station, and the photovoltaic defect detection model can be used to implement photovoltaic defect detection functions; Each platform access terminal responds to user access operations through the provided platform access interface, generates a platform call request matching the user access operation, and sends the platform call request to the local data processing platform to drive the local data processing platform to run according to the platform call request; The local data processing platform stores a computer program, and can implement the photovoltaic inspection data processing method according to any one of claims 1-9 by running the computer program in conjunction with the at least one platform access terminal.

Images