Large-scale photovoltaic power station string fault positioning and processing system and method

Abstract

The invention discloses a large-scale photovoltaic power station string fault positioning and processing system and method, and relates to the technical field of photovoltaic power station operation and maintenance, and the method comprises the steps: obtaining a photovoltaic power station layout planning map, building a three-dimensional real scene model, endowing each assembly with a three-dimensional coordinate, and building an assembly coordinate database; carrying out inspection according to a visible light and thermal imaging camera carried by the unmanned aerial vehicle, identifying a fault group string, pairing three-dimensional coordinates of components, and generating a fault group string thermodynamic diagram; based on the fault string thermodynamic diagram and the three-dimensional live-action model, planning an optimal arrival path of the fault string, and issuing navigation information to a terminal system in real time; navigation information is issued to a terminal system in real time, a maintainer wears terminal equipment to receive three-dimensional live-action navigation, transmission type visual guidance is achieved, a maintenance result is transmitted back through the terminal after maintenance is completed, the system automatically updates the real-time state of a photovoltaic power station assembly, and operation and maintenance closed-loop management is achieved. The method has the advantages that the method adapts to a large-scale power station, and positioning efficiency is not affected by scale expansion.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, specifically to a system and method for locating and handling string faults in large-scale photovoltaic power plants. Background Technology

[0002] Existing technologies for locating and handling string faults in large-scale photovoltaic power plants largely rely on manual inspections or fault detection based on traditional monitoring systems. These methods suffer from slow response times, inaccurate location, and low processing efficiency. Due to the massive scale of photovoltaic power plants, traditional fault detection methods often cannot obtain the precise location of faulty components in real time and cannot adequately consider factors such as terrain and road obstacles, making it difficult for maintenance personnel to reach the fault area in a timely manner. Existing systems typically lack intelligent path planning and dynamic navigation functions after fault diagnosis and cannot effectively achieve closed-loop management during the fault handling process. The efficiency of post-maintenance data updates and fault record management is also low, making it difficult to provide comprehensive operation and maintenance support. Summary of the Invention

[0003] To address the aforementioned technical issues, this technical solution provides a system and method for locating and handling string faults in large-scale photovoltaic power plants. This solution resolves the problems of low inspection efficiency, poor positioning accuracy, lack of navigation, and the disconnect between the physical location and digital information of equipment, which leads to an exponential decrease in fault location efficiency as the power plant grows larger.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Methods for locating and handling string faults in large-scale photovoltaic power plants include: Obtain the layout planning map of the photovoltaic power station, establish a 3D real scene model of the photovoltaic power station, assign 3D coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. Based on the inspection by drone equipped with visible light and thermal imaging cameras, the faulty strings of the photovoltaic power station are automatically identified, the three-dimensional coordinates of the photovoltaic power station components are associated and matched, and a heat map of the faulty strings of the photovoltaic power station is generated. Based on the heat map of the faulty strings of the photovoltaic power plant and the 3D real scene model of the photovoltaic power plant, the optimal arrival path of the faulty strings of the photovoltaic power plant is planned, and navigation information is sent to the terminal system in real time. Based on real-time navigation information sent to the terminal system, maintenance personnel wear terminal devices to receive 3D real-scene navigation, achieving penetrating visual guidance, accurately locating faulty strings in the photovoltaic power station, and transmitting maintenance results back through the terminal after the maintenance is completed. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance.

[0005] Preferably, a layout plan of a photovoltaic power plant is obtained based on satellite imagery; Based on oblique photography by drones, POS data of different images of photovoltaic power stations are obtained, and an unordered image sequence of POS data is constructed. The POS data includes: location: longitude, latitude, altitude; attitude: pitch, roll, heading; The SIFT scale-invariant feature transform algorithm is used to perform multi-scale processing on the disordered image sequence of POS data and construct the scale space of the disordered image sequence of POS data. Based on the scale space of the disordered image sequence of POS data, continuous Gaussian blurring is performed to generate multi-scale images, and the local extrema of each pixel are detected. Using the second derivative matrix of Gaussian, we evaluate whether the local extrema of each pixel are stable. If they are unstable, we remove the unstable local extrema. Based on eliminating unstable local extrema, retaining the key local extrema of each pixel, and calculating the gradient direction and magnitude of each pixel; Based on the gradient direction and magnitude of each pixel, local features of each pixel are extracted; Calculate the gradient information of the neighborhood around the key local extrema of each pixel to generate a feature descriptor; The feature descriptor, taking a 16x16 region as an example, is divided into 16 sub-blocks, and the gradient histograms of each sub-block are calculated in 8 directions.

[0006] Preferably, the similarity between two feature descriptors is calculated using the Euclidean distance formula, and a reasonable threshold is set to determine whether the two feature descriptors match. The reasonable threshold is 0.8; The fundamental matrix of unknown camera intrinsic parameters is estimated using the random sample consensus algorithm. Based on the known camera intrinsic parameters from the POS data, the essential matrix is ​​calculated, and the relative rotation and translation matrix between each pair of images is calculated using the least squares method. Based on the relative rotation and translation matrix between each pair of images, triangulation is performed to calculate the three-dimensional spatial position of the matching feature points; Using the three-dimensional spatial position of the matching feature points as input, the MVS multi-view stereo algorithm is used for dense reconstruction, and registration and alignment are performed for images from different perspectives. For each pixel of the matching feature point, calculate the cost of matching in the corresponding view, construct the cost volume, optimize the cost volume using a dynamic programming algorithm, and generate a disparity map. Based on the disparity map, it is converted into a depth map and fused with point cloud to establish a 3D reality model of the photovoltaic power station.

[0007] Preferably, based on a 3D real-scene model of a photovoltaic power station, semantic segmentation technology is used to classify and identify photovoltaic power station components; Based on the depth map, the center spatial coordinates of each photovoltaic power station module are calculated through three-dimensional geometric operations, and a unique identifier is assigned to each photovoltaic power station module. A photovoltaic power station component coordinate database is established based on the unique identifier assigned to each photovoltaic power station component. The photovoltaic power station component coordinate database includes: photovoltaic power station components, component coordinates, component type, model, inverter information, and the string to which they belong.

[0008] 5. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 1, wherein step S2 comprises: Based on the inspection of the photovoltaic power station by the drone equipped with visible light and thermal imaging cameras, the visible light and thermal imaging data are obtained and the data is preprocessed. Image recognition algorithms are used to identify components in preprocessed visible light images to determine the spatial location and bounding box of each component in a photovoltaic power station. Based on thermal imaging data, dynamic temperature thresholds are set to identify areas of photovoltaic power plant strings with abnormal temperatures. The dynamic temperature threshold is taken as an example of the average temperature of the photovoltaic power plant string. For each photovoltaic power plant string, calculate the average temperature of all components in each photovoltaic power plant string; Based on the average temperature of all components in each photovoltaic power station string, the average temperature of all components in adjacent normal strings is calculated and compared. If the temperature of a certain photovoltaic power station string is significantly higher than that of other photovoltaic power station strings, then the photovoltaic power station string is determined to be faulty. A photovoltaic power station string is considered faulty if its temperature exceeds three times the standard deviation of the average temperature of all components in a neighboring normal string. Based on the determination of the photovoltaic power station string fault results, a color mapping relationship is established according to the degree of temperature anomaly, and the three-dimensional coordinates of the paired photovoltaic power station modules are associated to generate a heat map of the faulty photovoltaic power station string. The heat map of the faulty strings of the photovoltaic power station: blue indicates low temperature, green indicates medium temperature, and yellow to red indicates that the temperature gradually increases.

[0009] Preferably, the formula for the dynamic fault index of photovoltaic power plant strings is as follows:

[0010] in, Let be the dynamic fault index of the s-th photovoltaic power station string, where s is the s-th photovoltaic power station string. Let be the average temperature of the s-th photovoltaic power station string. Let be the average temperature of the neighboring normal strings of the s-th photovoltaic power station string. Let be the temperature standard deviation of the neighboring normal strings of the s-th photovoltaic power station string. To adjust the parameters, Let be the temperature gradient norm inside the s-th photovoltaic power station string.

[0011] Preferably, a global vertex set is constructed using the starting point of the maintenance personnel, the location of the faulty photovoltaic power station string, and the actual road obstacles as vertices; An edge is a path that connects any two nodes in the global set of vertices. A weighted directed graph model is constructed using actual distance, slope factor, and road factor as path weights; The actual road obstacles were obtained from the three-dimensional reality model of the photovoltaic power station. A set of random paths is generated using a genetic algorithm as the initial population; The fitness of each path is evaluated, with the shortest actual distance, the least impact of slope, and the best road conditions used as the fitness function. Through selection, crossover, and mutation operations, the population is continuously evolved, and the path selection is iteratively optimized to generate multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station. Based on generating multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station, and combining the path weights for comprehensive evaluation, the path with the shortest actual distance, the least impact of slope, and the best road conditions is selected as the optimal arrival path. Based on the optimal arrival path, the starting position of each maintenance personnel, the position of the faulty photovoltaic power station string and the actual road obstacles are obtained on the path, and coupled with the three-dimensional coordinates of the photovoltaic power station components to convert them into a three-dimensional coordinate waypoint sequence. Navigation information is transmitted to the terminal system in real time via wireless network.

[0012] Preferably, based on real-time transmission of navigation information to the terminal system, centimeter-level precise positioning of maintenance personnel is achieved using multi-source sensors; The multi-source sensors include: GPS, IMU inertial measurement unit, and vision sensor; Maintenance personnel wear terminal equipment equipped with 3D real-view navigation capabilities to receive 3D coordinate waypoint sequences and path planning information from the system. The terminal equipment uses spatial positioning technology to overlay the virtual navigation path with the actual photovoltaic power station scene, realizing penetrating visual guidance. Maintenance personnel can intuitively view the precise spatial location of the faulty string and the posture of the surrounding environment. During navigation, the terminal device continuously collects the real-time location data of the maintenance personnel and dynamically compares it with the preset path. When a deviation from the optimal path is detected, it immediately provides correction guidance through multimodal interaction. The multimodal interaction includes: 3D arrow guidance and spatial audio prompts; Upon reaching the location of the faulty cluster, the system automatically triggers the fault location confirmation process. Maintenance personnel scan the component's unique identifier using a terminal device to verify the fault information against the actual component status. After the maintenance operation is completed, the maintenance personnel enter the fault type and replacement part model information through the terminal interface, upload before and after comparison image data, and the system encrypts and sends the maintenance report and multimodal data back to the operation and maintenance management platform.

[0013] Preferably, after receiving the data, the platform automatically updates the real-time status database of photovoltaic power station components, including component operating status, maintenance history, and component replacement records, and intelligently triggers the correlation inspection process of adjacent strings to ensure that the scope of fault impact is fully investigated. Through a closed-loop operation and maintenance management mechanism, the system generates maintenance efficiency analysis reports, fault type statistical charts, and decision support data, providing a basis for optimizing preventive maintenance strategies for photovoltaic power plants and realizing full-process digital control from fault detection to maintenance and restoration.

[0014] Furthermore, a system for locating and handling string faults in large-scale photovoltaic power plants, and methods for locating and handling string faults in large-scale photovoltaic power plants, include: Data acquisition module, photovoltaic power station fault string heat map module, terminal system module and closed-loop management module; The data acquisition module is used to obtain the layout planning map of the photovoltaic power station, establish a three-dimensional real scene model of the photovoltaic power station, assign three-dimensional coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. The photovoltaic power station fault string heat map module is used to automatically identify photovoltaic power station fault strings based on inspections by drones equipped with visible light and thermal imaging cameras, associate and match the three-dimensional coordinates of photovoltaic power station components, and generate a photovoltaic power station fault string heat map. The terminal system module is electrically connected to the photovoltaic power station fault string heat map module and the data acquisition module. It is used to plan the optimal arrival path of the photovoltaic power station fault string based on the photovoltaic power station fault string heat map and the photovoltaic power station three-dimensional real scene model, and send navigation information to the terminal system in real time. The closed-loop management module is electrically connected to the terminal system module. It is used to send navigation information to the terminal system in real time. Maintenance personnel wear terminal devices to receive three-dimensional real-scene navigation, realize penetrating visual guidance, accurately locate the faulty string of the photovoltaic power station, and transmit the maintenance results back through the terminal after the maintenance is completed. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a system and method for fault location and handling in large-scale photovoltaic power plants. This solution integrates modeling and coordinate technology to solve the problem of the disconnect between physical and digital information; it improves fault location and navigation efficiency, reduces vehicle detours, and improves inspection efficiency; it accurately locates minor faults such as hot spots, avoiding blind searching; it is adaptable to large-scale power plants, and its location efficiency is not affected by the scale expansion; visual navigation lowers the operation and maintenance threshold and eliminates the need to rely on paper drawings. Attached Figure Description

[0016] Figure 1 Flowchart of methods for locating and handling string faults in large-scale photovoltaic power plants; Figure 2 A system framework diagram for fault location and handling in large-scale photovoltaic power plants. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, the method for locating and handling string faults in large-scale photovoltaic power plants includes: S1. Obtain the layout planning map of the photovoltaic power station, establish a three-dimensional real scene model of the photovoltaic power station, assign three-dimensional coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. Step S1 includes the following: Based on satellite imagery, obtain a layout planning map of photovoltaic power plants; Based on oblique photography by drones, POS data of different images of photovoltaic power stations are obtained, and an unordered image sequence of POS data is constructed. The POS data includes: location: longitude, latitude, altitude; attitude: pitch, roll, heading; The SIFT scale-invariant feature transform algorithm is used to perform multi-scale processing on the disordered image sequence of POS data and construct the scale space of the disordered image sequence of POS data. Based on the scale space of the disordered image sequence of POS data, continuous Gaussian blurring is performed to generate multi-scale images, and the local extrema of each pixel are detected. Using the second derivative matrix of Gaussian, we evaluate whether the local extrema of each pixel are stable. If they are unstable, we remove the unstable local extrema. Based on eliminating unstable local extrema, retaining the key local extrema of each pixel, and calculating the gradient direction and magnitude of each pixel; Based on the gradient direction and magnitude of each pixel, local features of each pixel are extracted; Calculate the gradient information of the neighborhood around the key local extrema of each pixel to generate a feature descriptor; The feature descriptor, taking a 16x16 region as an example, is divided into 16 sub-blocks, and the gradient histograms of each sub-block are calculated in 8 directions.

[0019] Step S1 also includes the following: Using the Euclidean distance formula, the similarity between two feature descriptors is calculated, and a reasonable threshold is set to determine whether the two feature descriptors match. The reasonable threshold is 0.8; The fundamental matrix of unknown camera intrinsic parameters is estimated using the random sample consensus algorithm. Based on the known camera intrinsic parameters from the POS data, the essential matrix is ​​calculated, and the relative rotation and translation matrix between each pair of images is calculated using the least squares method. Based on the relative rotation and translation matrix between each pair of images, triangulation is performed to calculate the three-dimensional spatial position of the matching feature points; Using the three-dimensional spatial position of the matching feature points as input, the MVS multi-view stereo algorithm is used for dense reconstruction, and registration and alignment are performed for images from different perspectives. For each pixel of the matching feature point, calculate the cost of matching in the corresponding view, construct the cost volume, optimize the cost volume using a dynamic programming algorithm, and generate a disparity map. Based on the disparity map, it is converted into a depth map and fused with point cloud to establish a 3D reality model of the photovoltaic power station.

[0020] Step S1 also includes the following: Based on a 3D real-scene model of a photovoltaic power station, semantic segmentation technology is used to classify and identify photovoltaic power station components. Based on the depth map, the center spatial coordinates of each photovoltaic power station module are calculated through three-dimensional geometric operations, and a unique identifier is assigned to each photovoltaic power station module. A photovoltaic power station component coordinate database is established based on the unique identifier assigned to each photovoltaic power station component. The photovoltaic power station component coordinate database includes: photovoltaic power station components, component coordinates, component type, model, inverter information, and the string to which they belong.

[0021] When using it, refer to the steps outlined above. Existing 3D modeling technologies for photovoltaic power plants mainly rely on traditional ground surveys and manual data input, which are inefficient and have limited accuracy. Although some technologies attempt to use satellite imagery and UAV oblique photography for layout planning and image data acquisition of photovoltaic power plants, these methods still suffer from cumbersome data processing, low accuracy, and inability to accurately obtain 3D coordinate information. Existing technologies cannot effectively handle large-scale data during multi-view reconstruction, resulting in significant errors in the reconstruction results and failing to meet the requirements for high-precision 3D modeling of photovoltaic power plants. This step combines satellite imagery and UAV oblique photography, utilizing SIFT feature point matching, continuous Gaussian blur processing, and multiple... Scale-based image analysis effectively improves the accuracy and robustness of image processing. Based on this, combined with the calculation of camera intrinsic parameters, rotation and translation matrices, and optimization of the multi-view stereo (MVS) algorithm, more precise 3D spatial positioning and modeling can be achieved. Through semantic segmentation technology and depth map computation, photovoltaic power station components can be automatically classified and identified, and a unique identifier can be assigned to each component, constructing a detailed component coordinate database. The beneficial effects of this method are that it improves the accuracy and automation level of 3D real-scene model construction for photovoltaic power stations, significantly enhances data acquisition and processing efficiency, reduces manual intervention, and improves the management and maintenance efficiency of photovoltaic power station operation.

[0022] S2. Based on the inspection by the UAV equipped with visible light and thermal imaging cameras, automatically identify the faulty strings of the photovoltaic power station, associate and match the three-dimensional coordinates of the photovoltaic power station components, and generate a heat map of the faulty strings of the photovoltaic power station. Step S2 includes the following: Based on the inspection of the photovoltaic power station by the drone equipped with visible light and thermal imaging cameras, the visible light and thermal imaging data are obtained and the data is preprocessed. Image recognition algorithms are used to identify components in preprocessed visible light images to determine the spatial location and bounding box of each component in a photovoltaic power station. Based on thermal imaging data, dynamic temperature thresholds are set to identify areas of photovoltaic power plant strings with abnormal temperatures. The dynamic temperature threshold is taken as an example of the average temperature of the photovoltaic power plant string. For each photovoltaic power plant string, calculate the average temperature of all components in each photovoltaic power plant string; Based on the average temperature of all modules in each photovoltaic power station string, the average temperature of all modules in adjacent normal strings is calculated and compared. If the temperature of a certain photovoltaic power station string is significantly higher than that of other photovoltaic power station strings, then that photovoltaic power station string is determined to be faulty. The formula is as follows:

[0023] in, Let be the dynamic fault index of the s-th photovoltaic power station string, where s is the s-th photovoltaic power station string. Let be the average temperature of the s-th photovoltaic power station string. Let be the average temperature of the neighboring normal strings of the s-th photovoltaic power station string. Let be the temperature standard deviation of the neighboring normal strings of the s-th photovoltaic power station string. To adjust the parameters, Let be the temperature gradient norm inside the s-th photovoltaic power station string; A photovoltaic power station string is considered faulty if its temperature exceeds three times the standard deviation of the average temperature of all components in a neighboring normal string. Based on the determination of the photovoltaic power station string fault results, a color mapping relationship is established according to the degree of temperature anomaly, and the three-dimensional coordinates of the paired photovoltaic power station modules are associated to generate a heat map of the faulty photovoltaic power station string. The heat map of the faulty strings of the photovoltaic power station: blue indicates low temperature, green indicates medium temperature, and yellow to red indicates that the temperature gradually increases.

[0024] When using it, refer to the steps outlined above. Existing photovoltaic power plant fault detection technologies mostly rely on manual inspections or periodic checks, which are inefficient and prone to missing some faults. This is especially true for large-scale photovoltaic power plants, where traditional methods struggle to meet the demands for efficient and comprehensive detection. Current technologies primarily focus on visible light image processing, neglecting in-depth analysis of thermal imaging data. Existing fault detection algorithms often fail to pinpoint individual components when identifying temperature anomalies and locating faulty components, and their judgment criteria are somewhat vague. This step, by combining data collected from visible light and thermal imaging cameras and utilizing image recognition and temperature anomaly detection algorithms, can accurately identify and locate faulty components and determine faults using dynamic temperature thresholds, significantly improving the accuracy and efficiency of fault detection. Through 3D coordinate correlation technology combined with heat map generation, the location and temperature distribution of faulty strings are visually displayed. This method not only improves the accuracy of photovoltaic power plant fault diagnosis but also enables the completion of large-scale photovoltaic power plant inspections in a shorter time, effectively reducing manual intervention and improving the operation and maintenance efficiency and management level of the power plant.

[0025] S3. Based on the heat map of the faulty strings of the photovoltaic power station and the three-dimensional real scene model of the photovoltaic power station, plan the optimal arrival path of the faulty strings of the photovoltaic power station and send navigation information to the terminal system in real time. Step S3 includes the following: A global vertex set is constructed using the starting point of the maintenance personnel, the location of the faulty photovoltaic power station string, and the actual road obstacles as vertices; An edge is a path that connects any two nodes in the global set of vertices. A weighted directed graph model is constructed using actual distance, slope factor, and road factor as path weights; The actual road obstacles were obtained from the three-dimensional reality model of the photovoltaic power station. A set of random paths is generated using a genetic algorithm as the initial population; The fitness of each path is evaluated, with the shortest actual distance, the least impact of slope, and the best road conditions used as the fitness function. Through selection, crossover, and mutation operations, the population is continuously evolved, and the path selection is iteratively optimized to generate multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station. Based on generating multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station, and combining the path weights for comprehensive evaluation, the path with the shortest actual distance, the least impact of slope, and the best road conditions is selected as the optimal arrival path. Based on the optimal arrival path, the starting position of each maintenance personnel, the position of the faulty photovoltaic power station string and the actual road obstacles are obtained on the path, and coupled with the three-dimensional coordinates of the photovoltaic power station components to convert them into a three-dimensional coordinate waypoint sequence. Navigation information is transmitted to the terminal system in real time via wireless network.

[0026] When using it, refer to the steps outlined above. Currently, fault repair route planning for photovoltaic (PV) power plants largely relies on manual planning or traditional map navigation systems, resulting in low route selection efficiency and a lack of dynamic adjustment capabilities. This is especially true in large-scale PV power plants, where maintenance personnel may face complex terrain obstacles and construction difficulties, making it impossible for traditional route planning methods to efficiently and accurately plan the optimal repair route. Existing route planning technologies typically fail to fully utilize the 3D reality model of the PV power plant, lacking detailed consideration of factors such as road slope and traffic conditions, and failing to achieve real-time route optimization and dynamic navigation. This step, by combining the heat map of the faulty string in the PV power plant with the 3D reality model, uses a genetic algorithm to optimize the optimal route in real time. It can automatically plan the optimal route from the maintenance personnel's starting point to the faulty string, considering multiple factors such as actual road obstacles, slope factors, and road traffic conditions, thereby effectively improving the efficiency and accuracy of route selection. The navigation information generated based on the 3D coordinates and heat map can be transmitted to the terminal system in real time, ensuring that maintenance personnel reach the fault area in the shortest possible time, reducing manual intervention and scheduling costs, and improving the operation and maintenance efficiency and emergency response capabilities of the PV power plant.

[0027] S4. Based on real-time navigation information sent to the terminal system, maintenance personnel wear terminal devices to receive three-dimensional real-scene navigation, realize penetrating visual guidance, accurately locate the faulty string of the photovoltaic power station, and transmit the maintenance results back through the terminal after the maintenance is completed. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance. Step S4 includes the following: Based on real-time navigation information sent to the terminal system, and using multi-source sensors, the maintenance personnel are located with centimeter-level precision. The multi-source sensors include: GPS, IMU inertial measurement unit, and vision sensor; Maintenance personnel wear terminal equipment equipped with 3D real-view navigation capabilities to receive 3D coordinate waypoint sequences and path planning information from the system. The terminal equipment uses spatial positioning technology to overlay the virtual navigation path with the actual photovoltaic power station scene, realizing penetrating visual guidance. Maintenance personnel can intuitively view the precise spatial location of the faulty string and the posture of the surrounding environment. During navigation, the terminal device continuously collects the real-time location data of the maintenance personnel and dynamically compares it with the preset path. When a deviation from the optimal path is detected, it immediately provides correction guidance through multimodal interaction. The multimodal interaction includes: 3D arrow guidance and spatial audio prompts; Upon reaching the location of the faulty cluster, the system automatically triggers the fault location confirmation process. Maintenance personnel scan the component's unique identifier using a terminal device to verify the fault information against the actual component status. After the maintenance operation is completed, the maintenance personnel enter the fault type and replacement part model information through the terminal interface, upload before and after comparison image data, and the system encrypts and sends the maintenance report and multimodal data back to the operation and maintenance management platform.

[0028] Step S4 also includes the following: After receiving the data, the platform automatically updates the real-time status database of photovoltaic power station components, including component operating status, maintenance history, and component replacement records. It also intelligently triggers the associated inspection process of adjacent strings to ensure that the scope of fault impact is fully investigated. Through a closed-loop operation and maintenance management mechanism, the system generates maintenance efficiency analysis reports, fault type statistical charts, and decision support data, providing a basis for optimizing preventive maintenance strategies for photovoltaic power plants and realizing full-process digital control from fault detection to maintenance and restoration.

[0029] When using it, refer to the steps outlined above. Currently, the maintenance of photovoltaic power plants largely relies on manual labor or traditional equipment to guide maintenance personnel, resulting in problems such as low positioning accuracy, inaccurate path planning, and low maintenance efficiency. Traditional navigation systems often fail to fully integrate with the site environment, lacking high-precision real-time positioning and dynamic path adjustment capabilities, causing maintenance personnel to frequently deviate from the optimal path. Existing systems typically cannot automatically update the real-time status of the power plant and lack an effective closed-loop management mechanism, leading to poor efficiency and data traceability in the maintenance process. This step, through multi-source sensors combined with 3D real-scene navigation technology, provides centimeter-level precise positioning and dynamic path adjustment, ensuring that maintenance personnel can accurately and quickly reach the location of the faulty string. Through penetrating visual guidance and multimodal interaction, maintenance personnel can clearly view the spatial location of the faulty component and its surrounding environment, improving the accuracy and efficiency of operations. The system automatically updates the status of photovoltaic power plant components by transmitting maintenance results in real time and triggers related checks, ensuring closed-loop management of photovoltaic power plant operation and maintenance. Comprehensive recording and analysis of operation and maintenance data can support the optimization of preventive maintenance strategies, greatly improving the management level and fault response capabilities of photovoltaic power plants.

[0030] Reference Figure 2 As shown, the system for locating and handling string faults in a large-scale photovoltaic power plant includes: Data acquisition module, photovoltaic power station fault string heat map module, terminal system module and closed-loop management module; The data acquisition module is used to obtain the layout planning map of the photovoltaic power station, establish a three-dimensional real scene model of the photovoltaic power station, assign three-dimensional coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. The photovoltaic power station fault string heat map module is used to automatically identify photovoltaic power station fault strings based on inspections by drones equipped with visible light and thermal imaging cameras, associate and match the three-dimensional coordinates of photovoltaic power station components, and generate a photovoltaic power station fault string heat map. The terminal system module is electrically connected to the photovoltaic power station fault string heat map module and the data acquisition module. It is used to plan the optimal arrival path of the photovoltaic power station fault string based on the photovoltaic power station fault string heat map and the photovoltaic power station three-dimensional real scene model, and send navigation information to the terminal system in real time. The closed-loop management module is electrically connected to the terminal system module. It is used to send navigation information to the terminal system in real time. Maintenance personnel wear terminal devices to receive three-dimensional real-scene navigation, realize penetrating visual guidance, accurately locate the faulty string of the photovoltaic power station, and transmit the maintenance results back through the terminal after the maintenance is completed. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for locating and handling string faults in large-scale photovoltaic power plants, characterized in that, include: S1. Obtain the layout planning map of the photovoltaic power station, establish a three-dimensional real scene model of the photovoltaic power station, assign three-dimensional coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. S2. Based on the inspection by the UAV equipped with visible light and thermal imaging cameras, automatically identify the faulty strings of the photovoltaic power station, associate and match the three-dimensional coordinates of the photovoltaic power station components, and generate a heat map of the faulty strings of the photovoltaic power station. S3. Based on the heat map of the faulty strings of the photovoltaic power station and the three-dimensional real scene model of the photovoltaic power station, plan the optimal arrival path of the faulty strings of the photovoltaic power station and send navigation information to the terminal system in real time. S4. Based on real-time navigation information sent to the terminal system, maintenance personnel wear terminal devices to receive 3D real-scene navigation, achieving penetrating visual guidance, accurately locating faulty strings in the photovoltaic power station, and transmitting maintenance results back through the terminal after completing the maintenance. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance.

2. The method for locating and handling string faults in large-scale photovoltaic power plants according to claim 1, characterized in that, S1 includes: Based on satellite imagery, obtain a layout planning map of photovoltaic power plants; Based on oblique photography by drones, POS data of different images of photovoltaic power stations are obtained, and an unordered image sequence of POS data is constructed. The POS data includes: location: longitude, latitude, altitude; attitude: pitch, roll, heading; The SIFT scale-invariant feature transform algorithm is used to perform multi-scale processing on the disordered image sequence of POS data and construct the scale space of the disordered image sequence of POS data. Based on the scale space of the disordered image sequence of POS data, continuous Gaussian blurring is performed to generate multi-scale images, and the local extrema of each pixel are detected. Using the second derivative matrix of Gaussian, we evaluate whether the local extrema of each pixel are stable. If they are unstable, we remove the unstable local extrema. Based on eliminating unstable local extrema, retaining the key local extrema of each pixel, and calculating the gradient direction and magnitude of each pixel; Based on the gradient direction and magnitude of each pixel, local features of each pixel are extracted; Calculate the gradient information of the neighborhood around the key local extrema of each pixel to generate a feature descriptor; The feature descriptor, taking a 16x16 region as an example, is divided into 16 sub-blocks, and the gradient histograms of each sub-block are calculated in 8 directions.

3. The method for locating and handling string faults in large-scale photovoltaic power plants according to claim 2, characterized in that, S1 further includes: Using the Euclidean distance formula, the similarity between two feature descriptors is calculated, and a reasonable threshold is set to determine whether the two feature descriptors match. The reasonable threshold is 0.8; The fundamental matrix of unknown camera intrinsic parameters is estimated using the random sample consensus algorithm. Based on the known camera intrinsic parameters from the POS data, the essential matrix is ​​calculated, and the relative rotation and translation matrix between each pair of images is calculated using the least squares method. Based on the relative rotation and translation matrix between each pair of images, triangulation is performed to calculate the three-dimensional spatial position of the matching feature points; Using the three-dimensional spatial position of the matching feature points as input, the MVS multi-view stereo algorithm is used for dense reconstruction, and registration and alignment are performed for images from different perspectives. For each pixel of the matching feature point, calculate the cost of matching in the corresponding view, construct the cost volume, optimize the cost volume using a dynamic programming algorithm, and generate a disparity map. Based on the disparity map, it is converted into a depth map and fused with point cloud to establish a 3D reality model of the photovoltaic power station.

4. The method for locating and handling string faults in large-scale photovoltaic power plants according to claim 3, characterized in that, S1 further includes: Based on a 3D real-scene model of a photovoltaic power station, semantic segmentation technology is used to classify and identify photovoltaic power station components. Based on the depth map, the center spatial coordinates of each photovoltaic power station module are calculated through three-dimensional geometric operations, and a unique identifier is assigned to each photovoltaic power station module. A photovoltaic power station component coordinate database is established based on the unique identifier assigned to each photovoltaic power station component. The photovoltaic power station component coordinate database includes: photovoltaic power station components, component coordinates, component type, model, inverter information, and the string to which they belong.

5. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 1, characterized in that, S2 includes: Based on the inspection of the photovoltaic power station by the drone equipped with visible light and thermal imaging cameras, the visible light and thermal imaging data are obtained and the data is preprocessed. Image recognition algorithms are used to identify components in preprocessed visible light images to determine the spatial location and bounding box of each component in a photovoltaic power station. Based on thermal imaging data, dynamic temperature thresholds are set to identify areas of photovoltaic power plant strings with abnormal temperatures. The dynamic temperature threshold is taken as an example of the average temperature of the photovoltaic power plant string. For each photovoltaic power plant string, calculate the average temperature of all components in each photovoltaic power plant string; Based on the average temperature of all components in each photovoltaic power station string, the average temperature of all components in adjacent normal strings is calculated and compared. If the temperature of a certain photovoltaic power station string is significantly higher than that of other photovoltaic power station strings, then the photovoltaic power station string is determined to be faulty. A photovoltaic power station string is considered faulty if its temperature exceeds three times the standard deviation of the average temperature of all components in a neighboring normal string. Based on the determination of the photovoltaic power station string fault results, a color mapping relationship is established according to the degree of temperature anomaly, and the three-dimensional coordinates of the paired photovoltaic power station modules are associated to generate a heat map of the faulty photovoltaic power station string. The heat map of the faulty strings of the photovoltaic power station: blue indicates low temperature, green indicates medium temperature, and yellow to red indicates that the temperature gradually increases.

6. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 5, characterized in that, The formula for the dynamic fault index of the photovoltaic power station string is as follows: in, Let be the dynamic fault index of the s-th photovoltaic power station string, where s is the s-th photovoltaic power station string. Let be the average temperature of the s-th photovoltaic power station string. Let be the average temperature of the neighboring normal strings of the s-th photovoltaic power station string. Let be the temperature standard deviation of the neighboring normal strings of the s-th photovoltaic power station string. To adjust the parameters, Let be the temperature gradient norm inside the s-th photovoltaic power station string.

7. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 5, characterized in that, S3 includes: A global vertex set is constructed using the starting point of the maintenance personnel, the location of the faulty photovoltaic power station string, and the actual road obstacles as vertices; An edge is a path that connects any two nodes in the global set of vertices. A weighted directed graph model is constructed using actual distance, slope factor, and road factor as path weights; The actual road obstacles were obtained from the three-dimensional reality model of the photovoltaic power station. A set of random paths is generated using a genetic algorithm as the initial population; The fitness of each path is evaluated, with the shortest actual distance, the least impact of slope, and the best road conditions used as the fitness function. Through selection, crossover, and mutation operations, the population is continuously evolved, and the path selection is iteratively optimized to generate multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station. Based on generating multiple feasible paths from the starting point of the maintenance personnel to the location of the faulty string in the photovoltaic power station, and combining the path weights for comprehensive evaluation, the path with the shortest actual distance, the least impact of slope, and the best road conditions is selected as the optimal arrival path. Based on the optimal arrival path, the starting position of each maintenance personnel, the position of the faulty photovoltaic power station string and the actual road obstacles are obtained on the path, and coupled with the three-dimensional coordinates of the photovoltaic power station components to convert them into a three-dimensional coordinate waypoint sequence. Navigation information is transmitted to the terminal system in real time via wireless network.

8. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 7, characterized in that, S4 includes: Based on real-time navigation information sent to the terminal system, and using multi-source sensors, the maintenance personnel are located with centimeter-level precision. The multi-source sensors include: GPS, IMU inertial measurement unit, and vision sensor; Maintenance personnel wear terminal equipment equipped with 3D real-view navigation capabilities to receive 3D coordinate waypoint sequences and path planning information from the system. The terminal equipment uses spatial positioning technology to overlay the virtual navigation path with the actual photovoltaic power station scene, realizing penetrating visual guidance. Maintenance personnel can intuitively view the precise spatial location of the faulty string and the posture of the surrounding environment. During navigation, the terminal device continuously collects the real-time location data of the maintenance personnel and dynamically compares it with the preset path. When a deviation from the optimal path is detected, it immediately provides correction guidance through multimodal interaction. The multimodal interaction includes: 3D arrow guidance and spatial audio prompts; Upon reaching the location of the faulty cluster, the system automatically triggers the fault location confirmation process. Maintenance personnel scan the component's unique identifier using a terminal device to verify the fault information against the actual component status. After the maintenance operation is completed, the maintenance personnel enter the fault type and replacement part model information through the terminal interface, upload before and after comparison image data, and the system encrypts and sends the maintenance report and multimodal data back to the operation and maintenance management platform.

9. The method for locating and handling string faults in a large-scale photovoltaic power plant according to claim 8, characterized in that, S4 further includes: After receiving the data, the platform automatically updates the real-time status database of photovoltaic power station components, including component operating status, maintenance history, and component replacement records. It also intelligently triggers the associated inspection process of adjacent strings to ensure that the scope of fault impact is fully investigated. Through a closed-loop operation and maintenance management mechanism, the system generates maintenance efficiency analysis reports, fault type statistical charts, and decision support data, providing a basis for optimizing preventive maintenance strategies for photovoltaic power plants and realizing full-process digital control from fault detection to maintenance and restoration.

10. A system for locating and handling string faults in large-scale photovoltaic power plants, characterized in that, The method for locating and handling string faults in a large-scale photovoltaic power plant, as described in any one of claims 1-9, includes: Data acquisition module, photovoltaic power station fault string heat map module, terminal system module and closed-loop management module; The data acquisition module is used to obtain the layout planning map of the photovoltaic power station, establish a three-dimensional real scene model of the photovoltaic power station, assign three-dimensional coordinates to each photovoltaic power station component, and establish a photovoltaic power station component coordinate database. The photovoltaic power station fault string heat map module is used to automatically identify photovoltaic power station fault strings based on inspections by drones equipped with visible light and thermal imaging cameras, associate and match the three-dimensional coordinates of photovoltaic power station components, and generate a photovoltaic power station fault string heat map. The terminal system module is electrically connected to the photovoltaic power station fault string heat map module and the data acquisition module. It is used to plan the optimal arrival path of the photovoltaic power station fault string based on the photovoltaic power station fault string heat map and the photovoltaic power station three-dimensional real scene model, and send navigation information to the terminal system in real time. The closed-loop management module is electrically connected to the terminal system module. It is used to send navigation information to the terminal system in real time. Maintenance personnel wear terminal devices to receive three-dimensional real-scene navigation, realize penetrating visual guidance, accurately locate the faulty string of the photovoltaic power station, and transmit the maintenance results back through the terminal after the maintenance is completed. The system automatically updates the real-time status of the photovoltaic power station components, realizing closed-loop management of photovoltaic power station operation and maintenance.

Images