Path planning platform of inspection robot for operation and maintenance of photovoltaic power station

Abstract

The invention discloses an inspection robot path planning platform for operation and maintenance of a photovoltaic power station, and relates to the related field of path planning control, and the platform comprises a global inspection path generation module which is used for generating a power station map according to the layout of photovoltaic panels, carrying out the coverage traversal of the photovoltaic panels, and generating a global inspection path; the photovoltaic panel operation dynamic analysis module is used for performing operation dynamic analysis and generating a plurality of identification points; the multi-path tracking module is used for obtaining a plurality of local inspection points, performing path tracking and generating a plurality of local inspection paths; the local inspection route dynamic updating module is used for performing path optimization, generating a first local inspection route, performing updating and determining an Nth local inspection route; and the global inspection optimization path generation module is used for performing feedback adjustment on the global inspection path. The technical problem of insufficient inspection efficiency and accuracy of existing inspection robot path planning for operation and maintenance of the photovoltaic power station is solved, and the technical effect of improving the inspection efficiency and accuracy is achieved.

Description

Technical Field

[0001] The present application relates to the field of path planning and control, and in particular to a path planning platform for inspection robots used in photovoltaic power station operation and maintenance. Background Art

[0002] Efficient and stable operation and maintenance of photovoltaic power stations are of great significance for ensuring power supply and improving energy efficiency. However, photovoltaic power stations usually occupy a large area and have a large number of photovoltaic panels. The traditional manual inspection method has problems such as low efficiency, high cost, and easy to miss inspections, which makes it difficult to meet the operation and maintenance needs of large-scale photovoltaic power stations. In order to solve the above problems, the industry has begun to explore the use of robots for automated inspections of photovoltaic power stations. The existing inspection robot path planning methods mainly rely on preset fixed routes or simple random traversal strategies. However, when the layout of the photovoltaic power station changes (such as the addition of new photovoltaic panels, damage to photovoltaic panels, etc.), the preset fixed routes may not cover all photovoltaic panels, resulting in incomplete inspections. The random traversal strategy lacks specificity, which may lead to a lot of time wasted in areas where photovoltaic panels are in good condition, while insufficient inspections are conducted in areas that need to be focused on (such as photovoltaic panels with frequent failures).

[0003] At the current stage, due to the lack of intelligent decision-making capabilities, relevant technologies cannot be dynamically adjusted according to the real-time operating status of photovoltaic panels, resulting in technical problems such as insufficient inspection efficiency and accuracy in the path planning of inspection robots used for photovoltaic power station operation and maintenance. Summary of the invention

[0004] The present application provides a patrol robot path planning platform for photovoltaic power station operation and maintenance, generates a power station map according to the layout of photovoltaic panels of the photovoltaic power station, covers and traverses multiple photovoltaic panels based on the power station map, generates a global patrol route, performs dynamic operation analysis on multiple photovoltaic panels according to the global patrol route, generates multiple identification points according to the analysis results, synchronizes the multiple identification points to the power station map for patrol analysis, obtains multiple local patrol points, and traces back to the identification points based on these points for multi-path tracking, generates multiple local patrol routes, performs path optimization based on multiple local patrol routes, generates a first local patrol route, and dynamically updates it according to the results of dynamic operation analysis, determines the Nth local patrol route, feedback adjusts the global patrol route based on the Nth local patrol route, generates a global patrol optimization path, and other technical means, which can monitor the operation status of photovoltaic panels in real time, dynamically adjust the patrol route according to the analysis results, ensure that key areas are focused on, and avoid wasting time and resources in areas with good conditions, thereby realizing intelligent path planning of the patrol robot of the photovoltaic power station and achieving the technical effect of improving the efficiency and accuracy of the patrol.

[0005] This application provides a patrol robot path planning platform for photovoltaic power station operation and maintenance, including: A global inspection route generation module, the global inspection route generation module is used to generate a power station map according to the photovoltaic panel layout of the photovoltaic power station, and to cover and traverse multiple photovoltaic panels based on the power station map to generate a global inspection route; a photovoltaic panel operation dynamic analysis module, the photovoltaic panel operation dynamic analysis module is used to perform operation dynamic analysis on multiple photovoltaic panels according to the global inspection route, and generate multiple identification points according to the operation dynamic analysis results; a multi-path tracking module, the multi-path tracking module is used to synchronize the multiple identification points to the power station map for inspection analysis, obtain multiple local inspection points, and generate multiple local inspection points based on the multiple inspection points. A local inspection point is traced back to the multiple identification points for multi-path tracking to generate multiple local inspection routes; a local inspection route dynamic update module, the local inspection route dynamic update module is used to perform path optimization based on the multiple local inspection routes to generate a first local inspection route, dynamically update the first local inspection route according to the operation dynamic analysis result, and determine the Nth local inspection route; a global inspection optimization path generation module, the global inspection optimization path generation module is used to feedback adjust the global patrol route based on the Nth local patrol route to generate a global inspection optimization path.

[0006] In a possible implementation, a dynamic operation analysis is performed on a plurality of photovoltaic panels according to the global inspection route, a plurality of identification points are generated according to the dynamic operation analysis results, and the following processing is performed: Based on multiple photovoltaic panels combined with the power station map, coordinate position analysis is performed to determine multiple coordinate points; an inspection cycle is formulated, and operation monitoring is performed by traversing the multiple coordinate points through the global inspection route according to the inspection cycle to obtain multiple operation data; based on the multiple operation data combined with environmental parameters, abnormality diagnosis is performed to determine multiple abnormality types, and dynamic analysis is performed based on the multiple abnormality types to generate the operation dynamic analysis results, which include multiple abnormal dynamic tags; the multiple abnormal dynamic tags are mapped to the multiple coordinate points for identification to generate the multiple identification points.

[0007] In a possible implementation, the multiple identification points are synchronized to the power station map for inspection analysis to obtain multiple local inspection points, and the following processing is performed: The multiple identification points are associated according to the multiple coordinate points to obtain multiple positioning association coefficients; the multiple identification points are synchronized to the power station map for inspection classification according to the multiple positioning association coefficients to obtain multiple categories to be inspected; the inspection time period is defined according to the multiple categories to be inspected, and the multiple identification points are inspected and managed according to the inspection time period to generate an inspection allocation management task; the inspection allocation management task is synchronized to the central monitoring unit for inspection recording to obtain multiple inspection monitoring record data, and the multiple inspection monitoring record data are added to the multiple local inspection points, and the multiple inspection monitoring record data have a corresponding relationship with the multiple local inspection points.

[0008] In a possible implementation, based on the multiple local inspection points, tracing back to the multiple identification points to perform multi-path tracking, multiple local inspection routes are generated, and the following processing is performed: According to the global inspection route, multiple photovoltaic panels are traversed to determine whether the multiple identification points have completed the traversal visit; when the multiple identification points have completed the traversal visit, the multiple local inspection points are edge monitored and the inspection end point is set; the inspection starting point is extracted, and the multiple identification points are reversely traced back according to the inspection end point until they are traced back to the inspection starting point, thereby generating multiple backtracking paths; path tracking analysis is performed based on the multiple backtracking paths to generate multiple path complexities, wherein the multiple path complexities include multiple time complexities and multiple space complexities; the multiple backtracking paths are weighted based on the multiple time complexities combined with the multiple space complexities to determine the multiple local inspection routes.

[0009] In a possible implementation, path optimization is performed based on the multiple local inspection routes to generate a first local inspection route, and the following processing is performed: Based on the multiple time complexities combined with the multiple weight coefficients, the multiple local inspection routes are screened to determine a time-optimized local inspection route; based on the multiple space complexities combined with the multiple weight coefficients, the multiple local inspection routes are screened to determine a space-optimized local inspection route; the time-optimized local inspection route is cross-combined with the space-optimized local inspection route to generate the first local inspection route.

[0010] In a possible implementation, the time-optimized local inspection route and the space-optimized local inspection route are cross-combined to generate the first local inspection route, and the following processing is performed: A time optimization target is defined based on the time-optimized local inspection route, and a space optimization target is defined based on the space-optimized local inspection route; the multiple local inspection points are calculated according to the time optimization target to construct a time matrix; local inspection areas are set based on the multiple local inspection points, and the local inspection areas are calculated according to the space optimization target to construct an inspection path network; the time matrix is ​​cross-matched with the inspection path network to generate multiple time-space inspection combinations, and path reachability verification is performed on the multiple time-space inspection combinations, and the first local inspection route is obtained according to the verification results.

[0011] In a possible implementation, the first local inspection route is dynamically updated according to the operation dynamic analysis result, the Nth local inspection route is determined, and the following processing is performed: Perform regular reinforcement learning based on the operation dynamic analysis results and construct a multi-period path learning log; S1: extract the target period path learning log based on the multi-period path learning log to perform matching evaluation on the first local inspection route and generate a path matching score; S2: determine whether the path matching score is greater than or equal to the expected score threshold. If the path matching score is less than the expected score threshold, generate an update instruction, and update the first local inspection route through the update instruction to generate a second local inspection route; S3: execute the second local inspection route based on the current inspection cycle. When in the next inspection cycle, repeat S1 and S2 for dynamic update to determine the Nth local inspection route.

[0012] In a possible implementation, regular reinforcement learning is performed according to the running dynamic analysis results, a multi-period path learning log is constructed, and the following processing is performed: The photovoltaic panel layout information of the photovoltaic power station is extracted based on the power station map, and the state space is defined according to the photovoltaic panel layout information; the reachability analysis is performed based on the global inspection route, and the action space is defined according to the reachability analysis result; the reward function is introduced, and the operation dynamic analysis result is mapped to the state space and the reward function is combined for reinforcement learning to generate an operation state learning record; the operation dynamic analysis result is mapped to the action space and the reward function is combined for reinforcement learning to generate an operation action learning record; the operation state learning record and the operation action learning record are associated and integrated according to the inspection cycle to construct a multi-period path learning log.

[0013] It is intended to use the inspection robot path planning platform for photovoltaic power station operation and maintenance proposed in this application, generate a power station map according to the photovoltaic panel layout of the photovoltaic power station through the global inspection route generation module, cover and traverse multiple photovoltaic panels based on the power station map to generate a global inspection route, perform operation dynamic analysis of multiple photovoltaic panels according to the global inspection route through the photovoltaic panel operation dynamic analysis module, generate multiple identification points according to the operation dynamic analysis results, synchronize the multiple identification points to the power station map through the multi-path tracking module for inspection analysis, obtain multiple local inspection points, trace back to multiple identification points based on the multiple local inspection points for multi-path tracking, generate multiple local inspection routes, perform path optimization based on multiple local inspection routes through the local inspection route dynamic update module, generate a first local inspection route, dynamically update the first local inspection route according to the operation dynamic analysis results, determine the Nth local inspection route, and feedback adjust the global inspection route based on the Nth local inspection route through the global inspection optimization path generation module to generate a global inspection optimization path, thereby achieving the technical effect of improving the efficiency and accuracy of inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention are briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the platform according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic diagram of the structure of a patrol robot path planning platform for photovoltaic power station operation and maintenance provided in an embodiment of the present application.

[0016] Figure 2 A schematic diagram of a process for generating multiple identification points in a patrol robot path planning platform for photovoltaic power station operation and maintenance provided in an embodiment of the present application.

[0017] Explanation of the reference numerals: global inspection route generation module 10 , photovoltaic panel operation dynamic analysis module 20 , multi-path tracking module 30 , local inspection route dynamic update module 40 , global inspection optimized path generation module 50 . DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, platform, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0021] The present application embodiment provides a patrol robot path planning platform for photovoltaic power station operation and maintenance, such as Figure 1 As shown, the platform includes: A global inspection route generation module 10 is used to generate a power station map according to the layout of photovoltaic panels in the photovoltaic power station, and to cover and traverse multiple photovoltaic panels based on the power station map to generate a global inspection route. Specifically, the photovoltaic panel layout information of the photovoltaic power station is collected, including the number, location, arrangement, etc. of the photovoltaic panels. Based on the collected photovoltaic panel layout information, a power station map (a diagram or model representing the layout and structure of the photovoltaic power station, used to guide the movement of the inspection robot) is generated. This map can be two-dimensional or three-dimensional, depending on the complexity of the power station. Next, a breadth-first search algorithm (or depth-first search, A* algorithm) is used to cover and traverse the photovoltaic panels on the power station map to ensure that each photovoltaic panel is visited. Finally, based on the results of the coverage traversal, a global inspection route is generated.

[0022] Photovoltaic panel operation dynamic analysis module 20, the photovoltaic panel operation dynamic analysis module 20 is used to perform operation dynamic analysis on multiple photovoltaic panels according to the global inspection route, and generate multiple identification points according to the operation dynamic analysis results. Specifically, the inspection robot visits each photovoltaic panel according to the global inspection route and collects its operation data, such as voltage, current, power, etc. Based on the collected operation data, abnormal diagnosis is performed to determine whether there is an abnormal type (such as overheating, overcurrent, undervoltage, etc.). Based on the operation dynamic analysis results, multiple abnormal points (identification points) are generated, which indicate which photovoltaic panels have abnormalities, as well as the type and degree of the abnormality.

[0023] like Figure 2 As shown, in a possible implementation, a dynamic operation analysis is performed on multiple photovoltaic panels according to the global inspection route, and multiple identification points are generated according to the results of the dynamic operation analysis, further including: performing coordinate position analysis based on multiple photovoltaic panels in combination with the power station map to determine multiple coordinate points; formulating an inspection cycle, traversing the multiple coordinate points through the global inspection route according to the inspection cycle to perform operation monitoring, and obtaining multiple operation data; performing abnormality diagnosis based on the multiple operation data in combination with environmental parameters to determine multiple abnormality types, performing dynamic analysis based on the multiple abnormality types, and generating the dynamic operation analysis results, which include multiple abnormal dynamic labels; mapping the multiple abnormal dynamic labels to the multiple coordinate points for identification to generate the multiple identification points.

[0024] Specifically, the photovoltaic panel layout information and power station map of the photovoltaic power station are received, which include the physical location, size, arrangement of the photovoltaic panels, etc. The precise coordinate points of each photovoltaic panel are determined on the power station map based on the layout information of the photovoltaic panels using a geographic information system (GIS) or similar positioning technology. These coordinate points are the basis for subsequent inspections and analysis. Based on the operating history of the photovoltaic power station, environmental conditions (such as irradiance, temperature, humidity, etc.), the type of photovoltaic panels, and maintenance requirements, an inspection cycle is formulated to ensure that the photovoltaic panels operate in the best condition while avoiding unnecessary frequent inspections.

[0025] When the set inspection cycle arrives, the inspection robot traverses the previously determined multiple coordinate points according to the global inspection route. At each coordinate point, the inspection robot performs operation monitoring and collects the operation data of the photovoltaic panel such as voltage, current, power, and environmental parameters (such as temperature, humidity, etc.). The collected operation data and environmental parameters are analyzed to identify any possible abnormalities. Based on the analysis results, multiple abnormality types are determined, such as overheating, undervoltage, overcurrent, etc. The abnormality types are further analyzed to predict their development trends and impacts on the operation of the photovoltaic power station. Based on the above analysis, an operation dynamic analysis result containing multiple abnormal dynamic tags is generated. These tags provide detailed information about the abnormality type, location, severity, and potential impact. The multiple abnormal dynamic tags in the operation dynamic analysis result are mapped to the previously determined multiple coordinate points, and these coordinate points with abnormal dynamic tags are regarded as identification points to guide subsequent local inspections and path planning. This implementation adopts an operation dynamic analysis method based on the global inspection route, which comprehensively monitors the operation status of the photovoltaic power station and can timely discover and locate potential abnormal points, providing strong support for subsequent local inspections and route planning, improving the efficiency and accuracy of inspections, and ensuring the safe and stable operation of the photovoltaic power station.

[0026] The multi-path tracking module 30 is used to synchronize the multiple identification points to the power station map for inspection analysis, obtain multiple local inspection points, and trace back to the multiple identification points based on the multiple local inspection points to perform multi-path tracking to generate multiple local inspection routes. Specifically, multiple identification points in the operation dynamic analysis results are synchronized to the power station map to perform more detailed inspection analysis. Based on the synchronized identification points, inspection classification and inspection management are performed to obtain multiple local inspection points, which are concentrated near the photovoltaic panels with abnormalities. Finally, multi-path tracking is performed based on the local inspection points to generate multiple local inspection routes, which focus on accessing the photovoltaic panels with abnormalities.

[0027] In a possible implementation, the multiple identification points are synchronized to the power plant map for inspection analysis to obtain multiple local inspection points, further including: associating the multiple identification points according to the multiple coordinate points to obtain multiple positioning correlation coefficients; synchronizing the multiple identification points to the power plant map for inspection classification according to the multiple positioning correlation coefficients to obtain multiple categories to be inspected; defining inspection time periods according to the multiple categories to be inspected, and performing inspection management on the multiple identification points according to the inspection time periods to generate inspection allocation management tasks; synchronizing the inspection allocation management tasks to the central monitoring unit for inspection recording to obtain multiple inspection monitoring record data, and adding the multiple inspection monitoring record data to the multiple local inspection points, and the multiple inspection monitoring record data have a corresponding relationship with the multiple local inspection points.

[0028] Specifically, the nearest neighbor algorithm is used to associate each identification point with the nearest coordinate point to ensure that the identification point can be accurately mapped to the specific location on the power station map, and multiple positioning correlation coefficients are generated. These coefficients represent the degree of correlation between the identification point and the coordinate point. Based on the positioning correlation coefficient and other relevant information (such as the type of photovoltaic panel, historical fault records, etc.), the identification points are inspected and classified. The classification criteria include priority, urgency, inspection difficulty, etc., and multiple categories to be inspected are generated. Each category contains a group of identification points with similar characteristics. A suitable inspection time period is set for each category to be inspected to ensure that all identification points can be inspected within the local inspection cycle, while avoiding excessive concentration or idleness of resources. A list of inspection time periods is generated, and each time period is associated with the corresponding category to be inspected.

[0029] According to the inspection time period, generate inspection allocation management tasks, including inspection points, inspection time, inspection robots and other information. Further optimize the tasks, such as merging adjacent inspection points to reduce movement time or adjusting the inspection order to optimize resource utilization, and generate an inspection allocation management task list to guide subsequent inspection work. Synchronize the inspection allocation management tasks to the central monitoring unit to ensure that the inspection robot can obtain the latest inspection information in real time. The inspection robot performs inspections according to the task requirements and records the data during the inspection process (such as photos, videos, measurements, etc., used to evaluate the operating status of photovoltaic panels). After the inspection is completed, the inspection monitoring record data is uploaded to the central monitoring unit, and the central monitoring unit associates the inspection monitoring record data with the corresponding local inspection points to ensure that each point has a complete inspection record. This implementation method associates identification points with coordinate points, ensuring that inspection tasks can accurately locate the specific location of the photovoltaic power station. By classifying inspections and defining inspection time periods, it optimizes the allocation and use of inspection resources and improves inspection efficiency. By real-time monitoring and updating of inspection tasks, it can flexibly respond to changes in the operating status of the photovoltaic power station and ensure the timeliness and effectiveness of inspection work.

[0030] In a possible implementation, multi-path tracking is performed based on the multiple local patrol points to trace back to the multiple identification points to generate multiple local patrol routes, further including: traversing the multiple photovoltaic panels according to the global patrol route to determine whether the multiple identification points have completed the traversal visit; when the multiple identification points have completed the traversal visit, edge monitoring is performed on the multiple local patrol points to set the patrol end point; the patrol starting point is extracted, and the multiple identification points are reversely traced back according to the patrol end point until they are traced back to the patrol starting point to generate multiple backtracking paths; path tracking analysis is performed based on the multiple backtracking paths to generate multiple path complexities, and the multiple path complexities include multiple time complexities and multiple space complexities; the multiple backtracking paths are weighted based on the multiple time complexities combined with the multiple space complexities to determine the multiple local patrol routes.

[0031] Specifically, it is determined whether the global inspection route has visited all the identification points to ensure that all potential problem points have been taken into consideration before generating the local inspection route. When the global inspection route completes the traversal of all identification points, edge monitoring is performed on multiple local inspection points (i.e., the area near the identification points) to find inspection points located at the edge of the local inspection point area. Based on the results of edge monitoring, an edge point is set as the inspection end point. This end point is the end point of the local inspection route and the point where the inspection robot needs to return to complete the inspection task.

[0032] The inspection starting point is the starting point of the global inspection route and the starting point of the local inspection route. Starting from the inspection end point, trace back in the opposite direction to the inspection starting point. In this process, multiple identification points will be passed and multiple backtracking paths will be formed. A detailed tracking analysis is performed on each backtracking path to determine its performance in the actual inspection process, and the time complexity and space complexity of each backtracking path are obtained. Among them, the time complexity is used to measure the length of time required to complete the backtracking path, which is related to factors such as the length of the path, the speed of the inspection robot, and possible obstacles. The space complexity is used to measure the complexity of the backtracking path in space, such as the tortuosity of the path, the required spatial range, etc. The backtracking path is weighted based on the time complexity and space complexity, and the efficiency and feasibility of the path are comprehensively considered. According to the weighted results, the best few backtracking paths are selected from multiple backtracking paths as local inspection routes. These routes are used to perform more detailed inspections on the identification points and their surrounding areas. This implementation method optimizes the inspection path through path tracking analysis and complexity calculation, reduces unnecessary inspection time and resource consumption, and improves the efficiency of photovoltaic power station operation and maintenance.

[0033] The local inspection route dynamic update module 40 is used to perform path optimization based on the multiple local inspection routes to generate a first local inspection route, dynamically update the first local inspection route according to the operation dynamic analysis result, and determine the Nth local inspection route. Specifically, based on the multiple local inspection routes, path optimization is performed to generate an optimal first local inspection route. As the operation dynamic analysis results change (such as new anomalies appearing, old anomalies being repaired, etc.), the first local inspection route is dynamically updated according to the global inspection route generation module 10, the photovoltaic panel operation dynamic analysis module 20 and the multi-path tracking module 30, that is, the existing path is modified or optimized to reflect the latest operation dynamic changes of the photovoltaic panel in real time, and the Nth local inspection route is generated.

[0034] In one possible implementation, path optimization is performed based on the multiple local inspection routes to generate a first local inspection route, which further includes: screening the multiple local inspection routes based on the multiple time complexities combined with multiple weight coefficients to determine a time-optimized local inspection route; screening the multiple local inspection routes based on the multiple space complexities combined with the multiple weight coefficients to determine a space-optimized local inspection route; cross-combining the time-optimized local inspection route with the space-optimized local inspection route to generate the first local inspection route.

[0035] Specifically, the weight coefficient refers to the value used to adjust or balance the relative importance of different evaluation indicators (such as time complexity and space complexity) in the path optimization process. For the time complexity and space complexity of each local inspection route, the corresponding weight coefficient is applied for weighted processing, that is, the time complexity and space complexity of each route will be multiplied by a specific weight coefficient to obtain the weighted time complexity and space complexity. According to the weighted time complexity, all local inspection routes are compared and screened, and the route with the lowest time complexity is selected as the time-optimized local inspection route. Similarly, according to the weighted space complexity, all local inspection routes are compared and screened, and the route with the lowest space complexity is selected as the space-optimized local inspection route. Finally, the time-optimized local inspection route and the space-optimized local inspection route are cross-combined, that is, some parts of the two routes are selected for splicing, or the characteristics of the two routes are combined according to a certain strategy (such as priority, complementarity, etc.) to generate a local inspection route that takes both time efficiency and space efficiency into consideration, that is, the first local inspection route. This implementation method obtains the routes with the best performance in time and space by screening the time-optimized local inspection routes and the space-optimized local inspection routes respectively. By cross-combining these two routes, the time efficiency and space efficiency are balanced to generate a first local inspection route that is both efficient and practical, thereby improving the accuracy and practicality of path optimization.

[0036] In one possible implementation, the time-optimized local inspection route is cross-combined with the space-optimized local inspection route to generate the first local inspection route, further comprising: defining a time optimization target based on the time-optimized local inspection route, and defining a space optimization target based on the space-optimized local inspection route; calculating the multiple local inspection points according to the time optimization target to construct a time matrix; setting local inspection areas based on the multiple local inspection points, calculating the local inspection areas according to the space optimization target to construct an inspection path network; cross-matching the time matrix with the inspection path network to generate multiple time-space inspection combinations, performing path reachability verification on the multiple time-space inspection combinations, and obtaining the first local inspection route based on the verification results.

[0037] Specifically, the time optimization goal is used to minimize the execution time of each task in the inspection process to ensure that the inspection robot can complete the task on time; the space optimization goal is used to minimize the moving distance of the inspection area without repeating the inspection points to ensure that each point can be covered in a reasonable order. According to the time optimization goal, multiple local inspection points are calculated to construct a time matrix, which records the time required from each inspection point to other inspection points and the time required to complete each inspection task. Local inspection areas are set based on multiple local inspection points, and the local inspection areas are calculated according to the space optimization goal to construct an inspection path network. The inspection path network is a graphical representation in which nodes represent inspection points and edges represent the connection paths between inspection points. The time matrix is ​​cross-matched with the inspection path network to generate multiple spatiotemporal inspection combinations, which take into account both time efficiency (through the time matrix) and space efficiency (through the inspection path network). Through the simulation of the inspection route, the path reachability of multiple spatiotemporal inspection combinations is verified, including checking whether the path in each combination is actually feasible, whether it meets the physical limitations of the inspection robot (such as maximum moving speed, turning radius, etc.), and whether it complies with the safety regulations of the photovoltaic power station. According to the verification results, the optimal spatiotemporal inspection combination is selected as the first local inspection route. This route takes into account both time efficiency and space efficiency, and is feasible in actual operation, maximizing the inspection efficiency and quality.

[0038] In a possible implementation, the first local inspection route is dynamically updated according to the operation dynamic analysis result to determine the Nth local inspection route, further including: performing regular reinforcement learning according to the operation dynamic analysis result to construct a multi-period path learning log; S1: extracting the target period path learning log based on the multi-period path learning log to perform matching evaluation on the first local inspection route to generate a path matching score; S2: judging whether the path matching score is greater than or equal to the expected score threshold, if the path matching score is less than the expected score threshold, generating an update instruction, updating the first local inspection route through the update instruction, and generating a second local inspection route; S3: executing the second local inspection route based on the current inspection cycle, and when in the next inspection cycle, repeatedly iterating S1 and S2 for dynamic update to determine the Nth local inspection route.

[0039] Specifically, according to the results of the dynamic analysis of the operation, the inspection robot path planning platform regularly conducts reinforcement learning, records the path planning, execution and feedback in each inspection cycle, and forms multi-period path learning logs. These logs contain data such as the selection of inspection routes, execution efficiency, and abnormal detection in different time periods. From the multi-period path learning logs, the current or recent latest inspection cycle is selected as the target period, and the path learning log of the target period is extracted for matching and evaluating the first local inspection route.

[0040] The first partial inspection route is matched with the target period path learning log, and the applicability of the first partial inspection route in the current or recent inspection cycle is evaluated, and finally a path matching score is generated. The path matching score is compared with the preset expected score threshold, which is a standard value set according to the operation and maintenance requirements of the photovoltaic power station, the performance of the inspection robot and other factors, and is used to determine whether the first partial inspection route meets the requirements.

[0041] If the path matching score is less than the expected score threshold, it means that the first local inspection route has performed poorly in the current or recent inspection cycle and needs to be updated. At this time, the platform will generate an update instruction to guide the update of the first local inspection route. According to the update instruction, the first local inspection route is updated to generate the second local inspection route. In the current inspection cycle, the inspection robot will perform inspections along the second local inspection route.

[0042] When entering the next inspection cycle, the platform will repeat steps S1 and S2 to match and evaluate the second local inspection route, and decide whether to continue updating based on the evaluation results. This process will continue to iterate until the Nth local inspection route that meets the expected score threshold is generated. This implementation method matches and evaluates the first local inspection route by extracting the target period path learning log. The platform can accurately evaluate the applicability of the current local inspection route and decide whether to update it based on the evaluation results. This dynamic update method can ensure that the local inspection route is always synchronized with the operation and maintenance needs of the photovoltaic power station, improving the inspection efficiency and accuracy.

[0043] In a possible implementation, regular reinforcement learning is performed according to the operation dynamic analysis results to construct a multi-period path learning log, which further includes: extracting the photovoltaic panel layout information of the photovoltaic power station based on the power station map, and defining the state space according to the photovoltaic panel layout information; performing reachability analysis based on the global inspection route, and defining the action space according to the reachability analysis results; introducing a reward function, mapping the operation dynamic analysis results to the state space and combining the reward function to perform reinforcement learning to generate an operation state learning record; mapping the operation dynamic analysis results to the action space and combining the reward function to perform reinforcement learning to generate an operation action learning record; associating and integrating the operation state learning record with the operation action learning record according to the inspection cycle to construct a multi-period path learning log.

[0044] Specifically, the layout information of the photovoltaic panels, such as the specific location, quantity, and arrangement, is extracted from the power station map. Based on the extracted photovoltaic panel layout information, the state space is defined. The state space is all possible sets that describe the current state of the environment in reinforcement learning. For example, each state can be represented as the current operating status of a photovoltaic panel or a group of photovoltaic panels (such as power output, temperature, whether it is faulty, etc.). Reachable analysis refers to determining all possible states that can be reached by performing certain actions starting from the current state. Based on the global inspection route, all photovoltaic panel positions that the inspection robot can reach starting from the inspection starting point on the power station map are analyzed. Based on these reachable positions, the action space is defined. For example, the action can be moving to a specific photovoltaic panel position, adjusting the inspection speed, etc.

[0045] A reward function is introduced, which can reflect the effect of different actions performed by the inspection robot. For example, if the inspection robot successfully detects a faulty photovoltaic panel, it can get a positive reward; if it takes too much time or energy to reach a certain location, it will get a negative reward. The results of the operation dynamic analysis are mapped to the state space, and combined with the reward function for reinforcement learning to generate operation state learning records. Similarly, the results of the operation dynamic analysis are mapped to the action space, and combined with the reward function for reinforcement learning to generate operation action learning records. These records contain the reward values ​​and state transition information after taking different actions in different states.

[0046] At the end of each inspection cycle, the operation state learning records and operation action learning records within the cycle are associated and added to the multi-period path learning log to form a log containing the learning results of multiple inspection cycles for subsequent matching evaluation and path update. This implementation method extracts the photovoltaic panel layout information of the photovoltaic power station and defines the state space and action space based on this information, providing a clear learning framework for the inspection robot. After introducing the reward function, the inspection robot can learn how to inspect more effectively based on the reward value after executing the action. By generating the operation state learning records and the operation action learning records and associating and integrating them according to the inspection cycle, a log containing the learning results of multiple inspection cycles is constructed. This log provides data support for the subsequent matching evaluation and path update, realizes the intelligent and dynamic update of the inspection robot path planning, and improves the operation and maintenance efficiency and accuracy of the photovoltaic power station.

[0047] The global inspection optimization path generation module 50 is used to feedback-adjust the global inspection route based on the Nth local inspection route to generate a global inspection optimization path. Specifically, feedback-adjust the global inspection route based on the Nth local inspection route, that is, the global inspection route is regularly adjusted according to the updated route of the local inspection, so that the global inspection route always includes the local inspection route updated in real time, which can reflect the current status and demand of the photovoltaic power station in real time, thereby improving the inspection efficiency and effect. The embodiment of the present application adopts a method of generating a power station map according to the layout of photovoltaic panels of a photovoltaic power station, and covering and traversing multiple photovoltaic panels based on the power station map to generate a global inspection route, performing a dynamic operation analysis on multiple photovoltaic panels according to the global inspection route, generating multiple identification points according to the analysis results, synchronizing the multiple identification points to the power station map for inspection analysis, obtaining multiple local inspection points, and tracing back to the identification points based on these points for multi-path tracking to generate multiple local inspection routes, performing path optimization based on the multiple local inspection routes, generating a first local inspection route, and dynamically updating the first local inspection route according to the dynamic operation analysis results, determining the Nth local inspection route, performing feedback adjustment on the global inspection route based on the Nth local inspection route, generating a global inspection optimization path, and other technical means, which can monitor the operating status of photovoltaic panels in real time, dynamically adjust the inspection route according to the analysis results, ensure that key areas receive focus, and avoid wasting time and resources in areas in good condition, thereby realizing intelligent path planning of photovoltaic power station inspection robots and achieving the technical effect of improving the efficiency and accuracy of inspections.

[0048] Although the present application makes various references to certain modules in the platform according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0049] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. Inspection robot path planning platform for photovoltaic power station operation and maintenance, characterized by: The platform includes: A global inspection route generation module, the global inspection route generation module is used to generate a power station map according to the photovoltaic panel layout of the photovoltaic power station, and to cover and traverse multiple photovoltaic panels based on the power station map to generate a global inspection route; A photovoltaic panel operation dynamic analysis module, the photovoltaic panel operation dynamic analysis module is used to perform operation dynamic analysis on multiple photovoltaic panels according to the global inspection route, and generate multiple identification points according to the operation dynamic analysis results; A multi-path tracking module, the multi-path tracking module is used to synchronize the multiple identification points to the power station map for inspection analysis, obtain multiple local inspection points, trace back to the multiple identification points based on the multiple local inspection points, perform multi-path tracking, and generate multiple local inspection routes; A local inspection route dynamic update module, the local inspection route dynamic update module is used to perform path optimization based on the multiple local inspection routes to generate a first local inspection route, dynamically update the first local inspection route according to the operation dynamic analysis result, and determine the Nth local inspection route; A global inspection optimization path generation module is used to perform feedback adjustment on the global inspection route based on the Nth local inspection route to generate a global inspection optimization path.

2. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 1, characterized in that: Performing a dynamic operation analysis on multiple photovoltaic panels according to the global inspection route, and generating multiple identification points according to the dynamic operation analysis results, including: Performing coordinate position analysis based on multiple photovoltaic panels combined with the power station map to determine multiple coordinate points; Formulate an inspection cycle, and traverse the multiple coordinate points through the global inspection route according to the inspection cycle to perform operation monitoring and obtain multiple operation data; Perform abnormality diagnosis based on the multiple operating data combined with environmental parameters, determine multiple abnormality types, perform dynamic analysis according to the multiple abnormality types, and generate the operation dynamic analysis result, wherein the operation dynamic analysis result includes multiple abnormal dynamic tags; The multiple abnormal dynamic labels are mapped to the multiple coordinate points for identification, and the multiple identification points are generated.

3. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 2, characterized in that: The multiple identification points are synchronized to the power station map for inspection analysis to obtain multiple local inspection points, including: Associating the multiple identification points according to the multiple coordinate points to obtain multiple positioning association coefficients; Synchronizing the multiple identification points to the power station map according to the multiple positioning correlation coefficients for inspection classification to obtain multiple categories to be inspected; Defining a waiting inspection time period according to the multiple waiting inspection categories, performing inspection management on the multiple identification points according to the waiting inspection time period, and generating an inspection allocation management task; The inspection allocation management task is synchronized to the central monitoring unit for inspection recording, a plurality of inspection monitoring record data are obtained, and the plurality of inspection monitoring record data are added to the plurality of local inspection points, and the plurality of inspection monitoring record data correspond to the plurality of local inspection points.

4. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 1, characterized in that: Based on the multiple local inspection points, tracing back to the multiple identification points to perform multi-path tracking, and generating multiple local inspection routes, including: Traversing the plurality of photovoltaic panels according to the global inspection route, and determining whether the plurality of identification points have been traversed and visited; When the traversal access of the multiple identification points is completed, edge monitoring is performed on the multiple local inspection points to set the inspection end point; Extracting the inspection starting point, and tracing back the multiple identification points according to the inspection end point until tracing back to the inspection starting point, thereby generating multiple tracing paths; Performing path tracking analysis based on the multiple backtracking paths to generate multiple path complexities, wherein the multiple path complexities include multiple time complexities and multiple space complexities; The multiple backtracking paths are weighted based on the multiple time complexities combined with the multiple space complexities to determine the multiple local inspection routes.

5. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 4, characterized in that: Performing path optimization based on the multiple local inspection routes to generate a first local inspection route includes: Screening the multiple local inspection routes based on the multiple time complexities combined with multiple weight coefficients to determine a time-optimized local inspection route; Screening the multiple local inspection routes based on the multiple spatial complexities combined with the multiple weight coefficients to determine a spatially optimized local inspection route; The time-optimized local inspection route is cross-combined with the space-optimized local inspection route to generate the first local inspection route.

6. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 5, characterized in that: Cross-combining the time-optimized local inspection route with the space-optimized local inspection route to generate the first local inspection route includes: Defining a time optimization target based on the time-optimized local inspection route, and defining a space optimization target based on the space-optimized local inspection route; Calculating the multiple local inspection points according to the time optimization target to construct a time matrix; Setting a local inspection area based on the multiple local inspection points, calculating the local inspection area according to the spatial optimization target, and constructing an inspection path network; The time matrix is ​​cross-matched with the inspection path network to generate a plurality of time-space inspection combinations, path reachability verification is performed on the plurality of time-space inspection combinations, and the first local inspection route is obtained according to the verification result.

7. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 2, characterized in that: Dynamically updating the first local inspection route according to the operation dynamic analysis result to determine the Nth local inspection route includes: Perform regular reinforcement learning based on the operation dynamic analysis results and construct a multi-period path learning log; S1: extracting the target period path learning log based on the multi-period path learning logs to perform matching evaluation on the first local inspection route and generate a path matching score; S2: determining whether the path matching score is greater than or equal to an expected score threshold; if the path matching score is less than the expected score threshold, generating an update instruction, and updating the first local inspection route through the update instruction to generate a second local inspection route; S3: Execute the second local inspection route based on the current inspection cycle. When in the next inspection cycle, repeat the iterations of S1 and S2 to perform dynamic updates and determine the Nth local inspection route.

8. The inspection robot path planning platform for photovoltaic power station operation and maintenance according to claim 7, characterized in that: Regular reinforcement learning is performed based on the dynamic analysis results of the operation, and a multi-period path learning log is constructed, including: Extracting photovoltaic panel layout information of the photovoltaic power station based on the power station map, and defining a state space according to the photovoltaic panel layout information; Performing reachability analysis based on the global inspection route, and defining an action space according to the reachability analysis result; Introducing a reward function, mapping the operation dynamic analysis result to the state space and combining the reward function to perform reinforcement learning, and generating an operation state learning record; Mapping the running dynamic analysis result to the action space and combining it with the reward function to perform reinforcement learning, and generating a running action learning record; The operation status learning record and the operation action learning record are associated and integrated according to the inspection cycle to construct a multi-period path learning log.

Images