Photovoltaic power station intelligent inspection system and method based on deep learning

Abstract

The invention belongs to the technical field of photovoltaic power station inspection, and particularly relates to a photovoltaic power station intelligent inspection system and method based on deep learning, and the system comprises an unmanned plane which is used for carrying out the inspection and shooting an image; the data acquisition module is used for receiving images and time shot by the unmanned aerial vehicle; the digital photovoltaic power station module is used for establishing a corresponding virtual simulation scene image and a semantic instance image based on an actual photovoltaic power station; by monitoring the remaining electric quantity of the unmanned aerial vehicle and displaying the predicted remaining travel of the unmanned aerial vehicle, if the current predicted remaining travel of the unmanned aerial vehicle is smaller than the optimal inspection path planned by the path planning module, the unmanned aerial vehicle needs to be charged or replaced for inspection, and it can be ensured that the unmanned aerial vehicle does not need to be charged or replaced during one inspection task. All photovoltaic panels in a photovoltaic power station can be completely shot, the situation of charging in the midway is avoided, and the problem that defects possibly exist in the photovoltaic panels in a shot area due to time intervals generated by charging is also avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic power stations, and specifically relates to a photovoltaic power station intelligent inspection system and method based on deep learning. Background Art

[0002] With the rapid development of new energy, the operation and maintenance of photovoltaic power stations are facing new challenges. The traditional manual inspection method has problems such as low efficiency and high cost, which can hardly meet the operation and maintenance needs of large-scale power stations. With the rapid development of drones in various fields, the operation and maintenance of photovoltaic power stations at this stage also basically adopts drones for inspection.

[0003] For example, the publication number CN112633535A is a method and system for intelligent inspection of photovoltaic power stations based on drone images. The system includes a drone data import module, an inspection data analysis module, a deep learning algorithm module, a digital photovoltaic power station module, an inspection result management module and an inspection playback module. By introducing a deep learning algorithm, more accurate photovoltaic defect detection can be achieved. By establishing a digital photovoltaic power station, the virtual scene image and the real collected image are aligned, and then the position and number of the photovoltaic panels in the real scene are obtained. It is convenient for the operation and maintenance personnel to intuitively debug and inspect the management, greatly improve the level of automated operation and maintenance of photovoltaic power stations, and effectively improve the inspection efficiency of photovoltaic power stations.

[0004] Based on the above, from the perspective of inspection video and image processing and analysis of photovoltaic power stations, the existing technology can basically complete intelligent inspection. However, in the actual process of intelligent inspection of photovoltaic power stations, attention should also be paid to the control of drones. Since drones basically rely on battery drive and the battery size is fixed, when adapting to different photovoltaic power stations, due to the different areas of different photovoltaic power stations, there may be a possibility that the drone cannot fully collect images of all photovoltaic panels in the photovoltaic power station within a fixed time. In other words, if the above situation occurs, the drone will inevitably need to recharge in the middle and take a second shot. Then, due to the time interval, if the photovoltaic panel corresponding to the image taken by the drone once is blocked or has light spots during this period, it cannot be discovered in time, which will affect the operation and maintenance management of the photovoltaic power station.

[0005] To this end, the present invention provides a photovoltaic power station intelligent inspection system and method based on deep learning. Summary of the invention

[0006] In order to make up for the deficiencies of the prior art, at least one technical problem raised in the background technology is solved.

[0007] The technical solution adopted by the present invention to solve the technical problem is: a photovoltaic power station intelligent inspection system and method based on deep learning described in the present invention includes:

[0008] Drones, used to conduct inspections and take images;

[0009] Data acquisition module, used to receive the images and time taken by the drone;

[0010] Digital photovoltaic power station module, which establishes corresponding virtual simulation scene images and semantic instance images based on the actual photovoltaic power station;

[0011] The image processing module is used to process the captured images and analyze the captured images into the required inspection results, including the shooting time, altitude, longitude and latitude, defect type, defect location and photovoltaic panel number;

[0012] A deep learning algorithm module provides support for the image processing module, including a defect detection algorithm and an image matching algorithm. The defect detection algorithm is used to detect various defects in the image, and the image matching algorithm is used to match the virtual simulation scene image of the digital photovoltaic power station module with the actual captured image;

[0013] The data storage module is used to store and manage all inspection data of the photovoltaic power station, including the original inspection data and the parsed inspection results, and supports the query of the inspection results;

[0014] It also includes a path planning module for planning the optimal inspection path for drones;

[0015] It also includes a power and trip management module to monitor the remaining power of the drone and the estimated remaining trip;

[0016] The matching module selects drones to conduct inspections along the optimal inspection path.

[0017] Preferably, the method by which the path planning module plans the optimal inspection path with the minimum total travel distance is:

[0018] Use Dijkstra algorithm to draw multiple inspection paths passing through N points;

[0019] Calculate the total travel distance of multiple inspection paths;

[0020] The inspection path corresponding to the minimum total travel distance is selected as the optimal inspection path.

[0021] Preferably, the matching module further includes a calculation unit for calculating the utilization efficiency of the drone by:

[0022] Get the estimated remaining distance L of multiple drones i And the total travel distance L of the optimal inspection path z ;

[0023] According to the formula:

[0024] M i =Lz / L i ×100%

[0025] Among them, M i represents the utilization efficiency of any UAV, and i represents any UAV.

[0026] Preferably, the remaining power of the drone is Q i The estimated remaining distance L i The relationship satisfies the formula:

[0027] L i =Q i ·e -γt / A i 60

[0028] Among them, Q i represents the battery capacity of any drone, in mAh; A represents the average current consumption per unit distance of any drone, in mA / km; e -γt represents the attenuation parameter of the UAV battery over the service time, Q i ·e -γt Indicates the battery update capacity of any drone.

[0029] Preferably, it also includes an analysis module for analyzing the composition of the optimal inspection path, including the upwind section and the downwind section; the calculation unit is also used to calculate the estimated power consumption Q of the optimal inspection path. z ;

[0030] Among them, the downwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is less than 90°, and the headwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is greater than or equal to 90°.

[0031] Preferably, the estimated power consumption Q of the optimal inspection path is calculated z The method is:

[0032] Get the upwind segment travel L n 、Downwind section travel L s And the average current consumption of the downwind section is A s ;

[0033] According to the formula:

[0034] Q z =(L s +kL n )·A s 60

[0035] Among them, Q z Indicates the estimated power consumption, L s represents the downwind segment in the optimal inspection path, L nrepresents the upwind section in the optimal inspection path, k represents the average current consumption factor of the upwind section compared with the downwind section, A s Indicates the average current consumption during the downwind leg of the journey.

[0036] Preferably, based on the estimated power consumption Q z The calculation module is also used to update the drone utilization efficiency by:

[0037] According to the formula:

[0038]

[0039] Calculate the theoretical expression value of the optimal inspection path

[0040] Theoretical expression value Substitute into the formula: M i =L z / L i ×100%

[0041] The update formula is:

[0042] Calculate the updated utilization efficiency corresponding to multiple drones

[0043] Preferably, the matching module updates the utilization efficiency based on the drone The method to match any drone is:

[0044] According to the calculated updated utilization efficiency corresponding to multiple drones

[0045] Exclude update utilization efficiencies greater than 100% The corresponding drone;

[0046] Then select the one with the greatest update efficiency The corresponding drone is used as an inspection drone.

[0047] Preferably, when all updates utilize efficiency If both are greater than 100%, the path planning module is also used to replan the optimal inspection path with the lowest estimated power consumption, the method is:

[0048] Obtain multiple inspection paths drawn by the Dijkstra algorithm;

[0049] Use the analysis module to analyze the composition of the inspection path;

[0050] Calculate the estimated power consumption Q of multiple inspection paths j , where j represents any inspection path;

[0051] Select estimated power consumption Q j The inspection path corresponding to the minimum value is used as the re-planned UAV inspection path.

[0052] A photovoltaic power station intelligent inspection method based on deep learning, comprising the following steps:

[0053] S1: Obtain the optimal inspection path;

[0054] S2: Select a drone to perform the task corresponding to the optimal inspection path;

[0055] S3: take images of photovoltaic panels;

[0056] S4: Analyze the photovoltaic panel image;

[0057] S5: Process the image to obtain the photovoltaic panel number corresponding to the defective image.

[0058] The beneficial effects of the present invention are as follows:

[0059] 1. The present invention describes a deep learning-based intelligent inspection system and method for photovoltaic power stations. Through the power and travel management module, the remaining power of the drone is monitored and the estimated remaining travel of the drone is displayed. After the path planning module plans the optimal inspection path, it can be compared with the estimated remaining travel of the current drone. If the current estimated remaining travel of the drone is less than the optimal inspection path planned by the path planning module, the drone needs to be recharged or replaced for inspection. In this way, it can be ensured that the drone can fully photograph all photovoltaic panels in the photovoltaic power station during an inspection mission, avoiding the situation of recharging in the middle of the journey, thereby avoiding the time interval caused by recharging, which may lead to defects in the photovoltaic panels in the area where the photography has been completed.

[0060] 2. The photovoltaic power station intelligent inspection system and method based on deep learning described in the present invention introduces attenuation parameters to update the relationship between the remaining power of the drone and the travel distance, so as to obtain the corresponding relationship between the power of the drone and the travel distance at the current moment; obtain the updated estimated remaining travel distance L of any drone decreasing over time i , used to characterize the UAV service time and update the estimated remaining distance L i Therefore, when the matching module matches the optimal inspection path with the corresponding UAV, it can ensure that the UAV can fully execute the task of the optimal inspection path without performing the operation of recharging power in the middle, so as to prevent affecting the inspection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The present invention will be further described below in conjunction with the accompanying drawings.

[0062] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0063] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0064] like Figure 1 As shown, a photovoltaic power station intelligent inspection system and method based on deep learning according to an embodiment of the present invention includes a drone for inspection and taking images; a data acquisition module for receiving images taken by the drone and the time; a digital photovoltaic power station module for establishing corresponding virtual simulation scene images and semantic instance images based on the actual photovoltaic power station; an image processing module for processing the taken images and parsing the taken images into the required inspection results, including shooting time, altitude, longitude and latitude, defect type, defect location and photovoltaic panel number; a deep learning algorithm module for providing support for the image processing module, including a defect detection algorithm and an image matching algorithm, wherein the defect detection algorithm is used to detect various defects in the image, and the image matching algorithm is used to match the virtual simulation scene image of the digital photovoltaic power station module with the actually collected taken images; a data storage module for storing and managing all data of the photovoltaic power station inspection, including the original inspection data and the parsed inspection results, and supporting the query of the inspection results;

[0065] In one embodiment of the present invention, after the drone flies along the inspection path to obtain images of all photovoltaic panels in the photovoltaic power station area, the photovoltaic panel images are then analyzed to output the faulty photovoltaic panel number and location. First, a digital photovoltaic power station is established, and virtual simulation scene images and semantic instance images corresponding to the actual photovoltaic power station scene are generated from the bird's-eye view image of the actual photovoltaic power station. When the photovoltaic panel image is analyzed, pixel extraction is performed first, and then a defect detection algorithm is input based on the pixels. The defect detection algorithm outputs a judgment result, and then the judgment result output by the defect detection algorithm, including the defect type information and the bounding box coordinates corresponding to the defect, is mapped to the semantic instance image, and the numbering information and coordinate position of the corresponding photovoltaic panel are obtained from the semantic instance image, thereby completing the purpose of inspection and avoiding the failure to discover the photovoltaic panel fault in time.

[0066] It also includes a path planning module for planning the optimal inspection path for drones;

[0067] It also includes a power and trip management module to monitor the remaining power of the drone and the estimated remaining trip;

[0068] The matching module selects drones to conduct inspections along the optimal inspection path.

[0069] In addition, from the perspective of inspection photography and image processing and analysis of photovoltaic power stations, existing technologies are basically able to complete intelligent inspections. However, in the actual process of intelligent inspections of photovoltaic power stations, attention should also be paid to the control of drones. Since drones are basically battery-driven and the battery size is fixed, when adapting to different photovoltaic power stations, due to the different areas of different photovoltaic power stations, there may be a possibility that the drone cannot completely collect images of all photovoltaic panels in the photovoltaic power station within a fixed time. In other words, if the above situation occurs, the drone will inevitably need to recharge midway and take a second shot. Then, due to the time interval, if the photovoltaic panel corresponding to the image taken by the drone once is blocked or has light spots during this period, it cannot be discovered in time, which will affect the operation and maintenance management of the photovoltaic power station.

[0070] In one embodiment of the present invention, a drone is used to perform inspections according to the optimal inspection path planned by the path planning module. Before the inspection begins, the power and trip management module is relied upon to monitor the remaining power of the drone and display the estimated remaining trip of the drone. After the path planning module plans the optimal inspection path, it can be compared with the current estimated remaining trip of the drone. If the current estimated remaining trip of the drone is less than the optimal inspection path planned by the path planning module, the drone needs to be recharged or replaced for inspection. In the above manner, it can be ensured that the drone can fully photograph all photovoltaic panels in the photovoltaic power station during an inspection mission, avoiding the situation of recharging in the middle, and thus avoiding the time interval caused by recharging, which may lead to defects in the photovoltaic panels in the area where the photography has been completed.

[0071] In one embodiment, the method by which the path planning module plans the optimal inspection path with the minimum total travel distance is:

[0072] Use Dijkstra algorithm to draw multiple inspection paths passing through N points;

[0073] Calculate the total travel distance of multiple inspection paths;

[0074] The inspection path corresponding to the minimum total travel distance is selected as the optimal inspection path.

[0075] At present, AI algorithms are able to draw multiple paths passing through points based on preset points. In one embodiment of the present invention, with the help of the Dijkstra algorithm and the point map corresponding to the photovoltaic panels pre-constructed by the digital photovoltaic power station module, multiple paths passing through the photovoltaic panel point map can be drawn. However, there must be differences in the total strokes of the multiple paths. In order to complete the inspection of the photovoltaic power station as efficiently as possible, all inspection paths can be easily distinguished by calculating the total strokes of multiple inspection paths, and by selecting the inspection path corresponding to the minimum total stroke as the optimal inspection path, the efficiency of the inspection of the photovoltaic power station can be improved.

[0076] In one embodiment, the matching module further includes a calculation unit for calculating the utilization efficiency of the drone, and the method is:

[0077] Get the estimated remaining distance L of multiple drones i And the total travel distance L of the optimal inspection path z ;

[0078] According to the formula:

[0079] M i =L z / L i ×100%

[0080] Among them, M i represents the utilization efficiency of any UAV, and i represents any UAV.

[0081] In addition to the path planning module being able to construct multiple paths according to the preset photovoltaic panel point map, after obtaining the optimal inspection path corresponding to the minimum total travel distance by calculation, it is also necessary to select a drone that meets the optimal inspection path. In one embodiment of the present invention, the estimated remaining travel distance L of multiple drones is calculated. i , and obtain the total travel distance L of the optimal inspection path z , the utilization efficiency M corresponding to any UAV is calculated by the formula i , obtain the utilization efficiency M of multiple drones executing the optimal inspection path i It is worth noting that the utilization efficiency M is set here. i , which can intuitively reflect that multiple drones can efficiently complete the inspection task when executing the task of the optimal inspection path. For example, the total distance of the optimal inspection path is 10 kilometers, and there are currently four drone devices, and the estimated remaining distances are: A = 8 kilometers, B = 10 kilometers, C = 12 kilometers, and D = 6 kilometers;

[0082] According to the matching principle, the drones that meet the optimal inspection path are B and C;

[0083] Based on the principle of maximizing utilization efficiency, the utilization efficiency is calculated using the formula:

[0084] M = optimal inspection path / estimated remaining travel × 100%; according to the formula, M B =100%, and M C =83.3%;

[0085] Based on the above, M C Less than M B , which means that the utilization efficiency of drone C will be lower than that of drone B. More specifically, if there are multiple photovoltaic power stations that need to be inspected, if drone C is selected for inspection for the photovoltaic power station with the above-mentioned optimal inspection path of 10 kilometers, then assuming that the total inspection path of the optimal inspection path of another photovoltaic power station is 12 kilometers, the remaining drones will not be able to meet the requirement and will need to be recharged midway, which may cause defects in the photovoltaic panels in the photographed area.

[0086] In one embodiment, the remaining power of the drone is Q i The estimated remaining distance L i The relationship satisfies the formula:

[0087] L i =Q i ·e -γt / A i 60

[0088] Among them, Q i Indicates the remaining battery capacity of any drone, in mAh; A indicates the average current consumption per unit distance of any drone, in mA / km; e -γt represents the attenuation parameter of the UAV battery over the service time, Q i ·e -γt Indicates the battery update capacity of any drone.

[0089] Based on power and travel management, since the drone is powered by lithium batteries, after a period of continuous service, the estimated remaining travel time L of the drone lithium battery is i The relationship between the remaining power of the drone and the flight distance will be discounted, so it is necessary to update the corresponding relationship between the power of the drone and the flight distance through calculation; however, after the drone has been in service for a period of time, its battery density decreases and its internal resistance increases, which will cause the battery to be unable to support the theoretical flight distance of the drone. Therefore, by introducing the attenuation parameter, the relationship between the remaining power of the drone and the flight distance is updated, so that the corresponding relationship between the power of the drone and the flight distance at the current moment can be obtained; Based on the above formula, the updated estimated remaining flight distance L of any drone that decreases over time is obtained i , used to characterize the UAV service time and update the estimated remaining distance L iTherefore, when the matching module matches the optimal inspection path with the corresponding UAV, it can ensure that the UAV can fully execute the task of the optimal inspection path without performing the operation of recharging power in the middle, so as to prevent affecting the inspection results.

[0090] In one embodiment, an analysis module is further included for analyzing the composition of the optimal inspection path, including the upwind segment and the downwind segment; the calculation unit is also used to calculate the estimated power consumption Q of the optimal inspection path. z ;

[0091] Among them, the downwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is less than 90°, and the headwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is greater than or equal to 90°.

[0092] In addition to the fact that the drone's own battery decay may cause the drone to recharge midway during the inspection mission, when the drone is performing an inspection mission, encountering bad weather will also cause the drone's estimated remaining range L i Based on the above, in one embodiment of the present invention, the composition of the optimal inspection path planned by the path planning module needs to be analyzed by the analysis module. That is to say, on the date when the drone is scheduled to perform the inspection task, the weather conditions need to be considered. When there are multiple headwind sections in the optimal inspection path, the estimated remaining distance L of the drone will be shortened. i It is worth noting that in one embodiment, the composition of the optimal inspection path is divided according to the angle between the UAV's travel direction and the wind direction. Specifically, the downwind segment indicates a trip in which the angle between the UAV's travel direction and the wind direction is less than 90°, and the upwind segment indicates a trip in which the angle between the UAV's travel direction and the wind direction is greater than or equal to 90°.

[0093] In one embodiment, the estimated power consumption Q of the optimal inspection path is calculated z The method is:

[0094] Get the upwind segment travel L n 、Downwind section travel L s And the average current consumption of the downwind section is A s ;

[0095] According to the formula:

[0096] Q z =(L s +kL n )·A s 60

[0097] Among them, Q z Indicates the estimated power consumption, L srepresents the downwind segment in the optimal inspection path, L n represents the upwind section in the optimal inspection path, k represents the average current consumption factor of the upwind section compared with the downwind section, A s Indicates the average current consumption during the downwind leg of the journey.

[0098] According to the composition of the optimal inspection path obtained by the analysis module, namely, the downwind segment and the upwind segment, in one embodiment of the present invention, the estimated power consumption Q of the optimal inspection path needs to be calculated using the obtained downwind segment and the upwind segment. z , to prevent the use of UAVs to improve efficiency M i The matched drone is affected by bad weather, resulting in its estimated remaining range L i The power consumption Q is calculated by introducing the average current consumption factor k of the upwind segment compared with the downwind segment. z , according to the formula:

[0099] Q z =L s ·A s 60+L n ·kA s 60

[0100] Among them, kA s Indicates the average current consumption of the upwind segment, in mA / km;

[0101] By transforming the formula, we get the formula:

[0102] Q z =(L s +kL n )·A s 60

[0103] Based on the above formula, the estimated power consumption Q of the optimal inspection path is calculated. z , thus the remaining power Q of the drone i Compare to prevent the drone from estimating the remaining distance L i The task of executing the optimal inspection path cannot be fully executed.

[0104] In one embodiment, based on the estimated power consumption Q z The calculation module is also used to update the drone utilization efficiency by:

[0105] According to the formula:

[0106]

[0107] Calculate the theoretical expression value of the optimal inspection path

[0108] Theoretical expression value Substitute into the formula: M i =L z / L i ×100%

[0109] The update formula is:

[0110] Calculate the updated utilization efficiency corresponding to multiple drones

[0111] Based on the above formula, the estimated power consumption Q z There may be multiple drones, so in order to efficiently complete the inspection task, it is also necessary to estimate the power consumption Q z Recalculate the UAV utilization efficiency. In one embodiment of the present invention, the theoretical expression value of the optimal inspection path is obtained by the above formula: The theoretical expression value The actual power consumption corresponds to the estimated power consumption Q z The estimated travel distance is different from the calculated total travel distance of the optimal inspection path. The influence of the headwind section on the entire journey is also considered, and the theoretical expression value obtained is The estimated remaining distance L of the drone i , calculate the updated utilization efficiency Used to match drones that can complete the optimal inspection path task through the matching module.

[0112] In one embodiment, the matching module updates the utilization efficiency of the drone based on The method to match any drone is:

[0113] According to the calculated updated utilization efficiency corresponding to multiple drones

[0114] Exclude update utilization efficiencies greater than 100% The corresponding drone;

[0115] Then select the one with the greatest update efficiency The corresponding drone is used as an inspection drone.

[0116] As an embodiment of the present invention, when the updated utilization efficiency corresponding to multiple drones is calculated Finally, first exclude update utilization efficiency greater than 100% The corresponding drone, that is, excluding the drones whose estimated remaining range is less than the theoretical expression value The corresponding drone is then selected from the remaining drones to update the utilization efficiency Maximum corresponding drone, update utilization efficiency The closer it is to 100%, the closer it is to the theoretical value. The closer they are, the more efficient the drone can be in completing its inspection mission.

[0117] In one embodiment, when all updates utilize efficiency If both are greater than 100%, the path planning module is also used to replan the optimal inspection path with the lowest estimated power consumption, the method is:

[0118] Obtain multiple inspection paths drawn by the Dijkstra algorithm;

[0119] Use the analysis module to analyze the composition of the inspection path;

[0120] Calculate the estimated power consumption Q of multiple inspection paths j , where j represents any inspection path;

[0121] Select estimated power consumption Q j The inspection path corresponding to the minimum value is used as the re-planned UAV inspection path.

[0122] When all updates take advantage of efficiency All are greater than 100%, indicating that the estimated remaining range of all drones at this moment is less than the theoretical expression value. That is to say, according to the optimal inspection path at this time, no drone can complete the inspection task. Based on this, in one embodiment of the present invention, when no drone can complete the inspection task after obtaining the optimal inspection path with the minimum total travel distance, the theoretical expression value corresponding to the optimal inspection path calculated at this time is Considering the impact of the headwind segment on the entire journey, the optimal inspection path may not be the theoretical inspection path with the smallest total journey. In this case, it is necessary to re-obtain the inspection path and calculate the estimated power consumption Q of the multiple inspection paths planned by the path planning module. j , and then obtain the estimated power consumption Q j Theoretical expression value of the inspection path corresponding to the minimum value

[0123] According to the formula:

[0124]

[0125] Exclude update utilization efficiencies greater than 100% The corresponding drone;

[0126] Then select the one with the greatest update efficiency The corresponding drone is used as an inspection drone.

[0127] A photovoltaic power station intelligent inspection method based on deep learning, comprising the following steps:

[0128] S1: Obtain the optimal inspection path;

[0129] S2: Select a drone to perform the task corresponding to the optimal inspection path;

[0130] S3: take images of photovoltaic panels;

[0131] S4: Analyze the photovoltaic panel image;

[0132] S5: Process the image to obtain the photovoltaic panel number corresponding to the defective image.

[0133] Working principle: Through the power and travel management module, the remaining power of the drone is monitored and the estimated remaining travel of the drone is displayed. When the path planning module plans the optimal inspection path, it can be compared with the estimated remaining travel of the current drone. If the current estimated remaining travel of the drone is less than the optimal inspection path planned by the path planning module, the drone needs to be recharged or replaced for inspection. Through the above method, it can be ensured that the drone can fully photograph all photovoltaic panels in the photovoltaic power station during an inspection mission, avoiding the situation of recharging in the middle, and thus avoiding the time interval caused by recharging, which may lead to defects in the photovoltaic panels in the area where the shooting has been completed.

[0134] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A photovoltaic power station intelligent inspection system based on deep learning, characterized in that: include: Drones, used to conduct inspections and take images; Data acquisition module, used to receive the images and time taken by the drone; Digital photovoltaic power station module, which establishes corresponding virtual simulation scene images and semantic instance images based on the actual photovoltaic power station; The image processing module is used to process the captured images and analyze the captured images into the required inspection results, including the shooting time, altitude, longitude and latitude, defect type, defect location and photovoltaic panel number; A deep learning algorithm module provides support for the image processing module, including a defect detection algorithm and an image matching algorithm. The defect detection algorithm is used to detect various defects in the image, and the image matching algorithm is used to match the virtual simulation scene image of the digital photovoltaic power station module with the actual captured image; The data storage module is used to store and manage all inspection data of the photovoltaic power station, including the original inspection data and the parsed inspection results, and supports the query of the inspection results; It also includes a path planning module for planning the optimal inspection path for drones; It also includes a power and trip management module to monitor the remaining power of the drone and the estimated remaining trip; The matching module selects drones to conduct inspections along the optimal inspection path.

2. According to claim 1, a photovoltaic power station intelligent inspection system based on deep learning is characterized in that: The method for the path planning module to plan the optimal inspection path with the minimum total travel distance is: Use Dijkstra algorithm to draw multiple inspection paths passing through N points; Calculate the total travel distance of multiple inspection paths; The inspection path corresponding to the minimum total travel distance is selected as the optimal inspection path.

3. According to the deep learning-based intelligent inspection system for photovoltaic power stations of claim 2, it is characterized by: The matching module also includes a calculation unit for calculating the utilization efficiency of the drone, the method being: Get the estimated remaining distance L of multiple drones i And the total travel distance L of the optimal inspection path z ; According to the formula: M i =L z / L i ×100% Among them, M i represents the utilization efficiency of any UAV, and i represents any UAV.

4. According to claim 3, a photovoltaic power station intelligent inspection system based on deep learning is characterized in that: The remaining battery power of the drone is Q i The estimated remaining distance L i The relationship satisfies the formula: THE i =Q i ·And -γt TO i ·60 Among them, Q i Indicates the remaining battery capacity of any drone, in mAh; A indicates the average current consumption per unit distance of any drone, in mA / km; e -γt represents the attenuation parameter of the UAV battery over the service time, Q i ·e -γt Indicates the battery update capacity of any drone.

5. The photovoltaic power station intelligent inspection system based on deep learning according to claim 4 is characterized by: It also includes an analysis module for analyzing the composition of the optimal inspection path, including the upwind section and the downwind section; the calculation unit is also used to calculate the estimated power consumption Q of the optimal inspection path z ; Among them, the downwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is less than 90°, and the headwind segment trip refers to the trip where the angle between the drone's moving direction and the wind direction is greater than or equal to 90°.

6. The photovoltaic power station intelligent inspection system based on deep learning according to claim 5 is characterized by: Calculate the estimated power consumption Q of the optimal inspection path z The method is: Get the upwind segment travel L n 、Downwind section L s And the average current consumption of the downwind section is A s ; According to the formula: Q z =(L s +kL n )·A s ·60 Among them, Q z Indicates the estimated power consumption, L s represents the downwind segment in the optimal inspection path, L n represents the upwind section in the optimal inspection path, k represents the average current consumption factor of the upwind section compared with the downwind section, A s Indicates the average current consumption during the downwind leg of the journey.

7. The photovoltaic power station intelligent inspection system based on deep learning according to claim 6 is characterized by: Based on the estimated power consumption Q z The calculation module is also used to update the UAV utilization efficiency by: According to the formula: Calculate the theoretical expression value of the optimal inspection path Theoretical expression value Substitute into the formula: M i =L z / L i ×100% The update formula is: Calculate the updated utilization efficiency corresponding to multiple drones 8. The photovoltaic power station intelligent inspection system based on deep learning according to claim 7 is characterized by: The matching module updates the utilization efficiency based on the UAV The method to match any drone is: According to the calculated updated utilization efficiency corresponding to multiple drones Exclude update utilization efficiencies greater than 100% The corresponding drone; Then select the one with the greatest update efficiency The corresponding drone is used as an inspection drone.

9. The photovoltaic power station intelligent inspection system based on deep learning according to claim 8, characterized in that: When all updates take advantage of efficiency If both are greater than 100%, the path planning module is also used to replan the optimal inspection path with the lowest estimated power consumption, the method is: Obtain multiple inspection paths drawn by the Dijkstra algorithm; Use the analysis module to analyze the composition of the inspection path; Calculate the estimated power consumption Q of multiple inspection paths j , where j represents any inspection path; Select estimated power consumption Q j The inspection path corresponding to the minimum value is used as the re-planned UAV inspection path.

10. A photovoltaic power station intelligent inspection method based on deep learning, characterized in that: The following steps are involved: S1: Obtain the optimal inspection path; S2: Select a drone to perform the task corresponding to the optimal inspection path; S3: take images of photovoltaic panels; S4: Analyze the photovoltaic panel image; S5: Process the image to obtain the photovoltaic panel number corresponding to the defective image.

Images