CN-122026809-A - Offshore photovoltaic panel pollution state sensing method and system

CN122026809ACN 122026809 ACN122026809 ACN 122026809ACN-122026809-A

Abstract

The invention belongs to the technical field of photovoltaic detection, and provides a method and a system for sensing the pollution state of an offshore photovoltaic panel, which are used for solving the problem of inaccurate detection of the existing photovoltaic panel, inputting an offshore photovoltaic panel high-altitude image into a pollution coarse-granularity detection model to obtain the pollution probability, pollution coverage proportion and pollution level of the photovoltaic panel, and screening the photovoltaic panel with the pollution severity level to generate a suspected high-pollution target list; and acquiring and splicing the near RGB image and depth information of the high pollution target, inputting spliced data into a pollution fine granularity detection model to obtain a pollution area fine mask, a pollution type and a pixel level thickness map, and generating a cleaning strategy based on the pollution type, the fine coverage proportion and the thickness level to further realize cleaning. The invention can not only perform large-scale rapid pollution detection, but also provide high-precision pollution type and thickness information, and comprehensively improve the pollution monitoring capability.

Inventors

DUAN PEIYONG
CHANG SHULEI
Han Fanbing
FU QIANG
WU JIADONG
NING CHENGUANG
LIU LIXIA
XU LANGJUN

Assignees

齐鲁工业大学(山东省科学院)

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method for sensing a pollution state of an offshore photovoltaic panel, comprising: Inputting a pollution coarse granularity detection model to the collected high-altitude images of the offshore photovoltaic panels to obtain pollution probability, pollution coverage proportion and pollution level of the photovoltaic panels, and screening the photovoltaic panels with pollution severity level to generate a suspected high-pollution target list; planning a routing inspection path according to a suspected high-pollution target list, controlling an unmanned aerial vehicle to acquire a near RGB image and depth information at a target photovoltaic plate hovering point, splicing the near RGB image and the depth information, and inputting a pollution fine granularity detection model to obtain a pollution region fine mask, a pollution type and a pixel level thickness map; Fitting a photovoltaic panel reference plane, calculating depth difference of a polluted area, mapping the depth difference into physical thickness, counting component-level average thickness and thickness level, generating a cleaning strategy based on pollution type, fine coverage proportion and thickness level, and controlling the unmanned aerial vehicle to execute a cleaning task according to the cleaning strategy.
2. The method for sensing the pollution state of the offshore photovoltaic panel according to claim 1, wherein the pollution coarse-grained detection model adopts an improved YOLO network, and scalar regression branches and multiple classification branches are newly added on the basis of an original YOLO network detection head, wherein the scalar regression branches are used for predicting pollution coverage proportions, and the multiple classification branches are used for predicting pollution levels.
3. The method for sensing the pollution state of the offshore photovoltaic panel according to claim 1, further comprising preprocessing the acquired near RGB image and depth information, wherein the preprocessing is to uniformly scale the size of the near RGB image and normalize the size of the near RGB image, and aligning the depth information with the near RGB image on pixel coordinates, and clipping and interpolating abnormal values.
4. The method for sensing the pollution state of the offshore photovoltaic panel according to claim 1, wherein the pollution fine granularity detection model is trained by adopting a multi-task joint loss function, wherein the multi-task joint loss function comprises segmentation loss, regression loss of a pixel-level thickness map and cross entropy loss of a component-level thickness level.
5. An offshore photovoltaic panel pollution state sensing method according to claim 1, wherein the depth difference of the pollution area is calculated by fitting a photovoltaic panel reference plane and mapped to a physical thickness, specifically: dividing the photovoltaic panel area into a clean area and a pollution area by utilizing a fine binary mask of the pollution area; fitting a reference plane obtained by a plane model on depth data corresponding to the clean area by adopting a least square method; And calculating the difference value between the actual depth and the reference plane for the pixels in the polluted area, and mapping the height residual error into physical thickness to obtain a pixel-level thickness map.
6. The method for sensing the pollution state of the offshore photovoltaic panel according to claim 1, further comprising the steps of collecting close-range images at the same position of the photovoltaic panel after cleaning is completed, detecting by using a pollution fine granularity detection model to obtain a pollution coverage proportion and a thickness grade after cleaning, and determining the cleaning effect according to the pollution coverage proportion and the thickness grade after cleaning.
7. An offshore photovoltaic panel pollution state sensing system, comprising: The coarse granularity detection module is configured to input a pollution coarse granularity detection model to the collected high-altitude images of the offshore photovoltaic panels to obtain pollution probability, pollution coverage proportion and pollution level of the photovoltaic panels, and screen the photovoltaic panels with pollution severity level to generate a suspected high-pollution target list; The fine granularity detection module is configured to plan a routing inspection path according to a suspected high-pollution target list, control the unmanned aerial vehicle to acquire a near RGB image and depth information at a target photovoltaic board hovering point, splice the near RGB image and the depth information, and input a pollution fine granularity detection model to obtain a pollution area fine mask, a pollution type and a pixel level thickness map; the cleaning module is configured to fit a photovoltaic panel reference plane, calculate a depth difference of a polluted area and map the depth difference into physical thickness, count an average thickness and a thickness grade of a component level, generate a cleaning strategy based on a pollution type, a fine coverage proportion and the thickness grade, and control the unmanned aerial vehicle to execute a cleaning task according to the cleaning strategy.
8. An electronic device comprising a memory and a processor and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the method of any one of claims 1-6.
9. A computer readable storage medium storing computer instructions which, when executed by a processor, perform the method of any of claims 1-6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the method of any of claims 1-6.

Description

Offshore photovoltaic panel pollution state sensing method and system Technical Field The invention belongs to the technical field of photovoltaic detection, and particularly relates to a method and a system for sensing pollution state of an offshore photovoltaic panel. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. With the increasing global demand for renewable energy sources, offshore photovoltaic power generation has received a great deal of attention as an important green energy source. In the construction and operation process of an offshore photovoltaic power station, the pollution problem on the surface of a photovoltaic panel becomes one of key factors influencing the photovoltaic power generation efficiency. Due to the special offshore environment, sea salt, bird droppings, greasy dirt, algae and other pollutants are easy to adhere to the surface of the photovoltaic panel due to the factors of sea water evaporation, high air humidity and the like, so that the photovoltaic conversion efficiency of the photovoltaic panel is reduced, and even the long-term stability of the photovoltaic panel is influenced. The existing pollution detection technology mainly integrates a plurality of modes such as manual inspection, a traditional image recognition technology, an inspection system based on an unmanned aerial vehicle and the like, and although some existing methods realize automatic monitoring, the existing detection method still has the following problems that whether the photovoltaic panel is polluted or not can only be roughly judged, and the pollution type and the pollution degree are not accurately evaluated. Especially in large area offshore photovoltaic power plants, coarse particle size detection often fails to meet the requirements for accurate cleaning. Moreover, the thickness of the pollutant has a decisive influence on the selection of the cleaning mode, and the current technology cannot accurately evaluate the thickness of the pollutant, so that the cleaning strategy cannot be accurately matched with the actual condition of the pollution. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides a method and a system for sensing the pollution state of an offshore photovoltaic panel, which can rapidly detect pollution in a large range, provide high-precision pollution type and thickness information and comprehensively improve pollution monitoring capability. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, the present invention provides a method for sensing pollution status of an offshore photovoltaic panel, comprising: Inputting a pollution coarse granularity detection model to the collected high-altitude images of the offshore photovoltaic panels to obtain pollution probability, pollution coverage proportion and pollution level of the photovoltaic panels, and screening the photovoltaic panels with pollution severity level to generate a suspected high-pollution target list; planning a routing inspection path according to a suspected high-pollution target list, controlling an unmanned aerial vehicle to acquire a near RGB image and depth information at a target photovoltaic plate hovering point, splicing the near RGB image and the depth information, and inputting a pollution fine granularity detection model to obtain a pollution region fine mask, a pollution type and a pixel level thickness map; Fitting a photovoltaic panel reference plane, calculating depth difference of a polluted area, mapping the depth difference into physical thickness, counting component-level average thickness and thickness level, generating a cleaning strategy based on pollution type, fine coverage proportion and thickness level, and controlling the unmanned aerial vehicle to execute a cleaning task according to the cleaning strategy. In a second aspect, the present invention provides an offshore photovoltaic panel pollution status sensing system comprising: The coarse granularity detection module is configured to input a pollution coarse granularity detection model to the collected high-altitude images of the offshore photovoltaic panels to obtain pollution probability, pollution coverage proportion and pollution level of the photovoltaic panels, and screen the photovoltaic panels with pollution severity level to generate a suspected high-pollution target list; The fine granularity detection module is configured to plan a routing inspection path according to a suspected high-pollution target list, control the unmanned aerial vehicle to acquire a near RGB image and depth information at a target photovoltaic board hovering point, splice the near RGB image and the depth information, and input a pollution fine granularity detection model to obtain a pollution area fine mask, a pollution type and a pixel