CN115642877A

CN115642877A - Photovoltaic module abnormal shielding detection method and system based on deep learning

Info

Publication number: CN115642877A
Application number: CN202211379501.6A
Authority: CN
Inventors: 朱晓邦; 夏海洋; 彭合娟; 李嘉; 江春梅; 张瑜; 杨波; 雷蕾
Original assignee: Chongqing Zhongdian Self Energy Technology Co ltd
Current assignee: Chongqing Zhongdian Self Energy Technology Co ltd
Priority date: 2022-11-04
Filing date: 2022-11-04
Publication date: 2023-01-24

Abstract

The invention provides a photovoltaic string power loss calculation method, which comprises the following steps of obtaining a photovoltaic panel surface image; constructing a shelter detection model, inputting a surface image into the target detection model, and identifying a shelter; calculating the surface area of the shelter in the photovoltaic panel; and calculating the power generation data of the photovoltaic panel, and measuring and calculating the loss power generation. The invention also provides a photovoltaic module abnormal shielding detection system which comprises an image acquisition module, a data storage module, a data analysis module and a loss calculation module; the image acquisition module is used for acquiring a photovoltaic string surface image; the data storage module is used for storing surface image information, and the data analysis module is used for analyzing the obstruction in the surface image; the loss calculation module is used for calculating the power generation amount loss caused by the obstruction. The method solves the problem that the obstruction identification in the prior art cannot actually reflect the influence of the obstruction on the generating capacity.

Description

Photovoltaic module abnormal shielding detection method and system based on deep learning

Technical Field

The invention relates to the field of photovoltaic module monitoring, in particular to a photovoltaic module abnormal shielding detection method and a detection system.

Background

The energy of photovoltaic power generation is directly from the irradiation of sunlight, the solar irradiation on the earth surface is greatly influenced by the weather, and the power generation state of the system can be seriously influenced in long-term rainy and snowy days, cloudy days and foggy days. In addition, it is prominent that particles (such as dust) or other falling objects (such as bird droppings and leaves) in the environment sink on the surface of the solar cell module, and there may exist tree and building shadows in some distributed photovoltaic systems to block part of the light, and these abnormal shelters may reduce the conversion efficiency of the solar cell module, thereby causing a reduction in the power generation and even damage to the cell panel. Through research, bird droppings, shadows and other shielding problems become one of the most main problems influencing the photovoltaic performance. Shading can reduce the electrical energy generated by the photovoltaic array and cause potential safety hazards. When a portion of a photovoltaic panel is shaded, its shaded cell will become reverse biased and exist in the circuit as an energy consuming load. When the density of the accumulated dust is higher, the short-circuit current, the open-circuit voltage and the power generation efficiency are respectively greatly reduced. If the free shielding exists at the moment, the problems of hot spots and the like occur slightly, the service life of the photovoltaic module is shortened, and the fire disaster is caused seriously, so that the personal and property safety is harmed. Therefore, the identification of the abnormal shelter of the photovoltaic module is one of the problems to be solved urgently in the operation and maintenance of the photovoltaic power station.

At present, the operation and maintenance of a photovoltaic power station mainly carries out fault identification by collecting images through an unmanned aerial vehicle, but the traditional identification method cannot accurately identify the obstruction information. In the aspect of AI intelligent identification, the YOLO deep learning algorithm has good performance in the aspect of automatic identification of objects through multi-generation evolution. However, in the process of collecting the photovoltaic module image for unmanned aerial vehicle inspection, due to the influences of factors such as shooting distance and angle, different shelters may appear in the image in different shapes and sizes, and an improved identification algorithm is not adopted, so that the accuracy is not high, the detection rate is low, and the operation and maintenance requirements of a photovoltaic field station cannot be met. In contrast, chinese patent publication No. CN114648708A proposes a method and an apparatus for detecting a state of a photovoltaic module, where an initial detection model is constructed based on a YOLO target detection algorithm in combination with a feature pyramid network, and an identification algorithm of a target identification module is optimized through a deep learning algorithm to detect a type of an obstruction and the obstruction position, so as to generate an abnormal state detection result of the photovoltaic module. However, the influence of the shelter on the photovoltaic module cannot be reflected in multiple directions, and the operation and maintenance requirements of the photovoltaic station cannot be met.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a photovoltaic module abnormal shielding detection method and system based on deep learning, and solves the problem that the influence of a shielding object on the power generation amount cannot be reflected really in the identification of the shielding object in the prior art.

In order to achieve the purpose, the invention adopts the following technical scheme:

a photovoltaic string electric quantity loss calculation method comprises the following steps of obtaining a photovoltaic panel surface image; constructing a shelter detection model, inputting the surface image into the target detection model, and identifying the shelter; calculating the surface area of the shelter in the photovoltaic panel; and calculating the power generation data of the photovoltaic panel, and calculating the loss power generation amount.

The invention also provides a photovoltaic module abnormal shielding detection system which comprises an image acquisition module, a data storage module, a data analysis module and a loss calculation module; the image acquisition module is used for acquiring a surface image of the photovoltaic group string; the data storage module is used for storing surface image information, and the data analysis module is used for analyzing the obstruction in the surface image; the loss calculation module is used for calculating the power generation amount loss caused by the shelters.

Compared with the prior art, the invention has the following beneficial effects: according to the method, based on a depth recognition algorithm, the area of the shelters is calculated while the shelters are recognized, so that the power generation loss is obtained, and more accurate and comprehensive decision information is provided for operation and maintenance of the photovoltaic module.

Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention.

Drawings

FIG. 1 illustrates a process for detecting an obstruction in accordance with the present invention;

FIG. 2 is a power loss calculation process;

fig. 3 is a schematic diagram of a system configuration according to an embodiment of the present invention.

Detailed Description

In order to make the technical means, the creation features, the achievement purposes and the functions of the invention clearer and easier to understand, the technical scheme of the invention is further explained in the following with the accompanying drawings and the embodiments.

As shown in fig. 1-2, the invention provides a method for calculating the power loss of a photovoltaic string, which comprises the following steps: acquiring a surface image of a photovoltaic panel; the acquisition of the photovoltaic panel surface image is usually achieved by unmanned aerial vehicles and cameras on site. And preprocessing the surface image before inputting the surface image into the target detection model, wherein the preprocessing comprises the processing of adjusting the brightness and the contrast to a preset degree, denoising and the like. And (3) constructing a shelter detection model, wherein the shelter detection model can directly adopt a model in the prior art, and can also be obtained by continuously training the collection of different shelter shielding pictures based on a deep learning method. And inputting the surface image into an obtained target detection model, and identifying the obstruction.

Calculating the surface area of the shelter in the photovoltaic panel; the shelter occupies the surface area of the photovoltaic panel, can be measured and calculated through model identification data, and can also be obtained through directly performing image processing on the basis of the obtained surface image. After the surface area or the surface area ratio of the shelter is obtained, the power generation data of the photovoltaic panel can be calculated according to the surface area of the photovoltaic panel, and then the power generation loss is calculated to provide decision data for operation and maintenance.

In one embodiment, the loss electric quantity is automatically calculated by constructing an electric quantity loss model, and the loss electric quantity is calculated according to the surface area of the shelter and the illumination intensity. The power generation is generally influenced by the intensity of the light, the temperature and the wind speed, among which the intensity of the light plays a major decisive role. The illumination intensity is the key data of the photovoltaic power station for real-time monitoring. After the illumination intensity and the shielding area are known, the loss electric quantity can be calculated through various ways, and a person skilled in the art can select a proper formula to construct an electric quantity loss model for automatic measurement and calculation, such as a × B × k = Q, wherein a is the illumination intensity, B is the illumination area of the photovoltaic panel, k is a constant, and Q is the generated electric quantity.

In another embodiment, the obstruction information identified by the target detection model includes the type of obstruction and the location of the obstruction. The skilled person in the art trains an image recognition model to obtain a target monitoring model by manually classifying and indexing a plurality of groups of obstruction reference images according to the actual type of the obstruction. The position of the obstruction can then be determined by the anchor frame position. After the category and the position of the shielding object are obtained, the shielding object can be used as an influence factor of the electric quantity loss model, so that the result is more accurate.

In another embodiment, when the loss power generation amount is calculated, the percentage of the actual illumination area of the photovoltaic panel in the area of the photovoltaic panel is calculated, and the input illumination intensity is converted into the actual illumination intensity multiplied by the percentage to obtain the power generation amount of the photovoltaic panel. For a photovoltaic power station, an electric quantity prediction model is usually provided to help the photovoltaic power station to predict the generated energy, the model is learned and optimized more perfectly through more data accumulation, but when the illumination area is calculated, the area of a photovoltaic panel is used for measurement and calculation, so the illumination area in the model is not used as input, the illumination intensity is monitored in real time and is necessary input information.

In one embodiment, the power generation data is obtained by direct processing of the surface image, including the steps of converting the surface image color space to RGB to HVS color space for subsequent processing. And then threshold value binarization processing is carried out on the converted picture, the surface of a normal photovoltaic panel is relatively smooth, the surface of the normal photovoltaic panel has relatively large difference from the surface of the abnormally shielded photovoltaic panel, the normal photovoltaic panel is relatively easy to distinguish through setting a threshold value, the surface image after binarization processing only has two colors, for example, a shielding object part is black, and a non-shielding part is white, so that the contour extraction and the area calculation of a computer are facilitated.

In order to implement the above scheme, as shown in fig. 3, the invention further provides a photovoltaic module abnormal shielding detection system, which includes a main control module, a shielding identification module, and a lost electric quantity operation module; the main control module is composed of a picture acquisition module, an interface display module and a data storage module, wherein the picture acquisition module mainly comprises an unmanned aerial vehicle, a field camera and the like and is used for acquiring the surface image of the photovoltaic panel. The interface display module is mainly composed of display screens of mobile phones, computers and the like and is used for displaying and analyzing related data so as to facilitate decision making. The data storage module is used for storing initially obtained surface image data, model data and result data after analysis processing. The shielding identification module is mainly used for identifying information such as shielding objects, shielding positions and shielding areas, and the loss electric quantity operation module is mainly used for obtaining and calculating loss electric quantity through the shielding areas and other power generation information.

It is worth explaining that the method calculates the shelters, the sheltering positions and the sheltering areas through the models, can be directly used as decision information to be provided for operation and maintenance personnel to carry out operation and maintenance decisions, and can also be used as a loss power generation amount calculation consideration factor to obtain the actual power generation amount influence degree and provide a main decision reference for the operation and maintenance personnel; and transverse comparison of various shielding factors can be performed, the influence degree of the various shielding factors can be obtained, and more comprehensive reference can be given. In addition, the risk degree of factors such as bird droppings, dust accumulation, fallen leaf shielding and the like which cause additional heat loss, equipment damage and potential safety hazards can be obtained in comparison of all factors.

Finally, the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered in the claims of the present invention.

Claims

1. A photovoltaic module abnormal occlusion detection method based on deep learning is characterized by comprising the following steps:

acquiring a surface image of the photovoltaic panel;

constructing a shelter detection model, inputting the surface image into the target detection model, and identifying the shelter; calculating the surface area of the shelter in the photovoltaic panel;

and calculating the power generation data of the photovoltaic panel, and calculating the loss power generation amount.

2. The abnormal occlusion detection method of claim 1, wherein a power loss model is constructed, and a power loss amount is calculated based on the surface area of the occlusion object and the illumination intensity.

3. The abnormal occlusion detection method of claim 2, wherein the obstruction information identified by the target detection model includes a type of obstruction and a location of the obstruction.

4. The abnormal occlusion detection method of claim 3, wherein classification indexing is performed on several types of occlusion images manually while training the target detection model.

5. The abnormal occlusion detection method of claim 3, wherein a type of an occluding object and a position of the occluding object are used as influencing factors when the electric quantity loss model is constructed.

6. The abnormal occlusion detection method of claim 2, wherein when calculating the power generation loss, first calculating a percentage of an actual illumination area of the photovoltaic panel to the area of the photovoltaic panel, and multiplying the actual illumination intensity by the percentage to obtain the power generation loss of the photovoltaic panel.

7. The abnormal occlusion detection method of claim 1, wherein the surface image is preprocessed before being input into the target detection model, the preprocessing including adjusting brightness, contrast to a predetermined degree, and denoising.

8. The abnormal occlusion detection method of claim 1, wherein calculating an obstruction surface area comprises the steps of: converting the surface image color space, converting RGB into HVS color space, performing threshold value binarization processing, extracting the surface contour of the photovoltaic panel and the contour of the shielding object, and calculating the surface area of the photovoltaic panel occupied by the shielding object.

9. The photovoltaic module abnormal shielding detection system is characterized by comprising an image acquisition module, a data storage module, a data analysis module and a loss calculation module; the image acquisition module is used for acquiring a photovoltaic string surface image; the data storage module is used for storing surface image information, and the data analysis module is used for analyzing the obstruction in the surface image; the loss calculation module is used for calculating the power generation amount loss caused by the shelters.

10. The abnormal occlusion detection system of claim 9, further comprising a display module for displaying occlusion type and power loss.