TW202213939A - An intelligent diagnosis system and method for defects of solar power module - Google Patents

Abstract

The invention discloses an intelligent diagnosis system for defects of solar power module, which comprises: a first image capture module; a second image capture module; an image processing module; a deep learning module; and an information judgment module. The invention further discloses an intelligently diagnosing method for defects of solar power modules. Through the two-stage image depth training, the present invention can quickly and accurately identify the position and type of defects of the solar power module.

Description

Intelligent diagnosis system and method for solar power module defects

The present invention relates to the diagnosis of solar power generation modules, and more particularly, to an intelligent diagnosis system and method for applying deep learning to train solar power generation module defects.

Solar power generation has the following advantages: the system is easy to use for a long time, limited by the environment and geography, has a wide range of applications, can be combined with buildings, is easy to popularize, the power generation period changes with the intensity of sunlight, is helpful for suppressing peak power consumption, and is safe and free. The advantages of pollution, no noise, etc. have been widely installed in many countries around the world. The benefit of solar power generation is related to the scale of the installation. In order to effectively increase the overall power generation of a solar power plant, users usually lay panels on a large scale. However, solar modules can be damaged by weather, temperature changes, air pollution and UV rays. At present, the life of solar modules is at least 25 years. During operation, maintenance is the key to whether the life of the modules can reach the design value. The longer the system can be used and the more electricity it can generate, the cost per watt can be reduced on average.

At present, all solar photovoltaic system design and installation manufacturers are from evaluation, design planning, subsidies and loans, application for submission to installation and construction. Module maintenance and remote monitoring mostly need to find other cleaning companies and professional monitoring companies. Power station installation and power station operation and maintenance are two very important service-oriented links. The good operation of these two parts ensures the quality of the power station and helps the installation owner to obtain the maximum benefit.

Current system manufacturers' understanding of solar power module defect analysis technology Not many, resulting in a large number of failures after the operation of the power station, resulting in a lot of loss of manpower and material resources. In addition, due to the large area of the solar power plant, the traditional manual inspection has the disadvantages of difficulty in locating abnormal solar elements, low work efficiency, and high operation and maintenance costs.

Although some manufacturers now use drone aerial photography technology to perform inspection work, hoping to reduce maintenance costs, the drone detection method can only be said to be "quick screening", which only uses infrared (Infrared, IR) thermal cameras. Check whether the temperature of the solar power module is too high. This test can only detect the generalized hot spot phenomenon (that may be caused by the internal resistance and the dark current of the cell itself). This information can only evaluate the possibility of defects, but cannot accurately determine the defect processing method. In order to accurately grasp the operation efficiency and health status of solar power modules, occasional detection and quick inspection are not enough. Anytime, anywhere and remote monitoring may be the best way to maintain the normal operation of solar power modules.

In order to clarify whether the degree of defect is serious and needs immediate improvement, it is usually necessary to use other tools for further inspection, such as electroluminescence (EL) inspection to see cracks or finer problems. However, the determination of the defect mode of a solar power generation module relies on experience, subjective factors in the visual diagnosis process, the difficulty of quantifying the determination mode, and the time-consuming determination of a large amount of data.

In view of the above problems, it is necessary to propose a new solar power generation module defect assessment system and method to solve the above problems.

The main purpose of the present invention is to provide an intelligent diagnosis system for solar power module defects. The system can be used for large-scale solar power generation projects, using IR/EL image information at the same time, through deep learning training, to diagnose and find defect patterns, and to prevent solar power generation system failures, reduce maintenance accidents and costs, and ensure a profitable model. .

Another object of the present invention is to provide an intelligent method for diagnosing defects of solar power generation modules. This technology can be used for large-scale solar power generation projects, using IR/EL image information at the same time, through deep learning training, to diagnose and find defect patterns, avoid power loss and accidents, and greatly increase the power generation of solar power plants.

In order to achieve the main purpose of the present invention, the present invention provides a solar power generation module defect intelligent diagnosis system, which is used for defect analysis of a plurality of solar power generation modules, comprising:

a first image capture module to obtain a first image data of the solar power generation modules;

a second image capturing module to obtain a second image data of the solar power generation modules;

an image processing module for performing image processing on the first image data and the second image data to obtain a first image feature data and a second image feature data respectively;

A deep learning module, firstly performs a first period of in-depth training on the first image feature data, and then performs a second period of in-depth training on the second image feature data, and the first period is less than the second instalment; and

An information judging module analyzes the defects of the solar power generation modules according to the result of the deep training of the second phase.

In order to achieve another object of the present invention, the present invention provides a method for diagnosing defects of solar power generation modules, which is used for defect analysis of a plurality of solar power generation modules, comprising:

Step 1: obtaining a first image data of the solar power generation modules;

Step 2: obtaining a second image data of the solar power modules;

Step 3: performing image processing on the first image data and the second image data to obtain a first image feature data and a second image feature data respectively;

Step 4: Perform a first-stage in-depth training on the first image feature data, and then perform a second-stage in-depth training on the second image feature data, and the first stage is less than the second stage. number of periods; and

Step 5: Analyze the defects of the solar power generation modules based on the results of the second phase of in-depth training.

The present invention can quickly and accurately identify the location and type of defects in the solar power generation module through the two-stage image depth training. In order to achieve the effect of rapid and large-scale data determination, the intelligent diagnosis system and method of the solar power generation module defect of the present invention is further matched with the deep learning module of artificial intelligence to perform stable image identification. In view of the goals that the first image data and the second image data can achieve, the present invention sets different training methods. First perform a first epoch of depth training on the first image feature data, and then perform a second epoch of depth training on the second image feature data, and the first epoch is less than The second period.

Moreover, the deep learning training method used by the intelligent diagnosis system and method for solar power module defects of the present invention includes a plurality of convolutional layer training, which is a progressive training method in which the convolutional layers are in the order of the bottom layer, the middle layer and the top layer. . First of all, the weights of all convolutional layers will not be frozen. After training for several periods, the convolutional layers at the bottom layer are frozen to a greater degree to train the convolutional layers in the middle layer; after training for a few more periods, the convolutional layers in the middle layer are also frozen. , to train the top convolutional layers. It should be noted that the depth training of the first phase is to perform bottom training and middle training on the first image feature data. The depth training of the second phase is to perform middle-level training and top-level training on the second image feature data.

The intelligent diagnosis system and method for solar power generation module defects of the present invention can assist upstream and downstream operators (system manufacturers, installers, banks, and insurance companies) to more professionally and easily find faults that appear in the installation, operation and maintenance of power plants. various problems. With the construction of solar power plants, the intelligent diagnosis operation and maintenance of defect mode will have great market space and promotion value.

5: Intelligent diagnosis system for solar power module defects

10: The first image capture module

20: Second image capture module

30: Image processing module

40: Deep Learning Module

50: Information Judgment Module

In order to make the above-mentioned and other objects, features, and advantages of the present invention more clearly understood, several preferred embodiments are hereinafter described in detail in conjunction with the accompanying drawings.

Fig. 1 is a schematic diagram of the intelligent diagnosis system for defects of solar power generation modules according to the present invention;

FIG. 2 is a flow chart of the intelligent diagnosis method for solar power module defects according to the present invention.

While the present invention may be embodied in various forms of embodiment, those shown in the drawings and described herein are preferred embodiments of the invention. Those skilled in the art will appreciate that the apparatus and methods specifically described herein and illustrated in the accompanying drawings are considered to be exemplary, non-limiting, exemplary embodiments of the present invention, and that the scope of the present invention is limited only by the scope of the claims be defined. Features illustrated or described in connection with one exemplary embodiment may be combined with features of other embodiments. Such modifications and variations are intended to be included within the scope of the present invention.

The intelligent diagnosis system and method for solar power generation module defects of the present invention can quickly and accurately identify the position and type of solar power generation module defects through two-stage image depth training of two types of images. In addition, in order to achieve the effect of rapid and large-scale data determination, the intelligent diagnosis system and method of the solar power generation module defect of the present invention is further matched with the deep learning module of artificial intelligence to perform stable image identification.

Please refer to FIG. 1, which is a solar power generation module defect intelligent diagnosis system 5 according to the present invention, which is used for defect analysis of a plurality of solar power generation modules, which includes: a first image capture module 10; a second An image capture module 20 ; an image processing module 30 ; a deep learning module 40 ; and an information judgment module 50 .

In a solar power module, each component, including batteries, ribbons, EVA glue, glass/backplane, junction boxes (including bypass diodes), and frames, may have defects. Defects of each component of the module may be: battery rupture or crack, broken connecting wire (ribbon, solder) Cracks, Encapsulant discoloration and Delamination of encapsulated EVA materials, Front and rear panels: glass/back panel cracks, junction box and bypass diode cracks, Light induced degradation, etc.

The first image capturing module 10 obtains a first image data of the solar power generation modules. The second image capturing module 20 obtains a second image data of the solar power generation modules. The image processing module 30 performs image processing on the first image data from the first image capture module 10 and the second image data from the second image capture module 20 to obtain a The first image feature data and a second image feature data. The deep learning module 40 first performs a first period of deep training on the first image feature data from the image processing module 30 , and then performs the second image feature data from the image processing module 30 . A second period of deep training, and the first period is less than the second period. The information judging module 50 analyzes the defects of the solar power generation modules according to the result of the deep training of the second phase from the deep learning module 40 .

Wherein, the first image data is an infrared (Infrared, IR) image, and the second image data is an electroluminescence (Electroluminescence, EL) image. The standard for establishing infrared (Infrared, IR) images is mainly based on the thermal image detection standard IEC62446-3.

The first image data of the solar power generation modules is obtained by photographing the solar power generation modules during daytime. The first image capturing module 10 for acquiring the first image data of the solar power modules includes: an infrared lens: receiving and concentrating the infrared radiation emitted by the measured object; an infrared detector component: converting the thermal radiation signal into electrical signals; electronic components: process electrical signals; display components: convert electrical signals into visible light images: capture software: process the collected temperature data and convert them into temperature readings and images.

The second image data of the solar power generation modules are obtained by photographing the solar power generation modules at night or with low illumination (illuminance<200mW/cm ² ). Preferably, the second image data of the solar power generation modules are obtained when the illumination is between 100 mW/cm ² and 200 mW/cm ² . The second image capture module 20 for obtaining the second image data of the solar power modules includes: a photographic lens: receiving and concentrating the fluorescent radiation emitted by the object to be measured; The optical radiation signal becomes an electrical signal; the electronic component: processes the electrical signal; the display component: converts the electrical signal into a visible light image: capture software: processes the collected temperature data and converts it into temperature readings and images. The second image data, especially the electroluminescence second image data, can detect cell and module defects that cannot be seen by traditional visual inspection or optical image measurement, including 1. cell fabrication defects, 2. surface mesh Print defect, 3. Cell crack defect, 4. Cell rupture defect, 5. Low-efficiency conduction region defect.

The specifications of the first image data and the second image data obtained from the solar power modules include: sensing resolution: 640×512 pixels; pixel size 17 μm; infrared image using infrared bandwidth: 7.5-13.5 μm shooting , the electroluminescence images were taken with visible light; the image frame rates (Full Frame Rates): 30 Hz (National Television System Committee, NTSC) and 25 Hz (Phase Alternating Line, PAL).

The infrared lens and the electroluminescence lens for obtaining the first image data and the second image data of the solar power modules are mainly installed on the unmanned aerial vehicle, and the unmanned aerial vehicle obtains the image above the solar power generation modules and transmits them to the unmanned aerial vehicle. The first image capturing module 10 and the second image capturing module 20 . The relevant specifications of the UAV are roughly as follows: GPS hovering accuracy: vertical: ±0.5m; horizontal: ±1.5m; maximum descent speed: vertical: 5m/s; maximum horizontal flight speed: 54km/h or 15m/s; remote control method: Adopt FASST 2.4 G automatic scanning frequency locking; working environment temperature: -30 ℃ to 45 ℃.

In the present invention, the image processing tools used: Adobe Photoshop, Aphelion, ImageJ, OpenCV, Ulead PhotoImpact or Rapidminer. Typical image processing procedures include: 1. Image representation and model establishment (Image Modeling), 2. Image Enhancement, 3. Image Restoration, 4. Image Analysis, Five, image reconstruction (Image Reconstruction), six, image data compression (Image Compression).

Before training, it is necessary to discriminate various defect modes on the image data obtained by infrared/electroluminescence, so as to obtain various features and corresponding defect modes. In the image of the present invention, the first image data and the second image data image are manually extracted from the first image data, and the image of the feature extracted image is processed by image cutting to define the infrared/electroluminescence image of the solar module. The identification region (Region-of-Interest, ROI). These initial data are used as reference data of the image processing module 30 . And the first image data and the second image data images, that is, the images obtained by infrared/electroluminescence, are manually marked for failure features, integrated analysis and discrimination of various defect modes are performed. The image data after preliminary analysis and processing is imported into the deep learning module 40 for commissioning and training, and the resolution quantification and testing of the training results derived from the model are performed.

The image processing module 30 performs image processing on the first image data from the first image capture module 10 and the second image data from the second image capture module 20 .

The image processing includes the following:

1. Image Pre-processing: Perform Image Pre-processing on the obtained IR/EL image of the solar module. The main methods are: image binarization ( Binarization) to obtain a binary image, and perform Edge Detection and Isolation/Enhancement on the binary image respectively. (Enhancement).

2. Feature extraction (Feature Extraction), the pre-processed image is subjected to infrared/electroluminescence image feature extraction (Feature Extraction) of the solar module, including but not limited to: repeated line tracking (repeated line tracking), maximum Maximum curvature, wide line detector and Gabor filter.

3. Recognition region (Region-of-Interest, ROI), the image of the feature extraction is processed by image cutting to define the recognition region (Region-of-Interest, ROI) of the infrared/electroluminescence image of the solar module.

4. Data Augmentation, perform Data Augmentation on the ROI image, and perform various transformations on the ROI image, such as image blur (Gaussian Blur), image sharpening (Sharpen), imitation Affine transform, Shake, Gaussian Noise, and Coarse Dropout are used to process the data to increase the data volume and diversity of the training data set, which can avoid Overfitting occurs in subsequent deep learning training stages.

5. Data pre-processing (Data Pre-Processing), normalization (Data Normalization), standardization (Standardization) and labeling (Labeling) of the data obtained after data enhancement processing, which can help in the high-dimensional feature space. the rate of decline.

6. Generate a dataset. The first image feature data and the second image feature data include: a training dataset, a validation dataset, and a testing dataset. Add a first label to the processed first image feature data, add a second label to the processed second image feature data, and input the subsequent deep learning module 40 for training and performance evaluation and prediction. Classification.

Generally speaking, in the training of deep learning, the underlying features are relatively low-level and general. Top-level features are more high-order special, and high-order special features are closer to fully connected layers. It is therefore reasonable to speculate that neural networks are mainly classified by high-order features. The high-level features are formed from the middle-level features, and the middle-level features are formed from the low-level features. Therefore, it can be speculated that if the low- and medium-level features change too drastically, the high-level features will be difficult to form. The freezing allows the training of the neural network to focus on forming higher-order features, which is supposed to speed up the training of deep learning.

Therefore, the training strategy proposed by the deep learning module 40 of the present invention will be trained in two stages during the training process. Due to the first image data, especially the infrared image, the position of the possible defect of the solar power generation module can be quickly and easily known. The present invention uses the second image data, especially the electroluminescence image, to further distinguish the type of defect mode of the solar power module at the position where the possible defects of the solar power module are known. In view of the purpose that the first image data and the second image data can achieve, that is, to identify the position of the solar power module defect pattern and the type of the solar power module defect pattern, the present invention sets different training methods for different periods. . Therefore, the deep learning module 40 first performs a first epoch of depth training on the first image feature data, and then performs a second epoch of depth training on the second image feature data , and the number of the first phase is less than the number of the second phase.

In addition, in the deep learning module 40, the deep training includes a plurality of convolutional layer trainings, and the convolutional layer trainings are performed in a sequence of bottom-level training, middle-level training and top-level training. The depth training of the first phase is to perform bottom training and middle training on the first image feature data. Moreover, the depth training of the second phase is to perform middle-level training and top-level training on the second image feature data. That is, the first image feature data does not need to be top-level training, and the second image feature data is trained based on the first image feature data, so bottom-level training is not required.

The information judgment module 50 uses the second data from the deep learning module 40 The results of the in-depth training of the phases analyze the defects of these solar power generation modules.

Please refer to FIG. 2, which is an intelligent diagnosis method for solar power module defects of the present invention, which is used for defect analysis of a plurality of solar power modules, including:

Step 1: obtaining a first image data of the solar power generation modules;

Step 2: obtaining a second image data of the solar power modules;

Step 3: performing image processing on the first image data and the second image data to obtain a first image feature data and a second image feature data respectively;

Step 4: Perform a first-stage in-depth training on the first image feature data, and then perform a second-stage in-depth training on the second image feature data, and the first stage is less than the second stage. number of periods; and

Step 5: Analyze the defect analysis of the solar power generation modules according to the result of the second-stage in-depth training.

Wherein, the first image data is an infrared image, and the second image data is an electroluminescence image. The depth training includes training of a plurality of convolutional layers, and the trainings of the convolutional layers are performed in a sequence of bottom training, middle training and top training. The depth training of the first phase is to perform bottom training and middle training on the first image feature data. The depth training of the second phase is to perform middle-level training and top-level training on the second image feature data.

In step 4, the part of repeated training will be divided into two types. One is the number of recursive correction parameters in the model. This part is that the more recursive times are set, the longer the model will take to complete the training. The other is training with different numbers of samples or test models as the samples accumulate during the test model phase. Both parts will gradually improve the accuracy with the number of repetitions.

In step 4, two deep convolutional neural network architectures are used: ResNet-50 (Residual Network, residual network) and DenseNet (Dense Convolutional Network) Network, dense convolutional neural network) for training as a validation case.

In step 4, the weights of all convolutional layers will not be frozen first. After training for several periods, the bottom convolutional layers will be frozen to train the convolutional layers of the middle layer. After training for a few more epochs, the middle convolutional layers are also frozen to train the top convolutional layers. The basic concept is that the middle network is trained by the front network, the latter network is trained by the middle network, and finally the training method of the classifier is trained.

Due to the first image data, especially the infrared image, the position of possible defects in the solar power generation module can be quickly and easily known, but the types of defects cannot be accurately known. Therefore, the present invention uses the second image data, especially the electroluminescence image, to further distinguish the type of defect mode of the solar power module at the position where the possible defects of the solar power module are known. That is, the present invention can quickly and accurately identify the location and type of defects in the solar power generation module through the two-stage image depth training.

In addition, in order to achieve the effect of rapid and large-scale data determination, the intelligent diagnosis system and method of the solar power generation module defect of the present invention is further matched with the deep learning module of artificial intelligence to perform stable image identification. Different from the training method of the traditional deep learning module, various data are trained for a certain period of time. In view of the purpose that the first image data and the second image data can achieve, the present invention is to identify the solar power generation module respectively. The location of the defect mode and the type of defect mode of the solar power generation module, therefore, set different training methods for the number of stages. First perform a first epoch of depth training on the first image feature data, and then perform a second epoch of depth training on the second image feature data, and the first epoch is less than The second period.

Moreover, the deep learning training method used by the intelligent diagnosis system and method for solar power module defects of the present invention is a progressive training method of convolutional layers in the order of bottom layer, middle layer and top layer. First, the weights of all convolutional layers will not be frozen. After training for several periods, the The lower convolutional layers are frozen to a greater extent to train the convolutional layers of the middle layer; after training for several more epochs, the convolutional layers of the middle layer are also frozen to train the convolutional layers of the top layer. It should be noted that the depth training of the first phase is to perform bottom training and middle training on the first image feature data. The depth training of the second phase is to perform middle-level training and top-level training on the second image feature data. That is, the first image feature data does not need top-level training, and the second image feature data is trained based on the first image feature data, so bottom-level training is not required.

Although the present invention has been disclosed by the aforementioned preferred embodiments, it is not intended to limit the present invention, and any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. As explained above, various modifications and changes can be made without destroying the spirit of the invention. Therefore, the protection scope of the present invention should be determined by the scope of the appended patent application.

5: Intelligent diagnosis system for solar power module defects

10: The first image capture module

20: Second image capture module

30: Image processing module

40: Deep Learning Module

50: Information Judgment Module

Claims (10)

An intelligent diagnosis system for solar power generation module defects, used for defect analysis of a plurality of solar power generation modules, comprising: a first image capture module to obtain a first image data of the solar power generation modules; a second image capturing module to obtain a second image data of the solar power generation modules; an image processing module for performing image processing on the first image data and the second image data to obtain a first image feature data and a second image feature data respectively; A deep learning module, firstly performs a first period of in-depth training on the first image feature data, and then performs a second period of in-depth training on the second image feature data, and the first period is less than the second instalment; and An information judging module analyzes the defects of the solar power generation modules according to the result of the deep training of the second phase. The intelligent diagnosis system for solar power module defects according to claim 1, wherein the first image data is an infrared image, and the second image data is an electroluminescence image. The intelligent diagnosis system for solar power module defects according to claim 1, wherein the depth training includes a plurality of convolutional layer trainings, and the convolutional layer trainings are performed in a sequence of bottom training, middle training and top training. The intelligent diagnosis system for solar power module defects according to claim 1, wherein the first phase of in-depth training is to perform bottom-level training and middle-level training on the first image feature data. The intelligent diagnosis system for solar power module defects according to claim 1, wherein the second-stage deep training is to perform middle-level training and top-level training on the second image feature data. An intelligent method for diagnosing defects of solar power generation modules, which is used for defect analysis of a plurality of solar power generation modules, comprising: Step 1: obtaining a first image data of the solar power generation modules; Step 2: obtaining a second image data of the solar power modules; Step 3: performing image processing on the first image data and the second image data to obtain a first image feature data and a second image feature data respectively; Step 4: First perform a first period of in-depth training on the first image feature data, and then perform a second period of in-depth training on the second image feature data, and the first period is less than the second period. number of periods; and Step 5: Analyze the defects of the solar power generation modules based on the results of the second phase of in-depth training. The intelligent diagnosis method for solar power module defects according to claim 6, wherein the first image data is an infrared image, and the second image data is an electroluminescence image. The intelligent diagnosis method for solar power module defects according to claim 6, wherein the depth training includes a plurality of convolutional layer trainings, and the convolutional layer trainings are performed in a sequence of bottom-level training, middle-level training, and top-level training. The method for intelligently diagnosing defects of a solar power generation module according to claim 6, wherein the deep training of the first phase is a bottom-level training and a middle-level training for the first image feature data. The method for intelligently diagnosing defects of a solar power generation module according to claim 6, wherein the second-stage in-depth training is to perform middle-level training and top-level training on the second image feature data.

Images