Method and device for quantitatively evaluating short-circuit fault degree of photovoltaic panel cell
Patent Information
- Application Number
- CN202510352522.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-24
- Publication Date
- 2026-09-25
AI Technical Summary
现有计算机视觉方法技术水平主要停留在目标检测，即故障定位和分类，没有实现故障的轮廓分割，以及故障面积的定量评估
[0030]本发明提出的光伏板电池片短路故障程度定量评估方法及装置创新地将实例分割模型应用于光伏板故障检测，在模型性能、故障评估以及成本控制方面展现出显著技术效果，有力推动了光伏运维智能化发展。整体方案采用标注框提示的实例分割算法制作掩码数据集，并微调训练实例分割模型，有效提高在光伏板电池片短路区域故障检测的准确性及稳健性，为故障评估提供可靠依据。通过计算光伏板故障区域掩码面积占整体区域掩码面积的比例，能直观量化故障程度，为运维人员判断光伏板是否需要更换提供科学参考，改变了以往凭借经验判断的方式，使决策更精准、合理。相较于现有的目标检测和实例分割算法因缺乏光伏板故障掩码数据集难以训练微调，本发明利用实例分割算法自动制作掩码数据集，避免人工手动标注掩码，极大降低标注成本，提高数据准备效率，加速模型训练进程，促进相关技术在光伏运维领域的应用推广。
Smart Images
Figure CN122820528A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of image recognition and deep learning technology, and more particularly to a method and apparatus for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries. Background Technology
[0002] This section is intended to provide background or context for embodiments of the present invention. The description herein is not intended to imply that it is prior art simply because it is included in this section.
[0003] Solar energy, as a widely distributed and environmentally friendly energy source, has been gradually applied to power supply in oil fields. However, photovoltaic panels installed in oil fields, operating in open-air environments for extended periods, are prone to various malfunctions, such as localized short circuits in solar cells, thus reducing power generation efficiency. However, maintenance personnel do not immediately replace a malfunctioning panel. For example, a small-area short circuit on a single panel may have a relatively minor impact on the overall system's power generation efficiency. Therefore, accurately locating the short circuit and effectively assessing its severity are crucial for maintaining efficient operation and organizing timely replacement.
[0004] Electroluminescence (EL) is an optical phenomenon that photovoltaic (PV) system maintenance personnel can use to detect faults in PV panels. The principle is that when a forward voltage is applied to the silicon material in a PV cell, a photoelectric effect is generated. After the PN junction of the panel is forward-biased, infrared light is emitted from its surface, and an infrared CCD camera then captures the image. Under EL conditions, defective areas in PV cells will exhibit reduced brightness or no light emission. Images captured by the camera can be inspected manually or automatically analyzed using computer vision methods. Current computer vision techniques mainly focus on target detection, i.e., fault location and classification, and have not yet achieved fault contour segmentation or quantitative assessment of fault area.
[0005] In summary, there is an urgent need for a technical solution that can overcome the above-mentioned shortcomings and enable time-limited fault contour segmentation and quantitative assessment of fault area. Summary of the Invention
[0006] To address the problems existing in current technologies, this invention proposes a method and device for quantitatively assessing the severity of short-circuit faults in photovoltaic (PV) cells. This invention utilizes deep learning technology to locate and segment the short-circuit region of the PV cell, calculating the proportion of the faulted area to the total area of the PV panel. This provides a scientific basis for determining the optimal maintenance and replacement timing of the PV panel, thus optimizing operation and maintenance costs.
[0007] In a first aspect of the present invention, a method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell is proposed, the method comprising:
[0008] Acquire images of short-circuit solar cells in a photovoltaic panel;
[0009] The short-circuit image of the photovoltaic panel cell is input into the instance segmentation model to obtain the fault area mask and the mask of the entire photovoltaic panel area. The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set and testing it using the test set to obtain the instance segmentation model. The instance segmentation model takes the short-circuit image of the photovoltaic panel cell as input and the fault area mask and the mask of the entire photovoltaic panel area as output.
[0010] Based on the fault area mask and the mask of the entire photovoltaic panel area, the ratio of the fault area mask area to the mask area of the entire photovoltaic panel area is determined, and the degree of photovoltaic panel failure is evaluated by the ratio.
[0011] In one embodiment of the present invention, the short-circuit image of the photovoltaic cell is captured by an electroluminescent camera.
[0012] In one embodiment of the present invention, when training the instance segmentation model, the searched photovoltaic cell short-circuit images are labeled with fault regions.
[0013] In one embodiment of the present invention, training the instance segmentation model further includes:
[0014] The image dataset and the corresponding bounding boxes of the fault areas are input into an instance segmentation algorithm based on bounding box prompts for calculation and processing to obtain the fault area mask and the mask of the entire photovoltaic panel area.
[0015] In one embodiment of the present invention, the machine learning model is the YOLO11x-seg model.
[0016] In one embodiment of the present invention, training the instance segmentation model further includes:
[0017] A machine learning model is trained based on the training set and the fault area mask; wherein, by combining the fault area mask and the mask of the entire photovoltaic panel area, the machine learning model is trained to map the short-circuit image of the photovoltaic cell to the fault area mask and the mask of the entire photovoltaic panel area.
[0018] The test set is input into the trained machine learning model, and the prediction results are output, including the fault area mask and the mask of the entire photovoltaic panel area.
[0019] In one embodiment of the present invention, training the instance segmentation model further includes:
[0020] The accuracy of the model is evaluated based on the prediction results. If the evaluation results do not meet the preset requirements, the dataset samples are collected again to train the model. If the preset requirements are met, an instance segmentation model with satisfactory performance is obtained.
[0021] In one embodiment of the present invention, evaluating the accuracy of the model based on the prediction results includes:
[0022] Calculate metrics including the overlap and error rate between the predicted mask and the real mask; if the overlap does not reach the preset overlap threshold, or the error rate is greater than or equal to the preset error threshold, the model is deemed not to have met the preset requirements and is retrained; if the overlap reaches the preset overlap threshold and the error rate is less than the preset error threshold, the model is deemed to have met the preset requirements and training is completed.
[0023] In a second aspect of the present invention, a device for quantitatively assessing the degree of short-circuit fault in photovoltaic cell is provided, the device comprising:
[0024] The image acquisition module is used to acquire short-circuit images of photovoltaic panel cells;
[0025] An image analysis module is used to input the short-circuit image of the photovoltaic panel cell into an instance segmentation model to obtain a fault area mask and a mask for the entire photovoltaic panel area. The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set; combining the fault area mask and the mask for the entire photovoltaic panel area to perform mapping learning on the machine learning model from the short-circuit image of the photovoltaic panel cell to the fault area mask and the mask for the entire photovoltaic panel area; and testing using the test set to obtain the instance segmentation model. The instance segmentation model takes the short-circuit image of the photovoltaic panel cell as input and the fault area mask and the mask for the entire photovoltaic panel area as output.
[0026] The fault assessment module is used to determine the ratio of the area of the fault area mask to the area of the overall photovoltaic panel mask based on the fault area mask and the overall photovoltaic panel mask, and to assess the degree of photovoltaic panel fault through the ratio.
[0027] In a third aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell modules.
[0028] In a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries.
[0029] In a fifth aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements a method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell modules.
[0030] This invention proposes a method and device for quantitatively assessing the severity of short-circuit faults in photovoltaic (PV) cells. It innovatively applies an instance segmentation model to PV cell fault detection, demonstrating significant technical advantages in model performance, fault assessment, and cost control, thus powerfully promoting the intelligent development of PV operation and maintenance. The overall solution uses an instance segmentation algorithm with bounding box prompts to create a mask dataset and fine-tunes the training of the instance segmentation model, effectively improving the accuracy and robustness of fault detection in short-circuit areas of PV cells, providing a reliable basis for fault assessment. By calculating the proportion of the mask area of the PV cell fault region to the total mask area, the severity of the fault can be intuitively quantified, providing a scientific reference for maintenance personnel to determine whether PV cells need replacement. This changes the previous reliance on experience-based judgment, making decisions more accurate and reasonable. Compared to existing object detection and instance segmentation algorithms, which are difficult to train and fine-tune due to the lack of PV cell fault mask datasets, this invention automatically creates mask datasets using instance segmentation algorithms, avoiding manual mask annotation, greatly reducing annotation costs, improving data preparation efficiency, accelerating model training, and promoting the application and dissemination of related technologies in the PV operation and maintenance field. Attached Figure Description
[0031] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0032] Figure 1 This is a schematic diagram of a method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to an embodiment of the present invention.
[0033] Figure 2 This is a schematic diagram of an electroluminescent camera capturing images of a photovoltaic panel using a training set according to a specific embodiment of the present invention.
[0034] Figure 3 This is a schematic diagram of the mask outline output after processing an image and its corresponding label box using the SAM algorithm, according to a specific embodiment of the present invention.
[0035] Figure 4This is a schematic diagram of an image output by a trained YOLO11x-seg model according to a specific embodiment of the present invention.
[0036] Figure 5 This is a schematic diagram of the architecture of a photovoltaic panel cell short-circuit fault quantitative assessment device according to an embodiment of the present invention.
[0037] Figure 6 This is a schematic diagram of a computer device structure according to an embodiment of the present invention. Detailed Implementation
[0038] The principles and spirit of the invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are given merely to enable those skilled in the art to better understand and implement the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided to make this disclosure more thorough and complete, and to fully convey the scope of this disclosure to those skilled in the art.
[0039] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.
[0040] According to an embodiment of the present invention, a method and apparatus for quantitatively assessing the degree of short-circuit faults in photovoltaic panel cells are proposed, relating to the fields of image recognition and deep learning technologies. The main objective is to calculate the proportion of the short-circuit area in the total area of the photovoltaic panel cells, providing a reference for determining whether the photovoltaic panel needs replacement based on the severity of the fault. The overall scheme applies an instance segmentation model from the field of computer vision to photovoltaic panel images captured by an electroluminescent camera, achieving target segmentation of the fault area and calculating its area.
[0041] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.
[0042] Figure 1 This is a schematic flowchart of a method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to an embodiment of the present invention. Figure 1 As shown, the method includes:
[0043] S101, Obtain short-circuit image of photovoltaic panel cell;
[0044] S102, input the short-circuit image of the photovoltaic panel cell into the instance segmentation model to obtain the fault area mask and the mask of the entire photovoltaic panel area;
[0045] The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set and testing it using the test set to obtain the instance segmentation model; wherein the instance segmentation model takes short-circuit images of photovoltaic panel cells as input and takes a fault area mask and a mask of the entire photovoltaic panel area as output.
[0046] S103, based on the fault area mask and the mask of the entire photovoltaic panel area, determine the ratio of the fault area mask area to the mask area of the entire photovoltaic panel area, and evaluate the degree of photovoltaic panel failure through the ratio.
[0047] To provide a clearer explanation of the quantitative assessment method for the degree of short-circuit faults in photovoltaic cells, each step will be explained in detail below.
[0048] In one embodiment, the short-circuit image of the photovoltaic cell is captured by an electroluminescent camera.
[0049] When training the instance segmentation model, the searched photovoltaic cell short-circuit images are labeled with bounding boxes for the fault areas.
[0050] Furthermore, the image dataset and the corresponding bounding boxes of the fault regions are input into an instance segmentation algorithm based on bounding box prompts for calculation and processing to obtain the fault region mask and the mask of the entire photovoltaic panel region.
[0051] The machine learning model is the YOLO11x-seg model.
[0052] This invention utilizes the Instance Segmentation Algorithm Based on Box Cueing (SAM) to create a mask dataset for fault regions of photovoltaic panels, and uses this dataset to fine-tune and train an instance segmentation model (YOLO11x-seg). By processing photovoltaic panel images captured by an electroluminescent camera using the trained instance segmentation model (YOLO11x-seg), the mask for the fault region can be quickly generated, and the area ratio of the fault region to the overall photovoltaic panel can be calculated.
[0053] Specifically, a machine learning model is trained based on the training set and the fault area mask; wherein, by combining the fault area mask and the mask of the entire photovoltaic panel area, the machine learning model is trained to map the short-circuit image of the photovoltaic cell to the fault area mask and the mask of the entire photovoltaic panel area.
[0054] The test set is input into the trained machine learning model, and the prediction results are output, including the fault area mask and the mask of the entire photovoltaic panel area.
[0055] The accuracy of the model is evaluated based on the prediction results. If the evaluation results do not meet the preset requirements, the dataset samples are collected again to train the model. If the preset requirements are met, an instance segmentation model with satisfactory performance is obtained.
[0056] Specifically, the system calculates metrics including the overlap between the predicted mask and the real mask, and the error rate. If the overlap does not reach the preset overlap threshold, or the error rate is greater than or equal to the preset error threshold, the model is deemed not to have met the preset requirements and is retrained. If the overlap reaches the preset overlap threshold and the error rate is less than the preset error threshold, the model is deemed to have met the preset requirements and training is complete.
[0057] In practical applications, this invention employs data-driven learning. During the fine-tuning training phase of the segmentation model, YOLO11x-seg is trained using training set images with fault region masks generated based on SAM. The model learns the characteristics of normal areas of the photovoltaic panel and short-circuit fault areas of the solar cells from this data, including color, texture, and shape. For example, short-circuit areas may exhibit specific brightness and texture features in images captured by an electroluminescent camera. By continuously learning these features, the model gradually develops the ability to identify short-circuit areas.
[0058] An instance segmentation mechanism is employed, using the instance segmentation model (YOLO11x-seg) to identify each independent target instance in an image and generate an accurate segmentation mask for it. When processing photovoltaic panel images, the model analyzes each region in the image to determine whether it belongs to a short-circuit region. If so, a corresponding mask is generated for that short-circuit region, and a mask for the entire photovoltaic panel region is also generated. This process, based on the model's feature extraction, feature fusion, target classification, and localization operations on the input image, achieves the function of instance segmentation.
[0059] During the model inference and mask generation phases, test set images are input into the trained YOLO11x-seg model. Based on the feature knowledge learned during training, the model predicts the photovoltaic panels and short-circuit regions in the input images, outputting predicted fault region masks and masks of the entire photovoltaic panel area. These masks are used to subsequently calculate the proportion of the fault area to assess the degree of photovoltaic panel failure.
[0060] It should be noted that although the operation of the method of the present invention has been described in a specific order in the above embodiments and figures, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.
[0061] In one specific embodiment, the training model stage and the data processing stage are used as examples for illustration.
[0062] S1, Collect and organize data:
[0063] Search and compile publicly available datasets of short-circuit images of photovoltaic cells captured by electroluminescent cameras, requiring that all images include bounding boxes indicating the fault area.
[0064] Specifically, a broad search and compilation of publicly available datasets of short-circuit images of photovoltaic cells captured by electroluminescent cameras is necessary. This step is crucial because a high-quality dataset forms the foundation for subsequent model training. The collected images must meet specific requirements: each image must include an accurate bounding box labeling the fault region to provide precise supervision information for the model. Through this step, a dataset containing diverse fault scenarios can be constructed, laying the foundation for the model's generalization ability.
[0065] S2, Instance segmentation and mask generation based on the SAM algorithm:
[0066] The image dataset and its corresponding fault region bounding boxes are input into the Instance Segmentation Algorithm (SAM) based on bounding box prompts to obtain a fine-grained contour mask of the fault region and a mask of the entire photovoltaic panel area. The dataset is then divided into training and testing sets according to a predetermined ratio.
[0067] Specifically, the prepared image dataset and the corresponding fault region bounding boxes are input into the instance segmentation algorithm (SAM) based on bounding box prompts.
[0068] The SAM algorithm can generate detailed contour masks of fault areas and masks of the entire photovoltaic panel area using the information from the bounding boxes. The key to this step lies in the accuracy and high generalization of the SAM algorithm. This algorithm ensures that the generated masks closely match the actual situation, providing reliable data support for subsequent model training. Simultaneously, to verify the model's performance, the dataset is divided into training and test sets according to a preset ratio to ensure the objectivity and accuracy of the evaluation results. Since generating masks using the SAM algorithm is relatively slow and unsuitable for the real-time requirements of production inference, this step is only for creating the mask dataset, preparing for the next step of training a faster segmentation model.
[0069] S3, Training and Fine-tuning the Instance Segmentation Model:
[0070] A machine learning model is trained based on the image training set and the fault region mask obtained by S2. This machine learning model is built using the fine-tuned instance segmentation algorithm (YOLO11x-seg).
[0071] After obtaining the training set images and their corresponding fault region masks, these data were used for fine-tuning the YOLO11x-seg model. YOLO11x-seg is an advanced instance segmentation model that can achieve accurate segmentation results while maintaining high detection speed.
[0072] S4, Model Inference and Mask Generation:
[0073] Input the test set images into the model trained by S3, and output the predicted fault area mask and the mask of the entire photovoltaic panel area.
[0074] After model training is complete, the test set images are input into the trained YOLO11x-seg model for inference. The instance segmentation model will output the predicted fault region mask and the mask for the entire photovoltaic panel area.
[0075] S5, Evaluate the accuracy of the model's segmentation mask:
[0076] To ensure the reliability of the model in practical applications, the accuracy of the model's segmentation mask needs to be rigorously evaluated. This can be achieved by calculating metrics such as the overlap between the predicted mask and the true mask, and the error rate.
[0077] S6, Fault Area Proportion Calculation and Result Output:
[0078] Based on the fault area mask generated by the model and the mask of the entire photovoltaic panel area, the proportion of the fault area to the total area of the photovoltaic panel can be calculated. This proportion is an important indicator for assessing the degree of photovoltaic panel failure and can provide scientific decision-making basis for operation and maintenance personnel. By outputting this proportion result, a quantitative assessment of the degree of photovoltaic panel failure can be achieved, providing strong support for the operation and maintenance management of photovoltaic power plants.
[0079] The technical solution of the present invention will be described below using a practical application scenario as an example. First, 1379 images captured by an electroluminescent camera are collected and divided into training and test sets at a ratio of 8:2, of which 1103 images are for the training set and 276 images are for the test set.
[0080] Images of photovoltaic panels (e.g., taken by the electroluminescent camera used in the training set) Figure 2 (As shown) and its corresponding label are input into the SAM algorithm for segmentation. The output of the SAM algorithm is as follows: Figure 3 As shown. In Figure 2 In the diagram, the white areas represent normal battery cells, and the black areas represent short-circuited areas. Figure 3 In the diagram, the red outline represents the entire photovoltaic panel, the yellow outline represents short-circuited solar cells, and the green outline represents marker points.
[0081] use Figure 3The YOLO11x-seg model is fine-tuned using mask training. In the model inference and mask generation steps, the trained YOLO11x-seg model is used to generate masks as shown below. Figure 4 The image shown outputs a red outline representing the entire photovoltaic panel and a yellow outline representing short-circuited solar cells.
[0082] Compare Figure 3 and Figure 4 It can be seen that, Figure 4 Mask outline and Figure 3 The mask outlines are similar. The total pixel area of the red-outlined photovoltaic panel is calculated to be 236352.07 pixels, while the total pixel area of the short-circuited solar cell (yellow outline) is 32355.91 pixels. Dividing the short-circuited area by the total area gives the faulty area ratio, which is 13.7%. This value can be used to assess whether the photovoltaic panel needs to be replaced.
[0083] Table 1, for example, records the accuracy of the model in detecting various faults on the test set.
[0084] Table 1
[0085] category picture Example Mask_P Mask_R mAP50 mAP50-95 all 276 4353 0.724 0.571 0.52 0.374 photovoltaic panels 274 275 0.962 0.967 0.974 0.948 Short circuit string 23 82 0.838 0.88 0.919 0.648 Short-circuit plate 81 219 0.579 0.863 0.853 0.598 crack 82 106 0.53 0.84 0.778 0.313 microcracks 226 2017 0.305 0.568 0.372 0.113 Other faults 249 1638 0.223 0.592 0.315 0.121
[0086] The quantitative assessment method for the degree of short-circuit fault in photovoltaic cell proposed in this invention has made improvements in at least the following aspects, achieving the desired improvement effect:
[0087] 1. The model demonstrates good accuracy. Experimental results show that the YOLO11x-seg model trained using the SAM algorithm to create a masked dataset exhibits good accuracy and robustness on the photovoltaic panel cell short-circuit fault dataset. The accuracy for photovoltaic panels is 0.962, recall is 0.967, and mAP50 value is 0.974; for cell string short circuits, the accuracy is 0.838, recall is 0.88, and mAP50 value is 0.919; for individual cell short circuits, the accuracy is 0.579, recall is 0.863, and mAP50 value is 0.853. Accuracy results for other faults, including cracks and microcracks, are shown in Table 1.
[0088] 2. Fault severity assessment and replacement timing reference: The proportion of the faulty area is calculated by using the mask of the faulty area of the photovoltaic panel, which provides a scientific basis for assessing the degree and scope of the fault.
[0089] 3. It saves on the cost of manual mask annotation; the object detection and instance segmentation algorithms are already very powerful, but they cannot be trained and fine-tuned due to the lack of a photovoltaic panel fault mask dataset. This method uses the SAM algorithm to automatically generate the mask dataset, solving the problem of high costs associated with manual mask annotation.
[0090] After introducing the method of exemplary embodiments of the present invention, the following references are made. Figure 5This invention provides an exemplary embodiment of a device for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries.
[0091] The implementation of the device for quantitatively assessing the degree of short-circuit faults in photovoltaic cell solar panels can refer to the implementation of the method described above, and repeated details will not be elaborated upon. The terms "module" or "unit" used below can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.
[0092] Based on the same inventive concept, this invention also proposes a device for quantitatively assessing the degree of short-circuit faults in photovoltaic panel cells, such as... Figure 5 As shown, the device includes:
[0093] Image acquisition module 510 is used to acquire short-circuit images of photovoltaic panel cells;
[0094] The image analysis module 520 is used to input the short-circuit image of the photovoltaic panel cell into the instance segmentation model to obtain a fault area mask and a mask of the entire photovoltaic panel area. The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set; combining the fault area mask and the mask of the entire photovoltaic panel area to perform mapping learning from the short-circuit image of the photovoltaic panel cell to the fault area mask and the mask of the entire photovoltaic panel area; and testing using the test set to obtain the instance segmentation model. The instance segmentation model takes the short-circuit image of the photovoltaic panel cell as input and the fault area mask and the mask of the entire photovoltaic panel area as output.
[0095] The fault assessment module 530 is used to determine the ratio of the area of the fault area mask to the area of the mask of the whole photovoltaic panel based on the fault area mask and the mask of the whole photovoltaic panel, and to assess the degree of photovoltaic panel fault through the ratio.
[0096] In one embodiment, the short-circuit image of the photovoltaic cell is captured by an electroluminescent camera.
[0097] In one embodiment, when training the instance segmentation model, the searched photovoltaic cell short-circuit images are labeled with bounding boxes for the fault areas.
[0098] In one embodiment, training the instance segmentation model further includes:
[0099] The image dataset and the corresponding bounding boxes of the fault areas are input into an instance segmentation algorithm based on bounding box prompts for calculation and processing to obtain the fault area mask and the mask of the entire photovoltaic panel area.
[0100] In one embodiment, the machine learning model is the YOLO11x-seg model.
[0101] In one embodiment, training the instance segmentation model further includes:
[0102] A machine learning model is trained based on the training set and the fault area mask; wherein, by combining the fault area mask and the mask of the entire photovoltaic panel area, the machine learning model is trained to map the short-circuit image of the photovoltaic cell to the fault area mask and the mask of the entire photovoltaic panel area.
[0103] The test set is input into the trained machine learning model, and the prediction results are output, including the fault area mask and the mask of the entire photovoltaic panel area.
[0104] In one embodiment, training the instance segmentation model further includes:
[0105] The accuracy of the model is evaluated based on the prediction results. If the evaluation results do not meet the preset requirements, the dataset samples are collected again to train the model. If the preset requirements are met, an instance segmentation model with satisfactory performance is obtained.
[0106] In one embodiment, evaluating the model accuracy based on the prediction results includes:
[0107] Calculate metrics including the overlap and error rate between the predicted mask and the real mask; if the overlap does not reach the preset overlap threshold, or the error rate is greater than or equal to the preset error threshold, the model is deemed not to have met the preset requirements and is retrained; if the overlap reaches the preset overlap threshold and the error rate is less than the preset error threshold, the model is deemed to have met the preset requirements and training is completed.
[0108] It should be noted that although several modules of the photovoltaic panel cell short-circuit fault quantitative assessment device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules.
[0109] Based on the aforementioned inventive concept, such as Figure 6As shown, the present invention also proposes a computer device 600, including a memory 610, a processor 620, and a computer program 630 stored in the memory 610 and executable on the processor 620. When the processor 620 executes the computer program 630, it implements the aforementioned method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries.
[0110] Based on the aforementioned inventive concept, this invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries.
[0111] Based on the aforementioned inventive concept, this invention proposes a computer program product, which includes a computer program that, when executed by a processor, implements a method for quantitatively assessing the degree of short-circuit faults in photovoltaic cell batteries.
[0112] This invention proposes a method and device for quantitatively assessing the severity of short-circuit faults in photovoltaic (PV) cells. It innovatively applies an instance segmentation model to PV cell fault detection, demonstrating significant technical advantages in model performance, fault assessment, and cost control, thus powerfully promoting the intelligent development of PV operation and maintenance. The overall solution uses an instance segmentation algorithm with bounding box prompts to create a mask dataset and fine-tunes the training of the instance segmentation model, effectively improving the accuracy and robustness of fault detection in short-circuit areas of PV cells, providing a reliable basis for fault assessment. By calculating the proportion of the mask area of the PV cell fault region to the total mask area, the severity of the fault can be intuitively quantified, providing a scientific reference for maintenance personnel to determine whether PV cells need replacement. This changes the previous reliance on experience-based judgment, making decisions more accurate and reasonable. Compared to existing object detection and instance segmentation algorithms, which are difficult to train and fine-tune due to the lack of PV cell fault mask datasets, this invention automatically creates mask datasets using instance segmentation algorithms, avoiding manual mask annotation, greatly reducing annotation costs, improving data preparation efficiency, accelerating model training, and promoting the application and dissemination of related technologies in the PV operation and maintenance field.
[0113] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.
[0114] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0115] This invention is described with reference to flowchart illustrations and / or block diagrams of methods and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0118] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for quantitatively assessing the degree of short-circuit fault in photovoltaic panel cells, characterized in that, The method includes: Acquire images of short-circuit solar cells in a photovoltaic panel; The short-circuit image of the photovoltaic panel cell is input into the instance segmentation model to obtain the fault area mask and the mask of the entire photovoltaic panel area. The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set and testing it using the test set to obtain the instance segmentation model; the instance segmentation model takes the short-circuit image of the photovoltaic panel cell as input and the fault area mask and the mask of the entire photovoltaic panel area as output. Based on the fault area mask and the mask of the entire photovoltaic panel area, the ratio of the fault area mask area to the mask area of the entire photovoltaic panel area is determined, and the degree of photovoltaic panel failure is evaluated through the ratio.
2. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 1, characterized in that, The short-circuit image of the photovoltaic panel cell was captured by an electroluminescent camera.
3. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 1, characterized in that, When training the instance segmentation model, the searched photovoltaic cell short-circuit images are labeled with bounding boxes for the fault areas.
4. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 3, characterized in that, Training the instance segmentation model also includes: The image dataset and the corresponding bounding boxes of the fault areas are input into an instance segmentation algorithm based on bounding box prompts for calculation and processing to obtain the fault area mask and the mask of the entire photovoltaic panel area.
5. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 1, characterized in that, The machine learning model is the YOLO11x-seg model.
6. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 1, characterized in that, Training the instance segmentation model also includes: A machine learning model is trained based on the training set and the fault area mask; wherein, by combining the fault area mask and the mask of the entire photovoltaic panel area, the machine learning model is trained to map the short-circuit image of the photovoltaic cell to the fault area mask and the mask of the entire photovoltaic panel area. The test set is input into the trained machine learning model, and the prediction results are output, including the fault area mask and the mask of the entire photovoltaic panel area.
7. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 6, characterized in that, Training the instance segmentation model also includes: The accuracy of the model is evaluated based on the prediction results. If the evaluation results do not meet the preset requirements, the dataset samples are collected again to train the model. If the preset requirements are met, an instance segmentation model with satisfactory performance is obtained.
8. The method for quantitatively assessing the degree of short-circuit fault in photovoltaic cell according to claim 7, characterized in that, The accuracy of the model is evaluated based on the prediction results, including: Calculate metrics including the overlap and error rate between the predicted mask and the real mask; if the overlap does not reach the preset overlap threshold, or the error rate is greater than or equal to the preset error threshold, the model is deemed not to have met the preset requirements and is retrained; if the overlap reaches the preset overlap threshold and the error rate is less than the preset error threshold, the model is deemed to have met the preset requirements and training is completed.
9. A device for quantitatively assessing the degree of short-circuit fault in photovoltaic cell solar panels, characterized in that, The device includes: The image acquisition module is used to acquire short-circuit images of photovoltaic panel cells; An image analysis module is used to input the short-circuit image of the photovoltaic panel cell into an instance segmentation model to obtain a fault area mask and a mask for the entire photovoltaic panel area. The instance segmentation model is pre-trained using the following method: searching and organizing short-circuit images of photovoltaic panel cells to obtain an image dataset; dividing the image dataset into a training set and a test set according to a preset ratio; training the machine learning model using the training set; combining the fault area mask and the mask for the entire photovoltaic panel area to perform mapping learning on the machine learning model from the short-circuit image of the photovoltaic panel cell to the fault area mask and the mask for the entire photovoltaic panel area; and testing using the test set to obtain the instance segmentation model. The instance segmentation model takes the short-circuit image of the photovoltaic panel cell as input and the fault area mask and the mask for the entire photovoltaic panel area as output. The fault assessment module is used to determine the ratio of the area of the fault area mask to the area of the overall photovoltaic panel mask based on the fault area mask and the overall photovoltaic panel mask, and to assess the degree of photovoltaic panel fault through the ratio.
10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 8.
11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.
12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.