Material image analysis method for photovoltaic module subfissure characteristics

Abstract

The invention discloses a material image analysis method for the hidden crack characteristic of a photovoltaic module, and the method comprises the steps: obtaining an EL image from a photovoltaic module independently arranged in a to-be-detected region, and carrying out the preprocessing, thereby providing a basis for the rapid subsequent analysis. And inputting the to-be-analyzed subfissure data set into a preset risk assessment model, and quickly determining a plurality of risk values in the to-be-analyzed subfissure data set so as to distinguish the approximate damaged position of the photovoltaic module. Under different stress conditions in a natural environment corresponding to the photovoltaic module, a first layer weight value, a second layer weight value and a third layer weight value are weighted and given to the risk value, so that subsequent fusion calculation is facilitated. And finally, a first-layer weight value, a second-layer weight value and a third-layer weight value are calculated through fusion, and a risk assessment report of the independently arranged photovoltaic modules in the to-be-detected area is determined. And the expansion risk of the subfissure under the subsequent stress cannot be effectively evaluated.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic microcrack analysis technology, and specifically to a material image analysis method for the microcrack characteristics of photovoltaic modules. Background Technology

[0002] Currently, analyzing the microcrack characteristics of photovoltaic modules is a standard procedure in photovoltaic applications. Microcracks are tiny cracks inside the cells that can lead to module power degradation, hot spot effects, and even complete failure, severely impacting power generation efficiency and system lifespan. By analyzing the causes, propagation patterns, and electrical performance impacts of microcracks, production processes can be optimized, installation and transportation procedures improved, and effective testing standards established. This can enhance module reliability, reduce power plant operation and maintenance costs, and ensure the long-term stable operation and economic benefits of photovoltaic systems.

[0003] The detection of microcracks in photovoltaic modules mainly relies on electroluminescence imaging technology. EL images can clearly reveal microcracks and grid breakage defects inside the cells. However, this technology can only determine the presence or simple classification of microcracks, and cannot effectively assess the risk of microcrack propagation under subsequent stress. Summary of the Invention

[0004] The purpose of this invention is to provide a material image analysis method for the microcrack characteristics of photovoltaic modules, in order to solve the technical problem that the existing technology can only determine the presence or absence of microcracks or make simple classifications, but cannot effectively assess the risk of microcrack propagation under subsequent stress.

[0005] The technical solution of this invention is implemented as follows:

[0006] A material image analysis method for the microcrack characteristics of photovoltaic modules includes the following steps:

[0007] Step S1: Obtain EL images of independently arranged photovoltaic modules in the area to be detected, perform preprocessing, and generate a dataset of hidden cracks to be analyzed;

[0008] Step S2: Input the dataset of hidden cracks to be analyzed into a preset risk assessment model and extract multiple risk values ​​from the dataset of hidden cracks to be analyzed;

[0009] Step S3: The expansion rate of the risk value under different stress conditions is weighted to generate a first-layer weight value, a second-layer weight value, and a third-layer weight value;

[0010] Step S4: By fusing the first layer weight value, the second layer weight value, and the third layer weight value, output a risk assessment report for the photovoltaic modules arranged independently in multiple areas to be detected.

[0011] A further technical solution is that step S1 specifically includes:

[0012] Step S11: Divide the area with multiple photovoltaic modules into multiple regions of equal size to generate the area to be tested;

[0013] Step S12: Select the independently arranged photovoltaic modules in each of the areas to be detected;

[0014] Step S13: Obtain EL images of the independently arranged photovoltaic modules through the image acquisition unit;

[0015] Step S14: Identify hidden crack regions in the EL image and collect the hidden crack regions to generate a hidden crack dataset to be analyzed.

[0016] A further technical solution is that step S14 specifically includes:

[0017] Step S141: Extract the maximum length, average width, main grid lines, number per unit area, relative position of the main grid lines and fine grid lines, brightness value, shape of the hidden crack, and the proportion of broken fine grid lines in the hidden crack region;

[0018] Step S142: The maximum length, average width, main grid line, number per unit area, relative position of the main grid line and the fine grid line, brightness value, shape of the crack, and the ratio of the number of broken fine grid lines in the crack region are marked as crack characteristic parameters.

[0019] Step S143: Pack and collect the hidden crack feature parameters to generate the hidden crack dataset to be analyzed.

[0020] A further technical solution is that step S2 specifically includes:

[0021] Step S21: Based on the historical hidden crack dataset, construct the mapping relationship between hidden crack feature parameters and expansion rate;

[0022] Step S22: Train the mapping relationship between the hidden crack feature parameters and the propagation rate using a physical information neural network, and generate the preset risk assessment model;

[0023] Step S23: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model;

[0024] Step S24: Extract the primary data set of different hidden crack characteristic parameters and propagation rates from the hidden crack dataset to be analyzed in the preset risk assessment model;

[0025] Step S25: If multiple different primary data sets exceed the warning line, they are classified into multiple risk values.

[0026] A further technical solution is that step S24 specifically includes:

[0027] Step S241: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model and call the feature recognition module in the model;

[0028] Step S242: Based on the feature recognition module, extract multiple feature parameters related to the microcracks one by one from the microcrack dataset to be analyzed. The feature parameters include the length, orientation, location and morphological features of the microcracks.

[0029] Step S243: Call the propagation rate analysis module in the preset risk assessment model to calculate the propagation rate of the hidden crack associated with the multiple characteristic parameters;

[0030] Step S244: Combine each extracted hidden crack feature parameter and its associated hidden crack propagation rate to generate the primary data set.

[0031] A further technical solution is that step S244 specifically includes:

[0032] Step S2441: Receive the asynchronous data stream of the hidden crack characteristic parameters and the spread rate;

[0033] Step S2442: Based on preset data verification rules, perform matching verification on the asynchronous data stream and remove invalid data pairs that cannot be matched or have logical conflicts.

[0034] Step S2443: Logically associate the valid data pairs that have passed the preset data verification rules, and encapsulate them into data objects;

[0035] Step S2444: Assign a unique index to the data object and map it to its corresponding time-series position in the dataset to be analyzed;

[0036] Step S2445: Collect all the unique indexes and add set metadata to generate the primary data set.

[0037] A further technical solution is that step S2443 specifically includes:

[0038] Step S24431: Based on the physical properties of the hidden crack characteristic parameters in the data pair, construct a hierarchical field including basic properties and derived properties;

[0039] Step S24432: Logically bind the hierarchical field with the corresponding valid data pair, injecting data fingerprints and timestamps during the binding process to complete the encapsulation of the data object.

[0040] A further technical solution is that step S3 specifically includes:

[0041] Step S31: Using the conditional attributes under different stress conditions and historical fault data, evaluate the intensity of the influence of stress conditions on the propagation rate.

[0042] Step S32: Combine the risk values ​​of multiple time-series nodes under the same stress condition with the propagation rate to generate the primary weighting coefficients under the same stress condition.

[0043] Step S33: Perform nonlinear mapping processing on the influence intensity and the primary weight coefficient to generate the first layer weight value, the second layer weight value and the third layer weight value corresponding to structural risk, material degradation risk and instantaneous failure risk, respectively.

[0044] A further technical solution is that step S4 specifically includes:

[0045] Step S41: Construct a three-dimensional weight space, whose orthogonal coordinate axes correspond to the first layer weight value, the second layer weight value, and the third layer weight value, respectively. Then, synthesize the first layer weight value, the second layer weight value, and the third layer weight value into a risk vector and calculate the magnitude and direction of the risk vector.

[0046] Step S42: Based on the direction of the risk vector, determine the main failure factors of the photovoltaic module, and based on the magnitude of the risk vector and historical data of the area where the photovoltaic module is located, assess the probability level and expected severity level of the main failure factors.

[0047] Step S43: Based on the probability level and expected hazard level of the occurrence of the main failure factors, a spatial interpolation algorithm is used to generate a risk probability distribution map and a hazard level distribution map of the area to be detected, and high-risk clusters are identified.

[0048] Step S44: Call the report generation engine to overlay and render the information of the high-risk cluster, risk probability distribution map and hazard degree distribution map with the photovoltaic module location map to generate a risk assessment report.

[0049] A further technical solution is that step S42 further includes:

[0050] Step S421: Based on the main failure factors, predict the number of cycles required for the crack in the photovoltaic module to expand from its current size to the size of complete failure under day and night thermal cycling.

[0051] Step S422: Combining the local climate data of the photovoltaic module, convert the number of cycles into actual time, and estimate the remaining lifespan of the photovoltaic module.

[0052] The beneficial effects of this invention are as follows:

[0053] EL images are obtained from independently arranged photovoltaic modules within a segmented detection area and preprocessed to provide a foundation for subsequent rapid analysis. The dataset of microcracks to be analyzed is input into a pre-defined risk assessment model to quickly determine multiple risk values ​​within the dataset, thus identifying the approximate location of damage to the photovoltaic modules. Under different stress conditions in the natural environment corresponding to the photovoltaic modules, the risk values ​​are weighted and assigned first, second, and third-level weight values ​​to facilitate subsequent fusion calculations. Finally, by fusing the first, second, and third-level weight values, a risk assessment report for the independently arranged photovoltaic modules within the detection area is generated. This addresses the technical problem that existing technologies can only determine the presence or absence of microcracks or provide simple classification, but cannot effectively assess the risk of microcrack propagation under subsequent stress. Attached Figure Description

[0054] Figure 1 A flowchart of a material image analysis method for the microcrack characteristics of photovoltaic modules provided by the present invention;

[0055] Figure 2 A flowchart of step S1 of a material image analysis method for the microcrack characteristics of photovoltaic modules provided by the present invention;

[0056] Figure 3 A flowchart of step S2 of a material image analysis method for the microcrack characteristics of photovoltaic modules provided by the present invention;

[0057] Figure 4 A flowchart of step S3 of a material image analysis method for the microcrack characteristics of photovoltaic modules provided by the present invention;

[0058] Figure 5 The flowchart of step S4 of the material image analysis method for the microcrack characteristics of photovoltaic modules provided by the present invention. Detailed Implementation

[0059] To better understand the technical content of this invention, specific embodiments are provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0060] See Figures 1 to 2 This invention provides a material image analysis method for the microcrack characteristics of photovoltaic modules, comprising the following steps:

[0061] Step S1: Obtain EL images of independently arranged photovoltaic modules in the area to be detected, perform preprocessing, and generate a dataset of hidden cracks to be analyzed;

[0062] Step S2: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model and extract multiple risk values ​​from the dataset of hidden cracks to be analyzed;

[0063] Step S3: The expansion rate of the risk value under different stress conditions is weighted to generate the first layer weight value, the second layer weight value, and the third layer weight value;

[0064] Step S4: By integrating the first-layer weight value, the second-layer weight value, and the third-layer weight value, output a risk assessment report for photovoltaic modules arranged independently in multiple areas to be tested.

[0065] It should be noted that EL images refer to images generated by capturing the physical phenomenon of light emitted by materials under the influence of electric current or electric field using a camera.

[0066] Preprocessing includes, but is not limited to, grayscale correction, contrast enhancement, and cell segmentation.

[0067] Specifically, EL images are obtained from independently arranged photovoltaic modules within the detection area and preprocessed to provide a foundation for subsequent rapid analysis. The dataset of microcracks to be analyzed is input into a pre-defined risk assessment model to quickly determine multiple risk values ​​within the dataset, thus identifying the approximate location of damage to the photovoltaic modules. Under different stress conditions in the natural environment corresponding to the photovoltaic modules, the risk values ​​are weighted and assigned first, second, and third-level weight values ​​to facilitate subsequent fusion calculations. Finally, by fusing the first, second, and third-level weight values, a risk assessment report for the independently arranged photovoltaic modules within the detection area is generated. This addresses the technical problem that existing technologies can only determine the presence or absence of microcracks or provide simple classifications, but cannot effectively assess the risk of microcrack propagation under subsequent stress.

[0068] Preferably, step S1 specifically includes:

[0069] Step S11: Divide the area with multiple photovoltaic modules into multiple regions of equal size to generate the area to be tested;

[0070] Step S12: Select the independently arranged photovoltaic modules in each area to be detected;

[0071] Step S13: Obtain EL images of independently arranged photovoltaic modules through the image acquisition unit;

[0072] Step S14: Identify hidden crack regions in the EL image and collect the hidden crack regions to generate a dataset of hidden cracks to be analyzed.

[0073] It should be noted that the image acquisition unit refers to a camera with extremely high sensitivity that can efficiently capture near-infrared light.

[0074] In this embodiment of the invention, a photovoltaic power plant comprises a large area composed of multiple photovoltaic modules. To enable rapid analysis of the modules, the large area is divided into multiple regions of equal size. This allows for the graphical analysis of each photovoltaic module within that region, and dividing it into multiple equal-sized regions reduces the analysis area and increases image clarity. Each selected region contains multiple independent photovoltaic modules arranged horizontally and vertically. An image acquisition unit captures EL images of each such independent photovoltaic module. If hidden cracks are identified in the EL images, these cracks will inevitably form an opening in the region. If there are multiple regions with hidden cracks, or multiple photovoltaic modules with multiple regions with hidden cracks, a dataset of hidden cracks to be analyzed is generated for subsequent analysis using a pre-defined risk assessment model.

[0075] Furthermore, step S14 specifically includes:

[0076] Step S141: Extract the maximum length, average width, main grid lines, number per unit area, relative position of main grid lines and fine grid lines, brightness value, shape of the crack, and proportion of broken fine grid lines in the crack region;

[0077] Step S142: The maximum length, average width, main grid lines, number per unit area, relative position of main grid lines and fine grid lines, brightness value, shape of the crack, and proportion of broken fine grid lines in the crack region are marked as the crack characteristic parameters.

[0078] Step S143: Pack and collect the characteristic parameters of the hidden cracks to generate a dataset of hidden cracks to be analyzed.

[0079] In this embodiment of the invention, microcrack regions are automatically identified from the EL images of photovoltaic modules using image analysis technology. Specific parameters include the crack's length, width, direction, number, brightness, shape, and its positional relationship with the current collection grid lines. These measured values ​​are then categorized and labeled as microcrack characteristic parameters. Finally, these microcrack characteristic parameters are packaged together to form a microcrack dataset that can be directly analyzed by a computer. This overcomes the drawbacks of low efficiency, poor consistency, and inability to perform large-scale data analysis associated with manual detection, providing a data foundation for subsequent automated and high-precision microcrack severity grading and lifetime prediction using a pre-set risk assessment model.

[0080] Preferably, step S2 specifically includes:

[0081] Step S21: Based on the historical hidden crack dataset, construct the mapping relationship between hidden crack feature parameters and expansion rate;

[0082] Step S22: Use a physical information neural network to train the mapping relationship between the characteristic parameters of hidden cracks and the propagation rate, and generate a preset risk assessment model;

[0083] Step S23: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model;

[0084] Step S24: Extract the primary data set of different hidden crack characteristic parameters and propagation rates from the hidden crack dataset to be analyzed in the preset risk assessment model;

[0085] Step S25: If multiple different primary data sets exceed the warning line, they are classified into multiple risk values.

[0086] It should be noted that physical information neural networks refer to the technology of directly embedding mechanical laws and conservation equations as constraints into neural networks.

[0087] In this embodiment of the invention, by analyzing a large amount of known historical microcrack data, the system learns and establishes the correlation between the length, width, direction, number, brightness, shape of the microcracks, their positional relationship with the current collection grid, and the rate of their subsequent expansion. Subsequently, a physical information neural network is used for training. This network not only learns from the data but also incorporates materials mechanics, thereby generating a high-precision risk assessment model that can accurately predict the expansion behavior of previously unseen microcracks with only a small amount of data. When new microcrack data is input, the risk assessment model can automatically extract the characteristics of rapidly appearing microcracks and predict their future expansion rate. That is, based on the length, width, direction, number, brightness, shape of minor microcracks and their positional relationship with the current collection grid, it predicts their expansion rate and trajectory, forming a preliminary data report. Finally, by analyzing the prediction results, once the predicted expansion rate of multiple components exceeds the safety threshold, they are automatically classified as high-risk objects. This improves the accuracy of microcrack risk prediction for photovoltaic modules, enabling the identification of high-risk components before a failure occurs, providing maintenance personnel with a clear priority list, and effectively avoiding power generation losses and safety accidents.

[0088] Furthermore, step S24 specifically includes:

[0089] Step S241: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model and call the feature recognition module in the model;

[0090] Step S242: Based on the feature recognition module, extract multiple feature parameters related to the microcracks one by one from the microcrack dataset to be analyzed. The feature parameters include the length, orientation, location and morphological features of the microcracks.

[0091] Step S243: Call the propagation rate analysis module in the preset risk assessment model to calculate the propagation rate of hidden cracks associated with multiple characteristic parameters;

[0092] Step S244: Combine each extracted microcrack feature parameter and its associated microcrack propagation rate to generate a primary dataset.

[0093] In this embodiment of the invention, the microcrack data to be analyzed is input into a pre-trained risk assessment model. First, the feature recognition function in the model is activated to automatically and accurately identify the key features of each microcrack from the data. The key features include specific information such as its length, direction, location, and shape. Then, the model activates the propagation rate analysis function to calculate the potential propagation speed of each microcrack based on these feature parameters. Finally, the features of each microcrack are precisely matched with its propagation rate and packaged to generate a primary dataset, providing a data foundation for accurately judging the safety risks of photovoltaic modules and improving the objectivity of the assessment results.

[0094] Furthermore, step S244 specifically includes:

[0095] Step S2441: Receive asynchronous data streams of hidden crack characteristic parameters and spread rate;

[0096] Step S2442: Based on the preset data verification rules, perform matching verification on the asynchronous data stream and remove invalid data pairs that cannot be matched or have logical conflicts.

[0097] Step S2443: Logically associate the valid data pairs that have passed the preset data verification rules, and encapsulate them into data objects;

[0098] Step S2444: Assign a unique index to the data object and map it to its corresponding time-series position in the dataset to be analyzed;

[0099] Step S2445: Collect all unique indexes and add collection metadata to generate a primary data collection.

[0100] In this embodiment of the invention, asynchronous data streams of hidden crack features and propagation rates from different analysis modules are received. Then, based on preset verification rules (i.e., whether the length and rate are within a reasonable range), these data are automatically paired and screened, discarding obviously erroneous or mismatched invalid data, such as leaves, filaments, watermarks, or other pseudo-shapes resembling cracks attached to photovoltaic modules. Next, the verified valid data are paired, bundled, and packaged into a complete data unit. Each data unit is then assigned a unique ID number, which can accurately trace its specific location and timing in the original dataset. Finally, all numbered data units are aggregated, and overall descriptive information about the dataset is added, ultimately forming a primary database. Through data cleaning, association, and standardized encapsulation, the accuracy, consistency, and traceability of the data used for subsequent risk analysis are ensured, fundamentally avoiding analysis errors caused by data mismatch, omissions, or format confusion.

[0101] Furthermore, step S2443 specifically includes:

[0102] Step S24431: Based on the physical properties of the hidden crack characteristic parameters in the data pair, construct a hierarchical field including basic properties and derived properties;

[0103] Step S24432: Logically bind the hierarchical fields with the corresponding valid data pairs, injecting data fingerprints and timestamps during the binding process to complete the encapsulation of data objects.

[0104] It should be noted that the basic attributes refer to the length value and orientation angle, while the derived attributes refer to the risk index inferred based on the length and orientation.

[0105] In this embodiment of the invention, after obtaining effective data pairs of microcrack characteristics and propagation rates, a data framework is automatically constructed based on the principle that length represents the potential hazard scale and orientation affects the degree of stress concentration. The data framework not only includes the basic attributes obtained through direct measurement, but also automatically calculates derived attributes with predictive significance. Subsequently, this data framework is firmly bound to the original data pair, and a unique data fingerprint and precise time stamp are embedded in this process, thereby ultimately forming a complete, independent, and self-verifiable data pair. This not only improves the accuracy and efficiency of subsequent data analysis, but also provides a reliable guarantee for data tracking and verification.

[0106] Preferably, step S3 specifically includes:

[0107] Step S31: Using conditional properties under different stress conditions and historical fault data, assess the intensity of the influence of stress conditions on the propagation rate.

[0108] Step S32: Combine the risk values ​​and propagation rates of multiple time-series nodes under the same stress condition to generate primary weighting coefficients under the same stress condition;

[0109] Step S33: Perform nonlinear mapping processing on the influence intensity and the primary weight coefficient to generate the first layer weight value, the second layer weight value and the third layer weight value corresponding to structural risk, material degradation risk and instantaneous failure risk, respectively.

[0110] It should be noted that stress conditions include thermal expansion and contraction, and wind and snow loads.

[0111] In this embodiment of the invention, the properties of different stress conditions and their historical failure records are first analyzed to assess the degree of harm of each stress to the accelerated propagation of microcracks. Then, for the same stress, its risk data and propagation rate at multiple different time points are examined, and a basic importance coefficient for that stress is obtained through fusion calculation. Finally, this basic coefficient is intelligently and nonlinearly combined with the previously assessed degree of harm to generate independent weight values ​​representing three different risk types: structural damage, material aging failure, and instantaneous fracture, namely, the first-level weight value, the second-level weight value, and the third-level weight value. Through nonlinear fusion, the stress influence is transformed into a hierarchical risk weight system, so that the risk assessment results can not only reflect the overall risk level, but also accurately indicate the specific type and main source of the risk.

[0112] Preferably, step S4 specifically includes:

[0113] Step S41: Construct a three-dimensional weight space, whose orthogonal coordinate axes correspond to the first layer weight value, the second layer weight value, and the third layer weight value, respectively. Then, synthesize the first layer weight value, the second layer weight value, and the third layer weight value into a risk vector and calculate the magnitude and direction of the risk vector.

[0114] Step S42: Based on the direction of the risk vector, determine the main failure factors of the photovoltaic module, and based on the magnitude of the risk vector and historical data of the area where the photovoltaic module is located, assess the probability level and expected severity level of the main failure factors.

[0115] Step S43: Based on the probability level of the occurrence of the main failure factors and the expected hazard level, a spatial interpolation algorithm is used to generate a risk probability distribution map and a hazard level distribution map of the area to be detected, and high-risk clusters are identified.

[0116] Step S44: Call the report generation engine to overlay and render the information of high-risk clusters, risk probability distribution map and hazard degree distribution map with the photovoltaic module location map to generate a risk assessment report.

[0117] It should be noted that the length of the vector includes the total risk magnitude, and the direction of the vector includes the main risk factor.

[0118] In this embodiment of the invention, a spatial model is created, consisting of three dimensions: structural risk, material aging risk, and instantaneous fracture risk. The first, second, and third layer weight values ​​of each photovoltaic module are transformed into a vector in this space. The length and direction of this vector are calculated. Then, based on the vector direction, the most significant failure factor of the module is determined, i.e., whether it is a frame problem or a cell problem. Based on the vector length and historical operation and maintenance data of its location, the probability of occurrence and the potential damage caused by the most significant failure factor are predicted. Then, the evaluation results of all photovoltaic modules are summarized using spatial algorithms to generate a visual distribution map of the risk probability and severity of the entire photovoltaic power station, and areas where high-risk photovoltaic modules are concentrated are automatically identified. Finally, the analysis results are overlaid with the module location map by a reporting engine to generate a comprehensive evaluation report containing specific risk locations and corresponding response suggestions, enabling operation and maintenance personnel to quickly identify high-risk areas, accurately determine risk types, and formulate targeted maintenance strategies.

[0119] Furthermore, step S42 also includes:

[0120] Step S421: Based on the main failure factors, predict the number of cycles required for the crack in the photovoltaic module to expand from its current size to the size of complete failure under day and night thermal cycling.

[0121] Step S422: Combine the local climate data of the photovoltaic module to convert the number of cycles into actual time and estimate the remaining life of the photovoltaic module.

[0122] In this embodiment of the invention, after determining the main failure factors of the photovoltaic module, based on the principles of materials science and fracture mechanics, the number of thermal cycles required for the cracks on the module to grow from their current size to the critical size that leads to complete functional failure under the alternating day and night thermal cycling stress is predicted. Subsequently, the specific climate data of the photovoltaic power station location, i.e., the actual daily temperature variation range, is accessed to convert the above-mentioned abstract number of thermal cycles into actual days or years, thereby accurately predicting the remaining safe service life of the photovoltaic module under the current damage state, thus effectively avoiding power generation loss and safety risks caused by sudden module failure.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A material image analysis method for the microcrack characteristics of photovoltaic modules, characterized in that, Includes the following steps: Step S1: Obtain EL images of independently arranged photovoltaic modules in the area to be detected, perform preprocessing, and generate a dataset of hidden cracks to be analyzed; Step S2: Input the dataset of hidden cracks to be analyzed into a preset risk assessment model and extract multiple risk values ​​from the dataset of hidden cracks to be analyzed; Step S3: The expansion rate of the risk value under different stress conditions is weighted to generate a first-layer weight value, a second-layer weight value, and a third-layer weight value; Step S4: By fusing the first layer weight value, the second layer weight value, and the third layer weight value, output a risk assessment report for the photovoltaic modules arranged independently in multiple areas to be detected.

2. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Divide the area with multiple photovoltaic modules into multiple regions of equal size to generate the area to be tested; Step S12: Select the independently arranged photovoltaic modules in each of the areas to be detected; Step S13: Obtain EL images of the independently arranged photovoltaic modules through the image acquisition unit; Step S14: Identify hidden crack regions in the EL image and collect the hidden crack regions to generate a hidden crack dataset to be analyzed.

3. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 2, characterized in that, Step S14 specifically includes: Step S141: Extract the maximum length, average width, main grid lines, number per unit area, relative position of the main grid lines and fine grid lines, brightness value, shape of the hidden crack, and the proportion of broken fine grid lines in the hidden crack region; Step S142: The maximum length, average width, main grid line, number per unit area, relative position of the main grid line and the fine grid line, brightness value, shape of the crack, and the ratio of the number of broken fine grid lines in the crack region are marked as crack characteristic parameters. Step S143: Pack and collect the hidden crack feature parameters to generate the hidden crack dataset to be analyzed.

4. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that, Step S2 specifically includes: Step S21: Based on the historical hidden crack dataset, construct the mapping relationship between hidden crack feature parameters and expansion rate; Step S22: Train the mapping relationship between the hidden crack feature parameters and the propagation rate using a physical information neural network, and generate the preset risk assessment model; Step S23: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model; Step S24: Extract the primary data set of different hidden crack characteristic parameters and propagation rates from the hidden crack dataset to be analyzed in the preset risk assessment model; Step S25: If multiple different primary data sets exceed the warning line, they are classified into multiple risk values.

5. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 4, characterized in that, Step S24 specifically includes: Step S241: Input the dataset of hidden cracks to be analyzed into the preset risk assessment model and call the feature recognition module in the model; Step S242: Based on the feature recognition module, extract multiple feature parameters related to the microcracks one by one from the microcrack dataset to be analyzed. The feature parameters include the length, orientation, location and morphological features of the microcracks. Step S243: Call the propagation rate analysis module in the preset risk assessment model to calculate the propagation rate of the hidden crack associated with the multiple characteristic parameters; Step S244: Combine each extracted hidden crack feature parameter and its associated hidden crack propagation rate to generate the primary data set.

6. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 5, characterized in that, Step S244 specifically includes: Step S2441: Receive the asynchronous data stream of the hidden crack characteristic parameters and the spread rate; Step S2442: Based on preset data verification rules, perform matching verification on the asynchronous data stream and remove invalid data pairs that cannot be matched or have logical conflicts. Step S2443: Logically associate the valid data pairs that have passed the preset data verification rules, and encapsulate them into data objects; Step S2444: Assign a unique index to the data object and map it to its corresponding time-series position in the dataset to be analyzed; Step S2445: Collect all the unique indexes and add set metadata to generate the primary data set.

7. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 6, characterized in that, Step S2443 specifically includes: Step S24431: Based on the physical properties of the hidden crack characteristic parameters in the data pair, construct a hierarchical field including basic properties and derived properties; Step S24432: Logically bind the hierarchical field with the corresponding valid data pair, injecting data fingerprints and timestamps during the binding process to complete the encapsulation of the data object.

8. The material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that, Step S3 specifically includes: Step S31: Using the conditional attributes under different stress conditions and historical fault data, evaluate the intensity of the influence of stress conditions on the propagation rate. Step S32: Combine the risk values ​​of multiple time-series nodes under the same stress condition with the propagation rate to generate the primary weighting coefficients under the same stress condition. Step S33: Perform nonlinear mapping processing on the influence intensity and the primary weight coefficient to generate the first layer weight value, the second layer weight value and the third layer weight value corresponding to structural risk, material degradation risk and instantaneous failure risk, respectively.

9. A material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that, Step S4 specifically includes: Step S41: Construct a three-dimensional weight space, whose orthogonal coordinate axes correspond to the first layer weight value, the second layer weight value, and the third layer weight value, respectively. Then, synthesize the first layer weight value, the second layer weight value, and the third layer weight value into a risk vector and calculate the magnitude and direction of the risk vector. Step S42: Based on the direction of the risk vector, determine the main failure factors of the photovoltaic module, and based on the magnitude of the risk vector and historical data of the area where the photovoltaic module is located, assess the probability level and expected severity level of the main failure factors. Step S43: Based on the probability level and expected hazard level of the occurrence of the main failure factors, a spatial interpolation algorithm is used to generate a risk probability distribution map and a hazard level distribution map of the area to be detected, and high-risk clusters are identified. Step S44: Call the report generation engine to overlay and render the information of the high-risk cluster, risk probability distribution map and hazard degree distribution map with the photovoltaic module location map to generate a risk assessment report.

10. A material image analysis method for the microcrack characteristics of photovoltaic modules according to claim 9, characterized in that, Step S42 further includes: Step S421: Based on the main failure factors, predict the number of cycles required for the crack in the photovoltaic module to expand from its current size to the size of complete failure under day and night thermal cycling. Step S422: Combining the local climate data of the photovoltaic module, convert the number of cycles into actual time, and estimate the remaining lifespan of the photovoltaic module.

Images