Photovoltaic module fault diagnosis and performance degradation evaluation system for photovoltaic power station

Abstract

The invention relates to the technical field of photovoltaic fault diagnosis, in particular to a photovoltaic module fault diagnosis and performance degradation evaluation system for a photovoltaic power station, which comprises a data acquisition layer, a digital twin engine, an intelligent diagnosis core, a model dynamic optimization module and a decision output interface. The data acquisition layer acquires a component surface dust distribution image, working environment information and operation electrical parameters; constructing a photovoltaic module virtual model by a digital twin engine, and outputting a theoretical operation state vector; the intelligent diagnosis core performs mode recognition based on a fault mode library by comparing multi-dimensional deviation characteristics of actual data and theoretical vectors; the model dynamic optimization module verifies the diagnosis reliability through discrete distribution analysis, and dynamically optimizes model parameters and acquisition frequency; and the decision output interface generates a diagnosis report and adaptively corrects the fault mode library. According to the method, multi-source information fusion diagnosis is realized, external pollution and internal faults can be accurately distinguished, the accuracy and the operation and maintenance efficiency of fault diagnosis are remarkably improved, and meanwhile, self-learning and self-adaptive optimization capabilities are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic fault diagnosis, and particularly relates to a photovoltaic module fault diagnosis and performance degradation evaluation system for a photovoltaic power station. BACKGROUND

[0002] With the continuous deepening of the transformation of global energy structure to clean and low-carbon, photovoltaic power generation, as an important part of renewable energy, has a sustained and rapid growth in installed capacity. However, photovoltaic modules are exposed to complex and changeable natural environment for a long time, and their performance is affected by various factors such as dust deposition, material aging, hot spot effect, potential induced degradation, etc., resulting in a decrease in power generation efficiency and a loss of operating benefits. At present, the fault diagnosis and performance evaluation of photovoltaic modules in the industry mainly rely on regular manual inspection, infrared thermal imaging detection or analysis methods based on a single electrical parameter. These traditional methods have obvious limitations: manual inspection is inefficient and subjective; infrared detection can identify hot spots, but it is difficult to distinguish whether the cause is surface contamination or internal defect; analysis based on electrical parameters often cannot effectively separate the interference of environmental factors, resulting in a high misjudgment rate. In addition, the existing technology generally has problems such as single monitoring dimension, isolated data analysis, lack of dynamic optimization capability, etc., and cannot establish a complete closed loop from state perception to fault diagnosis to optimization decision, making it difficult to meet the urgent needs of intelligent operation and maintenance of large-scale photovoltaic power stations. Therefore, developing a new system that can integrate multi-source information and achieve accurate diagnosis and adaptive evaluation has become a key technical challenge to improve the operation and maintenance level of photovoltaic power stations. SUMMARY

[0003] To this end, the present application provides a photovoltaic module fault diagnosis and performance degradation evaluation system for a photovoltaic power station to overcome the problem of high misjudgment rate and inability to accurately distinguish between external contamination and internal fault due to reliance on a single data source and lack of multi-dimensional information fusion capability in the prior art.

[0004] To achieve the above-mentioned purpose, the present application provides a photovoltaic module fault diagnosis and performance degradation evaluation system for a photovoltaic power station, comprising:

[0005] a data acquisition layer for acquiring surface dust distribution images of each photovoltaic module of the photovoltaic power station, simultaneously collecting working environment information of the photovoltaic power station, and reading output current and output voltage of each photovoltaic module;

[0006] a digital twin engine for constructing a virtual model of each photovoltaic module in a virtual space, updating virtual dust distribution and virtual environment load in the virtual model at a predetermined simulation frequency according to each surface dust distribution image and working environment information transmitted by the data acquisition layer, and outputting a theoretical operating state vector of the virtual model in the current state;

[0007] an intelligent diagnosis core configured to fuse and compare actual operation electrical data read by the data acquisition layer with the theoretical operation state vector output by the digital twin engine to obtain a multi-dimensional deviation feature vector; and perform pattern recognition on the multi-dimensional deviation feature vector based on a predefined fault mode library;

[0008] a model dynamic optimization module configured to perform discrete distribution analysis based on the multi-dimensional deviation feature vector output by the intelligent diagnosis core, and correct the simulation process of the virtual model of the digital twin engine based on the pattern recognition according to the result of the discrete distribution analysis, including,

[0009] adjusting a parameter calibration strategy of the virtual model according to the result of the pattern recognition;

[0010] or, adjusting a collection frequency of the dust distribution image on the surface of the photovoltaic module and an update frequency of virtual dust distribution in the virtual model according to the result of the pattern recognition;

[0011] a decision output interface configured to output a diagnosis report including a fault type identifier and a maintenance operation instruction based on the result of the pattern recognition and the optimized state of the virtual model, and adjust the recognition standard in the fault mode library based on the result of the discrete distribution analysis.

[0012] As a preferred technical solution of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the data acquisition layer includes an environment monitoring unit, an electrical monitoring unit and a spatial imaging unit deployed on the site of the photovoltaic power station:

[0013] the spatial imaging unit is configured to obtain a high-resolution surface dust distribution image of the surface of the photovoltaic module;

[0014] the environment monitoring unit is configured to monitor the working environment information in real time, and the working environment information includes environmental parameters such as irradiance, ambient temperature and backboard temperature of the module;

[0015] the electrical monitoring unit is configured to monitor the output current and output voltage of the photovoltaic module in real time.

[0016] As a preferred technical solution of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the digital twin engine is configured to calculate a set of theoretical electrical parameters including at least theoretical maximum power point power, theoretical open circuit voltage and theoretical short circuit current through the virtual model simulation based on the surface dust distribution image and the working environment information;

[0017] combine the set of theoretical electrical parameters into the theoretical operation state vector.

[0018] As a preferred technical scheme of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the intelligent diagnosis core is configured to perform the following operations: extracting an actual maximum power point power, an actual open circuit voltage and an actual short circuit current from the actual operation electrical data;

[0019] Comparing the extracted actual parameters with corresponding theoretical parameters in the theoretical operation state vector to generate the multi-dimensional deviation feature vector containing power deviation, voltage deviation and current deviation.

[0020] As a preferred technical scheme of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the fault mode library is configured to perform the following mode recognition:

[0021] When the numerical signs of the power deviation and the current deviation are negative, and the numerical amplitudes thereof are positively correlated with the spatial density distribution of the surface dust distribution image, the surface dust pollution is recognized as cleanable;

[0022] When the infrared thermal imaging data indicates that there is a local hot spot, and the surface dust distribution image shows low dust density in the region corresponding to the hot spot, the uncleanability fault caused by hidden cracks or internal defects is recognized;

[0023] When the numerical sign of the voltage deviation is negative and the absolute value thereof is greater than the absolute value of the preset current deviation, the abnormal internal electrical connection of the module is recognized;

[0024] When the surface dust distribution image presents regular accumulation along the edges of the cell string or the grid line, and the numerical signs of the power deviation and the current deviation in the multi-dimensional deviation feature vector are negative, it is recognized that there is a potential induced attenuation risk.

[0025] As a preferred technical scheme of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the model dynamic optimization module performs discrete distribution analysis on the power deviation, voltage deviation and current deviation in the multi-dimensional deviation feature vector, and calculates the discrete degree index of each deviation element;

[0026] When the discrete degree index of each deviation element is lower than the corresponding preset threshold, it is determined that the recognition result of the fault mode library is accurate;

[0027] When the discrete degree index of any deviation element is higher than the corresponding predetermined threshold, it is determined that the recognition result has uncertainty.

[0028] As a preferred technical scheme of the photovoltaic module fault diagnosis and performance degradation evaluation system for the photovoltaic power station, the model dynamic optimization module is configured to perform the following control:

[0029] When the pattern recognition result indicates that the surface is cleanable dust contamination, the acquisition frequency of the surface dust distribution image and the update frequency of the virtual dust distribution are adjusted according to the change gradient of the surface dust distribution image.

[0030] When the pattern recognition result is an uncleanable fault or an abnormal electrical connection inside the component, the calibration of the series resistance and parallel resistance parameters in the virtual model is triggered.

[0031] When the pattern recognition result indicates a risk of potential-induced decay, the calibration of the insulation resistance and system voltage bias parameters in the virtual model is triggered.

[0032] As a preferred technical solution for a photovoltaic module fault diagnosis and performance degradation assessment system used in photovoltaic power plants, the model dynamic optimization module corrects the simulation process based on the results of discrete distribution analysis;

[0033] In response to accurate recognition results, subsequent control is performed based on the pattern recognition results;

[0034] In response to the uncertainty of the identification results, the following operations are triggered: calibrating the series resistance, parallel resistance, insulation resistance, and system voltage bias parameters in the virtual model; increasing the acquisition frequency of the surface dust distribution image; increasing the simulation frequency of the virtual model; and using this data as retraining data for the fault mode library.

[0035] As a preferred technical solution for a photovoltaic module fault diagnosis and performance degradation assessment system for photovoltaic power plants, the decision output interface receives the pattern recognition results output by the intelligent diagnosis core to determine the fault type identifier, and generates maintenance operation instructions by combining the virtual model optimization status output by the model dynamic optimization module.

[0036] The fault type identifier is combined with the corresponding maintenance operation instructions to generate a diagnostic report and output it.

[0037] As a preferred technical solution for a photovoltaic module fault diagnosis and performance degradation assessment system for photovoltaic power plants, the decision output interface analyzes the deviation of the deviation element dispersion index and the preset threshold based on the discrete distribution analysis results output by the model dynamic optimization module, generates an adjustment factor, and applies the adjustment factor to the identification standard parameters of the corresponding fault type in the fault mode library to update the threshold range or weight coefficient, thereby realizing the adaptive correction of the fault mode library.

[0038] Synchronously record the update history of the fault mode library and associate the current adjustment factor with the historical adjustment records.

[0039] Compared with existing technologies, the advantages of this invention lie in that the system integrates surface dust distribution images, environmental parameters, and electrical data, establishes a virtual model using digital twin technology, generates a multi-dimensional deviation feature vector by comparing theoretical operating state vectors with actual operating data, and achieves accurate fault type identification based on a predefined fault mode library. The model dynamic optimization module verifies the reliability of the diagnostic results through discrete distribution analysis and dynamically adjusts model parameters and acquisition frequency based on the identification results, forming a closed-loop optimization mechanism. The decision output interface transforms the diagnostic results into executable maintenance instructions and adaptively corrects the identification criteria of the fault mode library based on operational feedback. This system represents a leap from single electrical diagnostics to multi-physics field integrated diagnostics, effectively distinguishing between external contamination and internal faults, significantly improving the accuracy and reliability of fault diagnosis. Through adaptive model optimization and parameter calibration, it ensures the accuracy of the digital twin model under different operating conditions. Furthermore, the system possesses self-verification and self-evolution capabilities, continuously optimizing diagnostic criteria based on actual operating data. This effectively overcomes the limitations of fixed parameter systems in adapting to complex and changing field conditions, providing a complete technical solution for the intelligent operation and maintenance of photovoltaic power plants, and significantly improving operation and maintenance efficiency and economy. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of a photovoltaic module fault diagnosis and performance degradation assessment system for photovoltaic power plants according to an embodiment of the present invention;

[0041] Figure 2 This is a logic diagram showing whether the model dynamic optimization module corrects the simulation process in an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0043] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0044] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0045] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0046] Please see Figure 1 The diagram shown is a structural schematic of a photovoltaic module fault diagnosis and performance degradation assessment system for photovoltaic power plants according to an embodiment of the present invention. The present invention provides a photovoltaic module fault diagnosis and performance degradation assessment system for photovoltaic power plants, characterized in that it includes:

[0047] The data acquisition layer is used to acquire images of the surface dust distribution of each photovoltaic module in the photovoltaic power station, as well as to collect information on the working environment of the photovoltaic power station and read the output current and output voltage of each photovoltaic module.

[0048] A digital twin engine is used to construct virtual models of each photovoltaic module in virtual space. Based on the dust distribution images of each surface and the working environment information transmitted by the data acquisition layer, the virtual dust distribution and virtual environmental load are updated in the virtual model at a predetermined simulation frequency, and the theoretical operating state vector of the virtual model in the current state is output.

[0049] The intelligent diagnostic core is used to fuse and compare the actual operating electrical data read by the data acquisition layer with the theoretical operating state vector output by the digital twin engine to obtain a multi-dimensional deviation feature vector; and to perform pattern recognition on the multi-dimensional deviation feature vector based on a predefined fault mode library.

[0050] The model dynamic optimization module is used to perform discrete distribution analysis based on the multidimensional deviation feature vector output by the intelligent diagnostic core, and to correct the simulation process of the pattern recognition-based digital twin engine-optimized virtual model according to the results of the discrete distribution analysis. This includes...

[0051] The parameter calibration strategy of the virtual model is adjusted based on the pattern recognition results;

[0052] Alternatively, the acquisition frequency of the dust distribution image on the surface of the photovoltaic module and the update frequency of the virtual dust distribution in the corresponding virtual model may be adjusted based on the pattern recognition results.

[0053] The decision output interface is used to output a diagnostic report containing fault type identifiers and maintenance operation instructions based on the pattern recognition results and the optimization state of the virtual model, and to adjust the identification criteria in the fault mode library based on the discrete distribution analysis results.

[0054] In practice, the predetermined simulation frequency is dynamically set based on historical data analysis and system resource constraints.

[0055] Specifically, the data acquisition layer includes an environmental monitoring unit, an electrical monitoring unit, and a spatial imaging unit deployed at the photovoltaic power plant site:

[0056] The spatial imaging unit is configured to acquire high-resolution images of surface dust distribution on the surface of the photovoltaic module;

[0057] The environmental monitoring unit is used to monitor the working environment information in real time, including environmental parameters such as irradiance, ambient temperature, and component backplane temperature.

[0058] The electrical monitoring unit is used to monitor the output current and output voltage of the photovoltaic module in real time.

[0059] In practice, the actual operating electrical data includes the output current and output voltage of the photovoltaic modules.

[0060] The spatial imaging unit is a high-resolution multispectral camera deployed on inspection drones or fixed monitoring points. It captures the reflection and scattering characteristics of light of different wavelengths on the surface of the components to generate grayscale or optical density images that can quantify the degree and distribution of dust obstruction. The environmental monitoring unit, based on the physical characteristics of photovoltaic power generation, uses irradiance sensors, ambient temperature sensors, and temperature probes mounted on the back panel of the components to simultaneously acquire the core environmental variables that affect the power generation efficiency of the components. The electrical monitoring unit, as the link that directly senses the power generation status, uses high-precision current and voltage sensors to sample the output circuit of the components in real time to obtain its most direct working electrical parameters.

[0061] In this invention, the dust distribution images provided by the spatial imaging unit transform the difficult-to-quantify surface contamination status into analyzable visual data; the irradiance and temperature parameters captured by the environmental monitoring unit provide accurate model-driven inputs for the digital twin engine; and the real-time output voltage and current acquired by the electrical monitoring unit serve as a true benchmark for evaluating the actual operating performance of the components, enabling the system to accurately correlate component performance anomalies with specific environmental disturbances or their own faults, thereby laying a solid data foundation for subsequent accurate diagnosis and optimization.

[0062] Specifically, the digital twin engine is configured to simulate and calculate a set of theoretical electrical parameters based on the surface dust distribution image and the working environment information through the virtual model. The theoretical electrical parameters include at least the theoretical maximum power point power, the theoretical open-circuit voltage, and the theoretical short-circuit current.

[0063] The set of theoretical electrical parameters is combined into the theoretical operating state vector.

[0064] Understandably, by establishing a multiphysics virtual model that integrates electrical characteristics and optical effects, and by mapping external environmental parameters to the theoretical electrical output of components through dynamic simulation, an accurate comparison benchmark can be provided for subsequent fault diagnosis.

[0065] In implementation, the virtual model uses a photovoltaic cell engineering mathematical model built into the digital twin engine as its computational core. This model can simulate the inherent power generation characteristics of photovoltaic cells based on input irradiance and temperature conditions. The system analyzes and quantifies high-resolution surface dust distribution images acquired by the data acquisition layer, converting the grayscale or optical density information of different regions in the image into an assessment of the local transmittance of the module surface, and then calculates the overall attenuation degree of dust on incident irradiance. The attenuated effective irradiance and the module backsheet temperature are used together as key input parameters to drive the virtual model.

[0066] The mathematical model for photovoltaic cell engineering employs a single-diode / dual-diode equivalent circuit model or a high-precision data-driven model (such as a neural network) to quantify the influence of irradiance (G) and temperature (T) on the IV characteristics of photovoltaic cells. The mathematical expression is as follows:

[0067]

[0068] Where I is the output current, and the photogenerated current is the photocurrent. The diode saturation current is linearly related to irradiance. Related to the temperature index, V is the output voltage of the photovoltaic cell. Here, n is the series resistance, and n is the diode ideality factor. Thermoelectric voltage, It is a parallel resistor.

[0069] Image processing algorithms, such as grayscale thresholding and optical density inversion, are used to convert surface dust distribution images into local transmittance distribution maps, establishing a quantitative relationship between grayscale value and transmittance. .

[0070] Comprehensive Irradiance Attenuation Calculation: Calculating the Effective Irradiance of Each Region on the Component Surface Based on Transmittance Distribution And integrate it into the global effective irradiance input.

[0071] Effective irradiance With backplate temperature As dual input variables, the virtual model is driven to simulate the power generation characteristic IV curve of photovoltaic cells in real time, achieving real-time synchronization with the physical entity.

[0072] During the simulation phase, the digital twin engine loads the dust-corrected effective irradiance and real-time temperature into the virtual model. Based on the model's built-in physical formulas, the simulation calculates the theoretical short-circuit current and theoretical open-circuit voltage that an ideal, fault-free component should output under the corresponding operating conditions, and further deduces the theoretical maximum power point power. Finally, the engine encapsulates these three core theoretical electrical parameters into a structured theoretical operating state vector and outputs it.

[0073] Through the above implementation methods, the system transforms the impact of dust distribution, which is difficult to quantify directly, into accurately calculable electrical parameter deviations. This provides the core of intelligent diagnostics with a pure, ideal state reference system that reflects only the effects of dust and the environment. This allows the diagnostic process to effectively isolate external environmental interference, thereby more accurately revealing performance degradation caused by component aging, microcracks, or internal defects, ultimately improving the accuracy and reliability of fault diagnosis.

[0074] Specifically, the intelligent diagnostic core is configured to perform the following operations: extract the actual maximum power point power, actual open-circuit voltage, and actual short-circuit current from the actual operating electrical data;

[0075] The extracted actual parameters are compared with the corresponding theoretical parameters in the theoretical operating state vector to generate the multidimensional deviation feature vector containing power deviation, voltage deviation, and current deviation.

[0076] In implementation, by quantitatively comparing the actual operating state of the components with the theoretical ideal state, complex performance phenomena are transformed into categorizable fault modes, thereby achieving precise fault location and classification. From the raw electrical data monitored in real time, three key actual parameters that can characterize the power generation capacity of the components are analyzed and calculated: actual maximum power point power, actual open-circuit voltage, and actual short-circuit current. Subsequently, the diagnostic core compares these three actual parameters one by one with the corresponding theoretical parameters provided by the digital twin engine, which are simulated under the same environmental and dust conditions. By calculating their differences, a structured multidimensional deviation feature vector is generated.

[0077] This vector essentially decomposes component performance anomalies into quantified deviations across three orthogonal dimensions: power, voltage, and current. Ultimately, the system matches this deviation vector against a predefined fault mode library. This library is essentially a classifier built upon the physical laws of photovoltaic electrophysics, pre-defined with characteristic deviation combinations specific to different fault types. By identifying which pre-defined feature best matches the current deviation vector, the system outputs a clear and standardized fault type identifier.

[0078] Through the above implementation methods, the system can transform fault information hidden in complex operational data into clear and actionable diagnostic conclusions. This enables the system to move from perceiving performance degradation to diagnosing the cause of the fault, transforming operation and maintenance decisions from experience-based guesswork to data-driven, precise judgment. Consequently, it can trigger the most effective maintenance response for different types of faults, significantly improving the intelligence level and response efficiency of photovoltaic power plant operation and maintenance.

[0079] Specifically, the fault mode library is configured to perform the following pattern recognition:

[0080] When the power deviation and current deviation are negative and their magnitudes are positively correlated with the spatial density distribution of the surface dust distribution image, they are identified as cleanable surface dust contamination.

[0081] When infrared thermal imaging data indicates the presence of local hot spots, and the surface dust distribution image shows low dust density in the area corresponding to the hot spots, it is identified as an uncleanable fault caused by microcracks or internal defects.

[0082] When the sign of the voltage deviation is negative and its absolute value is greater than the absolute value of the preset current deviation, it is identified as an abnormal electrical connection inside the component.

[0083] When the surface dust distribution image shows regular accumulation along the edge of the battery string or grid line, and the power deviation and current deviation in the multidimensional deviation feature vector are negative, it is identified as having a risk of potential-induced degradation.

[0084] It is understandable that diagnostic criteria are constructed based on the differences in electrical characteristics and physical manifestations of different fault mechanisms. For cleanable surface dust contamination, the identification principle is that dust mainly blocks light, affecting the generation of photocurrent, thus manifesting as a synchronous decrease in power and current, and the degree of decrease is consistent with the spatial distribution of dust density in image recognition;

[0085] For non-cleanable faults caused by microcracks or internal defects, the identification principle is that such defects will cause local current mismatch and heat generation, but the surface dust coverage of the hot spot area shown on the infrared image is not serious. This mismatch between the electrical thermal phenomenon and the optical phenomenon is the key to diagnosis.

[0086] The identification principle for abnormal electrical connections inside the component is that connection problems will increase the series resistance, which is mainly manifested as a significant reduction in output voltage, while the current is relatively less affected. Thus, the deviation vector shows a characteristic pattern in which the voltage deviation is much greater than the current deviation.

[0087] The identification principle for potential-induced degradation risk lies in the fact that abnormal potential distribution on the component surface will electrostatically attract dust, causing dust to accumulate regularly along the edges of the battery string or grid lines. At the same time, it is accompanied by a general decrease in power and current. This unique dust distribution pattern and the coexistence of electrical performance degradation constitute the basis for diagnosis.

[0088] In implementation, the system scans the surface of photovoltaic modules using an infrared thermal imager. When a region's temperature exceeds the module's average temperature and the temperature difference reaches a set threshold, it is identified as a hot spot. The identification process includes the following criteria: the temperature of the hot spot must be significantly higher than the surrounding normal area, or the area of ​​the hot spot must reach a specific proportion of the module's total area. This dual-criteria mechanism ensures the accuracy of hot spot identification, eliminating instantaneous temperature fluctuations caused by environmental factors while effectively capturing persistent localized overheating caused by microcracks or internal defects.

[0089] The criterion for judgment is a temperature difference of 10-20°C between the module's average temperature and the surrounding normal cell temperature. The International Electrotechnical Commission (IEC) standard IEC 61215 stipulates that photovoltaic modules must pass a hotspot durability test. This test essentially verifies that under localized shading, the hotspot temperature will not cause permanent damage to the module. The long-term heat resistance temperature of photovoltaic module encapsulation materials typically has an upper limit. Sustained excessively high local temperatures will accelerate material aging, yellowing, and even delamination. A temperature difference of 10°C is a significant anomaly signal, while a temperature difference exceeding 20°C indicates high risk and can easily lead to irreversible damage.

[0090] The system calculates the ratio of the absolute value of the voltage deviation to the absolute value of the current deviation. When this ratio exceeds the preset absolute value of the current deviation, it is determined that the voltage deviation is significantly greater than the current deviation. Based on the fault characteristics of photovoltaic modules, internal electrical connection abnormalities mainly affect voltage characteristics, while external factors such as dust contamination mainly affect current characteristics.

[0091] The preset absolute value of the current deviation is the ratio of the absolute value of the voltage deviation to the absolute value of the current deviation, which is greater than 1.5. Increased series resistance caused by factors such as microcracks and broken solder strips will linearly lower the maximum power point voltage and open-circuit voltage. For example, a 20% power loss may stem from a 2% current drop and an 18% voltage drop, resulting in a ratio as high as 9.0. Through extensive IV curve testing and data analysis on components with known faults, the probability distribution of this ratio under different fault types can be statistically determined. A ratio greater than 1.5 is a statistically critical point that can distinguish series resistance faults from parallel resistance or dust faults with a high probability. This means that when the voltage drop is more than 1.5 times the current drop, the system has a very high probability of identifying it as an internal connection problem.

[0092] The system processes surface dust distribution images as follows: it identifies battery string boundaries and grid line locations using an edge detection algorithm, and then analyzes the dust distribution characteristics in these specific areas. When the dust distribution density near the battery string edges or grid lines is significantly higher than in other areas, and this distribution pattern repeats on the component surface, it is determined to be regular accumulation. This identification process comprehensively considers multiple features such as the spatial continuity, directional consistency, and density concentration of dust distribution to ensure accurate capture of characteristic dust distribution patterns caused by potential-induced decay.

[0093] The dust density in the battery string or grid edge areas is significantly higher than in other areas, and is more than 1.5 times the average dust density in the module center area.

[0094] The physical phenomenon of electrostatic adsorption causes the module to generate a high voltage to ground, forming an electrostatic field on the glass surface. Dust particles are thus polarized and adsorbed in the region of strongest electric field, namely the edges of the cell strings and near the grid lines. By analyzing a large number of module images known to have PID risk using computer vision algorithms, the intensity of this adsorption effect can be quantified. The 1.5x threshold is a decision boundary that can effectively distinguish between the two types of patterns, derived by training a machine learning model on two types of images: normal uniform dust accumulation and PID-characteristic dust accumulation.

[0095] Through the above implementation methods, the system achieves a leap from single electrical diagnostics to multi-physics fusion diagnostics. It can effectively distinguish between external contamination and internal damage, and differentiate between connection faults and material degradation, thus providing accurate fault classification information for operation and maintenance decisions. This not only avoids misjudging internal faults as surface contamination and delaying maintenance, but also promptly detects potential degradation risks, significantly improving the accuracy of fault diagnosis and the targeting of operation and maintenance strategies.

[0096] Please see Figure 2As shown, it is a logic diagram of whether the model dynamic optimization module corrects the simulation process in an embodiment of the present invention. The model dynamic optimization module performs discrete distribution analysis on the power deviation, voltage deviation and current deviation in the multidimensional deviation feature vector and calculates the discreteness index of each deviation element.

[0097] When the dispersion index of each deviation element is lower than the corresponding preset threshold, the identification result of the fault mode library is determined to be accurate.

[0098] When the dispersion index of any deviation element is higher than the corresponding predetermined threshold, it is determined that the identification result has uncertainty.

[0099] In implementation, the system continuously collects multidimensional deviation feature vectors for multiple diagnostic cycles within a set time window, and calculates the dispersion index of three elements: power deviation, voltage deviation, and current deviation. This dispersion index is calculated as the ratio of the average deviation to the average value; the degree of data dispersion is quantified by calculating the ratio of the average deviation of each deviation element to its average value. The time frame is set to one day.

[0100] The specific thresholds for the dispersion of each deviation element were determined through a limited number of tests using historical typical fault case data. By statistically analyzing the fluctuation range of known accurately diagnosed cases, an upper limit of dispersion that could cover the vast majority of stable fault cases was selected. The thresholds for voltage and current deviations were determined independently based on their inherent fluctuation characteristics under different fault modes, using a similar method. The determination of these thresholds employed statistical regression analysis based on historical fault data to ensure that the threshold settings conformed to the actual physical characteristics of each electrical parameter.

[0101] Through the above implementation methods, a self-verification closed loop for diagnostic results is constructed, which can automatically identify unreliable diagnoses caused by environmental transients, measurement noise, or complex compound faults. By triggering data re-acquisition, model parameter calibration, and other coping strategies, it effectively avoids making erroneous operation and maintenance decisions based on uncertain diagnostic results, and significantly improves the practicality and reliability of the entire system under actual complex working conditions.

[0102] Specifically, the model dynamic optimization module is configured to perform the following controls:

[0103] When the pattern recognition result indicates that the surface is cleanable dust contamination, the acquisition frequency of the surface dust distribution image and the update frequency of the virtual dust distribution are adjusted according to the change gradient of the surface dust distribution image.

[0104] When the pattern recognition result is an uncleanable fault or an abnormal electrical connection inside the component, the calibration of the series resistance and parallel resistance parameters in the virtual model is triggered.

[0105] When the pattern recognition result indicates a risk of potential-induced decay, the calibration of the insulation resistance and system voltage bias parameters in the virtual model is triggered.

[0106] In implementation, the control strategy for the model dynamic optimization module was refined, and an adaptive optimization mechanism linking fault diagnosis results was established. By dynamically adjusting system operating parameters according to different fault types, the accuracy of the digital twin model and the system operating efficiency were simultaneously optimized. The changing gradient was obtained through difference calculation of continuous time series images.

[0107] When the system detects cleanable surface dust contamination, the optimization module analyzes continuously acquired surface dust distribution images, calculates the gradient of its change over time, and dynamically adjusts the image acquisition frequency and the update frequency of dust distribution in the virtual model based on this gradient value. The change in dust accumulation rate directly reflects the change in environmental conditions. By adaptively adjusting the frequency, it can ensure timely response to rapid dust accumulation and save system resources under stable conditions.

[0108] When diagnostic results indicate non-cleanable faults or internal connection abnormalities, the calibration process for series and parallel resistance parameters in the virtual model is performed. Such faults directly change the internal impedance characteristics of the component. Through parameter calibration, the virtual model can be re-fitted to the actual degradation state of the component.

[0109] When the risk of potential-induced decay is identified, the insulation resistance and system voltage bias parameters in the virtual model are calibrated. Potential-induced decay will change the insulation characteristics of the component and the physical characteristics of the operating voltage point.

[0110] Through the above implementation methods, the system achieves closed-loop linkage between diagnosis and optimization, enabling the digital twin model to perform targeted self-correction based on the characteristics of different fault types. This not only ensures the accuracy of the model under different operating conditions but also improves the overall system efficiency through resource allocation optimization.

[0111] Specifically, the model dynamic optimization module corrects the simulation process based on the results of discrete distribution analysis;

[0112] In response to accurate recognition results, subsequent control is performed based on the pattern recognition results;

[0113] In response to the uncertainty of the identification results, the following operations are triggered: calibrating the series resistance, parallel resistance, insulation resistance, and system voltage bias parameters in the virtual model; increasing the acquisition frequency of the surface dust distribution image; increasing the simulation frequency of the virtual model; and using this data as retraining data for the fault mode library.

[0114] Specifically, the decision output interface receives the pattern recognition result output by the intelligent diagnostic core to determine the fault type identifier, and generates maintenance operation instructions by combining the virtual model optimization status output by the model dynamic optimization module.

[0115] The fault type identifier is combined with the corresponding maintenance operation instructions to generate a diagnostic report and output it.

[0116] In implementation, based on a predefined strategy library and a command mapping mechanism using historical maintenance records, the decision output interface, upon receiving the pattern recognition results from the intelligent diagnostic core, first determines the corresponding fault type identifier according to the built-in fault type-operation mapping table. This mapping table contains preset standard maintenance schemes based on historical maintenance records. Subsequently, the system queries effective handling measures for similar fault scenarios in historical maintenance records, and, combined with the virtual model optimization state provided by the model dynamic optimization module, matches the most suitable maintenance operation command for the current system state from the preset maintenance strategy library. The principle behind this mapping process is that different types of faults require differentiated handling methods, and historical maintenance records provide effective solutions that have been proven in practice. The combination of the two ensures that the generated maintenance commands not only comply with fault handling specifications but also incorporate actual operation and maintenance experience.

[0117] Through the above implementation methods, the system constructs a standardized and structured diagnostic output system, which transforms complex diagnostic data into clear instructions that on-site maintenance personnel can directly understand and execute. This effectively eliminates the cognitive gap between diagnostic information and maintenance operations, significantly improves the accuracy and efficiency of fault handling, and provides technical support for the standardization and normalization of maintenance management.

[0118] Specifically, the decision output interface analyzes the deviation between the discrete distribution analysis results output by the model dynamic optimization module and the deviation of the deviation element dispersion index and the preset threshold to generate an adjustment factor. The adjustment factor is then applied to the identification standard parameters of the corresponding fault type in the fault mode library to update the threshold range or weight coefficient, thereby achieving adaptive correction of the fault mode library.

[0119] Synchronously record the update history of the fault mode library and associate the current adjustment factor with the historical adjustment records.

[0120] In implementation, the decision output interface obtains the quantitative deviation of the dispersion index of each deviation element from the preset threshold by parsing the discrete distribution analysis results output by the dynamic optimization module of the parsing model. This deviation is then combined with the effective parameter range in historical adjustment records to generate corresponding adjustment factors. These adjustment factors are subsequently applied to the identification standard parameters of the corresponding fault type in the fault mode library, thereby achieving precise correction of the diagnostic standards. The system also establishes a complete version management mechanism, continuously recording the adjustment factors and historical context of each parameter adjustment, forming a traceable self-optimization trajectory.

[0121] Through the above implementation methods, the fault mode library can dynamically adjust the identification criteria according to changes in the actual operating environment and component status, effectively overcoming the problem of insufficient adaptability of the fixed parameter system when facing complex and ever-changing field conditions, significantly improving the accuracy and reliability of the system's long-term operation, and realizing continuous optimization of diagnostic performance.

[0122] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0123] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A photovoltaic module fault diagnosis and performance degradation assessment system for a photovoltaic power plant, characterized in that, Comprise: a data acquisition layer configured to acquire surface dust distribution images of each photovoltaic module of a photovoltaic power station, collect working environment information of the photovoltaic power station, and read output current and output voltage of each photovoltaic module; a digital twin engine configured to construct a virtual model of each photovoltaic module in a virtual space, update virtual dust distribution and virtual environment load in the virtual model at a predetermined simulation frequency according to the surface dust distribution images and the working environment information transmitted by the data acquisition layer, and output a theoretical operating state vector of the virtual model in a current state; an intelligent diagnosis core configured to fuse and compare actual operating electrical data read by the data acquisition layer with the theoretical operating state vector output by the digital twin engine to obtain a multi-dimensional deviation feature vector, and perform pattern recognition on the multi-dimensional deviation feature vector based on a predefined fault mode library; a model dynamic optimization module configured to perform discrete distribution analysis based on the multi-dimensional deviation feature vector output by the intelligent diagnosis core, correct a simulation process of the virtual model of the digital twin engine based on the pattern recognition according to a result of the discrete distribution analysis, including, adjusting a parameter calibration strategy of the virtual model according to the pattern recognition result; or, adjusting a collection frequency of the surface dust distribution images of the photovoltaic module and an update frequency of the virtual dust distribution in the corresponding virtual model according to the pattern recognition result; a decision output interface configured to output a diagnosis report containing a fault type identification and a maintenance operation instruction based on the pattern recognition result and an optimized state of the virtual model, and adjust identification criteria in the fault mode library based on a result of the discrete distribution analysis.

2. The photovoltaic module fault diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 1, characterized in that, The data acquisition layer comprises an environment monitoring unit, an electrical monitoring unit, and a space imaging unit deployed on site of the photovoltaic power station: The space imaging unit is configured to acquire high-resolution surface dust distribution images of a surface of the photovoltaic module; The environment monitoring unit is configured to monitor the working environment information in real time, and the working environment information includes environmental parameters such as irradiance, ambient temperature, and backboard temperature of the module; The electrical monitoring unit is configured to monitor the output current and the output voltage of the photovoltaic module in real time.

3. The photovoltaic module fault diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 2, characterized in that, The digital twin engine is configured to calculate a set of theoretical electrical parameters including at least theoretical maximum power point power, theoretical open circuit voltage, and theoretical short circuit current through the virtual model simulation based on the surface dust distribution images and the working environment information; The set of theoretical electrical parameters are combined into the theoretical operating state vector.

4. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 3, characterized in that, The intelligent diagnosis core is configured to perform the following operations: extract actual maximum power point power, actual open circuit voltage, and actual short circuit current from the actual operating electrical data; compare the extracted actual parameters with corresponding theoretical parameters in the theoretical operating state vector to generate the multi-dimensional deviation feature vector containing power deviation, voltage deviation, and current deviation.

5. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 4, characterized in that, The fault mode library is configured to perform the following pattern recognition: when the numerical sign of the power deviation and the current deviation is negative, and the numerical amplitude thereof is positively correlated with the spatial density distribution of the surface dust distribution image, the surface dust pollution is identified as cleanable; when the infrared thermal imaging data indicates that there is a local thermal spot, and the surface dust distribution image shows low dust density in the region corresponding to the thermal spot, the uncleanability failure caused by hidden cracks or internal defects is identified; when the numerical sign of the voltage deviation is negative and the absolute value thereof is greater than the absolute value of the preset current deviation, the internal electrical connection abnormality of the component is identified; when the surface dust distribution image presents regular accumulation along the edges of the battery string or grid line, and the numerical signs of the power deviation and the current deviation in the multi-dimensional deviation feature vector are negative, it is identified that there is a potential induced attenuation risk.

6. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 5, characterized in that, The model dynamic optimization module performs discrete distribution analysis on the power deviation, voltage deviation and current deviation in the multi-dimensional deviation feature vector, and calculates the discrete degree index of each deviation element; when the discrete degree index of each deviation element is lower than the corresponding preset threshold, it is determined that the identification result of the failure mode library is accurate; when the discrete degree index of any deviation element is higher than the corresponding predetermined threshold, it is determined that the identification result is uncertain.

7. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 6, characterized in that, The model dynamic optimization module is configured to perform the following controls: when the mode recognition result is cleanable surface dust pollution, the acquisition frequency of the surface dust distribution image and the update frequency of the virtual dust distribution are adjusted according to the change gradient of the surface dust distribution image; when the mode recognition result is uncleanability failure or internal electrical connection abnormality of the component, calibration of the series resistance and parallel resistance parameters in the virtual model is triggered; when the mode recognition result is a potential induced attenuation risk, calibration of the insulation resistance and system voltage bias parameters in the virtual model is triggered. 8.The system for diagnosing faults and assessing performance degradation of photovoltaic modules of a photovoltaic power plant according to claim 6, wherein, The model dynamic optimization module corrects the simulation process according to the results of discrete distribution analysis; in response to the accurate identification result, subsequent control is performed according to the mode recognition result; in response to the uncertainty of the identification result, the following operations are triggered: calibration of the series resistance, parallel resistance, insulation resistance and system voltage bias parameters in the virtual model, increase of the acquisition frequency of the surface dust distribution image, increase of the simulation frequency of the virtual model, and use of this data as retraining data for the failure mode library.

9. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 8, characterized in that, The decision output interface receives the mode recognition result output by the intelligent diagnosis core to determine the fault type identifier, and generates a maintenance operation instruction in combination with the virtual model optimization state output by the model dynamic optimization module; The fault type identifier and the corresponding maintenance operation instruction are combined to generate a diagnosis report and output.

10. The photovoltaic module failure diagnosis and performance degradation assessment system for a photovoltaic power plant according to claim 9, characterized in that, The decision output interface generates an adjustment factor based on the deviation amount between the discrete degree index of the deviation element and the preset threshold value, and applies the adjustment factor to the identification standard parameters of the corresponding fault type in the failure mode library to update the threshold range or the weight coefficient, so as to realize adaptive correction of the failure mode library; Synchronize the update history of the failure mode library and store the current adjustment factor in association with the history adjustment record.

Images