A multi-parameter fusion-based photovoltaic module contamination grading early warning method

Abstract

The application provides a kind of photovoltaic module dirty grading early warning method based on multi-parameter fusion, belongs to new energy centralized photovoltaic power station technical field, can significantly alleviate or solve the problems of low accuracy, poor real-time and weak adaptability of existing photovoltaic module dirty early warning.The method comprises: collecting environmental parameters, historical dirty data and component power generation data, constructing regional dirty feature database;Collect and preprocess component surface temperature distribution data, environmental parameter data and surface image data;Identify the dirty area and quantify the dirty degree;Combine real-time environmental parameters to dynamically adjust the early warning threshold;Trigger the grading early warning and output cleaning suggestions or cleaning tasks;Verify the early warning effect after cleaning and optimize the associated model and dirty identification algorithm.The method can improve the early warning accuracy and operation efficiency, reduce the power generation loss.

Description

Technical Field

[0001] This invention belongs to the field of new energy centralized photovoltaic power station technology, specifically involving a method for classifying and early warning of pollution levels in photovoltaic modules based on multi-parameter fusion. Background Technology

[0002] Photovoltaic power generation, as a crucial component of clean energy, has become a key direction in the construction of my country's new energy power system. In recent years, the construction scale of centralized photovoltaic power plants and large-scale ground-mounted power plants has continued to expand, the number of photovoltaic modules has steadily increased, and the operating environment of these power plants has become increasingly complex. Photovoltaic modules are installed outdoors for extended periods and must operate continuously under various natural environments, including high temperatures, low temperatures, wind and sand, rainfall, smog, and high humidity. Their operating status directly affects the power generation efficiency, operation and maintenance costs, and overall economic benefits of the photovoltaic power plant. Especially in the arid northwest, industrial areas of North China, and coastal areas with high salt spray, the surface of the modules is more susceptible to contamination from dust, sand, bird droppings, pollen, oil, and industrial particles, leading to problems such as attenuation of incident light, partial shading, increased risk of hot spots, and accelerated aging of encapsulation materials. These factors indicate that effective monitoring and timely early warning of contamination on the surface of photovoltaic modules have become an important technical requirement for the refined operation and maintenance of photovoltaic power plants.

[0003] In existing technologies, the management of photovoltaic (PV) module contamination issues typically employs methods such as manual inspection, periodic cleaning, single image recognition, or single electrical performance parameter analysis. Manual inspection relies primarily on maintenance personnel observing the surface contamination of the modules on-site and then combining this with experience to plan cleaning schedules. Some solutions identify contaminated areas by collecting images of the module surface; others determine the presence of contamination based on single indicators such as transmittance, output power, and temperature changes. While these methods have some application value in small-scale plants, for centralized, large-scale, and widely distributed PV power plants, the increasing number of modules and the greater diversity of operating environments make traditional monitoring methods insufficient to meet the requirements of real-time, precise, and intelligent operation and maintenance. This proposed solution targets centralized PV power plants in the new energy sector, aiming to provide early warnings of module contamination and assist maintenance personnel in making timely maintenance and cleaning decisions, thereby improving module power generation efficiency.

[0004] However, existing technologies still have significant shortcomings. First, manual inspection is highly subjective, inefficient, and difficult to achieve full coverage and continuous monitoring of large-scale photovoltaic power plants. Furthermore, it cannot accurately quantify the degree of contamination, leading to missed detections, misjudgments, and high labor costs. Second, existing image-based detection schemes rely heavily on complex deep learning models. While capable of identifying contaminated areas, they generally suffer from high computational complexity, high hardware deployment costs, and insufficient real-time performance. Some schemes only address the basic requirement of "identifying the presence of contamination," failing to assess the actual impact of contamination on power generation efficiency, potentially leading to over-cleaning or delayed cleaning, resulting in wasted maintenance resources. Third, using single parameters such as transmittance, power, or temperature for early warning is susceptible to interference from external factors such as light intensity, ambient temperature, weather changes, and rainfall, resulting in insufficient accuracy and hindering early prediction and tiered management of contamination levels. In addition, existing early warning systems generally lack the ability to adapt to different regional environmental characteristics. The early warning thresholds are often set in a fixed manner and cannot be dynamically adjusted according to environmental changes such as wind and sand, high temperature and humidity, and rainfall. They have poor adaptability and are difficult to meet the actual needs of photovoltaic module operation and maintenance management in different regions and under different operating conditions.

[0005] To address this, a method for classifying and warning about the contamination of photovoltaic modules based on multi-parameter fusion is proposed. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method for classifying and early warning of contamination in photovoltaic modules based on multi-parameter fusion.

[0007] This invention provides a method for classifying and early warning of contamination in photovoltaic modules based on multi-parameter fusion, comprising the following steps: S1: Collect environmental parameter data characterizing the conditions for surface contamination of photovoltaic modules within the target photovoltaic power station area, historical contamination data of photovoltaic modules formed during the historical operation of the target photovoltaic power station, and power generation data of photovoltaic modules in the target photovoltaic power station. Preprocess the collected environmental parameter data, historical contamination data of photovoltaic modules, and power generation data to construct a regional contamination feature database. Establish a correlation model between the contamination features of photovoltaic modules within the target photovoltaic power station area and the power generation efficiency loss of photovoltaic modules in the target photovoltaic power station. Determine basic early warning thresholds for different types of contamination. S2: Obtain surface temperature distribution data and surface image data of photovoltaic modules in the target photovoltaic power station, and preprocess the surface temperature distribution data, surface image data of photovoltaic modules, and environmental parameter data; S3: Perform dirt identification on the preprocessed photovoltaic module surface image data to identify dirty areas on the surface of photovoltaic modules in the target photovoltaic power station, and construct a quantitative index of the degree of dirtiness of photovoltaic modules based on the dirty areas and the surface temperature distribution data, so as to classify the degree of dirtiness of photovoltaic modules in the target photovoltaic power station according to the quantitative index of the degree of dirtiness of photovoltaic modules. S4: Based on the correlation model and combined with the environmental parameter data, the basic early warning thresholds for different types of dirt are dynamically adjusted to obtain dynamic early warning thresholds that conform to the environmental conditions of the area where the target photovoltaic power station is located and the dirt deposition characteristics of the photovoltaic modules. S5: Compare the quantitative result of the degree of dirtiness of the photovoltaic module with the dynamic early warning threshold, trigger the corresponding level of dirtiness warning, and output cleaning suggestions or cleaning tasks according to the warning level.

[0008] Furthermore, in step S1, the environmental parameter data includes at least wind speed, wind direction, rainfall, light intensity, and temperature and humidity; the historical dirt data of the photovoltaic module includes at least the type of dirt on the surface of the photovoltaic module and the percentage of dirt coverage area; and the power generation data includes at least the power generation capacity of the photovoltaic module.

[0009] Specifically, in step S2, infrared thermal imaging modules are deployed on the photovoltaic module array of the target photovoltaic power station at a preset spacing of 5 to 8 meters to collect surface temperature distribution data of the photovoltaic modules; environmental parameter data are collected using environmental sensors; surface image data of the photovoltaic modules are collected using drone inspection; and the monitoring terminal receives the surface temperature distribution data and environmental parameter data according to a preset sampling period, and acquires the surface image data of the photovoltaic modules.

[0010] Specifically, in step S2, the preset sampling period is 10 minutes to 30 minutes.

[0011] Preferably, in step S2, the acquired component surface image data, surface temperature distribution data, and environmental parameter data are preprocessed, including: performing grayscale and noise reduction processing on the component surface image data, and removing interference caused by environmental reflection and shadows; performing normalization processing on the surface temperature distribution data, and extracting temperature anomalies on the surface of the photovoltaic modules in the target photovoltaic power station; and filtering the environmental parameter data, retaining key environmental parameters that characterize the conditions for dirt formation on the surface of the photovoltaic modules in the target photovoltaic power station.

[0012] Specifically, in step S3, the quantification index of the degree of dirtiness is characterized by the following quantification value of dirtiness: Dirt quantification value = α × percentage of dirty coverage area + β × (1 - light transmittance) + γ × number of temperature anomalies; Wherein, α, β, and γ are weighting coefficients, and the value of α ranges from 0.4 to 0.5, the value of β ranges from 0.3 to 0.4, and the value of γ ranges from 0.1 to 0.2; the weighting coefficients are adjusted according to the type of pollution in the area where the target photovoltaic power station is located.

[0013] Further, in step S3, the degree of contamination of the photovoltaic module is classified according to the contamination quantification value, wherein: when the contamination quantification value is 0.1 to 0.3, the surface of the photovoltaic module is lightly contaminated; when the contamination quantification value is 0.3 to 0.6, the surface of the photovoltaic module is moderately contaminated; and when the contamination quantification value is greater than 0.6, the surface of the photovoltaic module is heavily contaminated.

[0014] Furthermore, in step S4, the dynamic warning threshold adjustment includes: lowering the light pollution warning threshold of the photovoltaic module during sandstorm weather; raising the light pollution warning threshold during rainy weather; and adjusting the moderate and heavy pollution warning thresholds during hot and humid weather.

[0015] Furthermore, in step S5, the output of cleaning suggestions or cleaning tasks corresponding to the warning level includes: When the photovoltaic modules in the target photovoltaic power station are determined to be slightly dirty, a warning is issued and suggestions for natural cleaning based on subsequent rainfall are provided. When the photovoltaic modules in the target photovoltaic power station are determined to be moderately dirty, a warning alert is issued, and a cleaning plan and cleaning priority are output; and When the photovoltaic modules in the target photovoltaic power station are determined to be heavily soiled, an emergency warning is issued, a cleaning task is output, and the cleaning equipment scheduling system is simultaneously linked to remind the operation and maintenance personnel to handle the situation.

[0016] Specifically, after cleaning the photovoltaic modules in the target photovoltaic power station, surface image data, surface temperature distribution data, power generation data, and environmental parameter data for the same period after cleaning are collected to verify the dirt removal effect of the photovoltaic modules and calculate the power generation efficiency recovery rate after cleaning. The surface image data, surface temperature distribution data, power generation data, environmental parameter data, early warning results, and cleaning effect data of the photovoltaic modules before and after cleaning are entered into the regional dirt feature database to optimize the association model and the dirt identification algorithm used to perform dirt identification in step S3.

[0017] The beneficial effects of this invention are as follows: This invention overcomes the shortcomings of existing photovoltaic (PV) module contamination warning technologies, such as low accuracy, poor real-time performance, weak adaptability, and high operation and maintenance costs. It provides a multi-parameter fusion-based method for graded PV module contamination warning. This method achieves accurate identification and quantitative analysis of PV module contamination status by fusing PV module surface image data, surface temperature distribution data, environmental parameter data, and power generation data. By constructing a regional contamination feature database and dynamically adjusting the warning threshold based on real-time environmental parameters, it achieves adaptive matching of the warning threshold to different regional environmental conditions and contamination deposition characteristics. By classifying PV module contamination levels into light, moderate, and heavy contamination and outputting corresponding cleaning suggestions or tasks, it achieves graded warning and differentiated operation and maintenance guidance. Further improvements in warning accuracy and system adaptability are achieved by verifying the cleaning effect after cleaning and continuously optimizing the correlation model and contamination identification algorithm. Therefore, this invention can improve the accuracy and real-time performance of PV module contamination warning, reduce operation and maintenance costs, minimize power generation losses caused by contamination, and extend the service life of PV modules. Attached Figure Description

[0018] Figure 1 The flowchart illustrates the steps of a photovoltaic module dirt classification and early warning method based on multi-parameter fusion, according to a specific embodiment of the present invention. Figure 2 The flowchart illustrates the operation of a photovoltaic module dirt classification and early warning method based on multi-parameter fusion, according to a specific embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown in the figure, a photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion provided by a specific embodiment of the present invention includes the following steps: S1: Collect environmental parameter data representing the conditions for the formation of surface contamination of photovoltaic modules in the target photovoltaic power station, historical contamination data of photovoltaic modules formed during the historical operation of the target photovoltaic power station, and power generation data of photovoltaic modules in the target photovoltaic power station within the area where the target photovoltaic power station is located. Preprocess the collected environmental parameter data, historical contamination data of photovoltaic modules, and power generation data and remove outliers. Construct a regional contamination feature database and establish a correlation model between the contamination features of photovoltaic modules in the area where the target photovoltaic power station is located and the power generation efficiency loss of photovoltaic modules in the target photovoltaic power station. Determine basic early warning thresholds for different types of contamination. S2: Deploy an infrared thermal imaging module on the photovoltaic module array of the target photovoltaic power station to collect surface temperature distribution data of the photovoltaic modules; collect environmental parameter data using environmental sensors; collect surface image data of the photovoltaic modules using drone inspection; receive surface temperature distribution data and environmental parameter data according to a preset sampling period by the monitoring terminal, and acquire surface image data of the photovoltaic modules, and preprocess the surface temperature distribution data, environmental parameter data and surface image data of the photovoltaic modules; S3: Dirt identification is performed on the pre-processed photovoltaic module surface image data to identify dirty areas on the surface of photovoltaic modules in the target photovoltaic power station. After identifying the dirty areas, the percentage of dirty coverage area and the average gray value of the dirt are calculated. Combined with the number of temperature anomalies on the surface of the photovoltaic modules in the target photovoltaic power station, a quantitative index of the degree of dirtiness of the photovoltaic modules is constructed. Based on this index, the degree of dirtiness of the photovoltaic modules in the target photovoltaic power station is classified into light, moderate, and heavy dirtiness. The dirt identification algorithm is used to identify dirty areas on the surface of photovoltaic modules from the pre-processed photovoltaic module surface image data. The dirt identification algorithm can be implemented using any one or a combination of gray-scale threshold segmentation, edge detection, region segmentation, texture feature extraction, or image recognition models obtained through training. First, brightness analysis and region segmentation are performed on the photovoltaic module surface image after grayscale conversion and denoising to identify candidate regions with significantly different gray-scale characteristics from clean module surfaces. Then, the candidate regions are screened based on their area, edge continuity, gray-scale distribution, and texture features to determine the dirty areas. For dirt with strong adhesion or irregular shape, the dirt area can be further corrected by combining historical sample features. The output of the dirt recognition algorithm includes at least one of the following: the location of the dirt area, the percentage of the dirt-covered area, the average gray value of the dirt, and the dirt type determination result. After cleaning, the dirt recognition algorithm can be updated or the model optimized based on the differences between the images before and after cleaning, the warning results, and the cleaning effect data to improve the accuracy of subsequent recognition. S4: Based on the correlation model established in step S1 and combined with the environmental parameter data collected in step S2, the basic warning thresholds corresponding to different types of contamination are dynamically adjusted to obtain dynamic warning thresholds that are compatible with the environmental conditions of the target photovoltaic power station area and the contamination deposition characteristics of photovoltaic modules. The basic warning thresholds are pre-set warning judgment benchmarks for different types of contamination. The basic warning thresholds can be determined based on the historical sample statistical results, historical power generation efficiency loss data, and preset calibration results in the regional contamination feature database. For different types of contamination, the degree of impact on the power generation efficiency of photovoltaic modules under different contamination coverage area ratios, different light transmittances, and different numbers of temperature anomalies can be statistically analyzed, and the basic warning thresholds for light, moderate, and heavy contamination can be determined accordingly. The dynamic warning thresholds are obtained by correcting the basic warning thresholds based on the current environmental parameter data through the correlation model. During dusty weather, the threshold for mild contamination can be lowered due to the faster rate of dirt deposition; during rainy weather, the threshold can be raised due to the possibility of natural cleaning; and during hot and humid weather, the thresholds for moderate and severe contamination can be adjusted because the characteristics of dirt adhesion and the risk of thermal anomalies may change. By employing these methods, the warning results can be made more consistent with the actual operating conditions of the target photovoltaic power station's location. S5: Compare the quantification result of the photovoltaic module dirt level obtained in step S3 with the dynamic early warning threshold obtained in step S4, trigger the corresponding level of dirt warning, and output the cleaning suggestion or cleaning task corresponding to the warning level; After cleaning the photovoltaic modules in the target photovoltaic power station, collect the module surface image data, surface temperature distribution data, power generation data, and environmental parameter data of the same period after cleaning to verify the dirt removal effect of the photovoltaic modules and calculate the power generation efficiency recovery rate of the photovoltaic modules after cleaning; Combine the module surface image data, surface temperature distribution data, power generation data, environmental parameter data, early warning results, and cleaning effect data before and after cleaning. The data is entered into a regional contamination feature database to optimize the correlation model. The effectiveness of contamination removal can be jointly determined based on the surface image data of photovoltaic modules before and after cleaning, surface temperature distribution data, and power generation data. When the proportion of contamination coverage area in the surface image of the photovoltaic modules decreases, the number of temperature anomalies decreases, and the power generation increases after cleaning, the cleaning measures can be determined to be effective. When the above indicators do not improve significantly after cleaning, it can be determined that there is still residual contamination or other abnormal factors affecting power generation performance. The cleaning effect determination results can be used as feedback samples in the regional contamination feature database to update the parameters of the correlation model and the contamination identification algorithm, thereby forming a closed loop of "early warning - cleaning - verification - optimization".

[0021] In one implementation, the correlation model is used to characterize the mapping relationship between the contamination characteristics of photovoltaic modules and the power generation efficiency loss of photovoltaic modules within the area where the target photovoltaic power station is located. The input parameters of the correlation model include at least environmental parameter data, historical contamination data of photovoltaic modules, and power generation data. The environmental parameter data includes at least wind speed, wind direction, rainfall, light intensity, and temperature and humidity. The historical contamination data of photovoltaic modules includes at least the type of contamination and the percentage of contamination-covered area. The power generation data includes at least the power output of the photovoltaic modules. The output of the correlation model includes at least the basic warning thresholds corresponding to different contamination types and the correction amounts for the basic warning thresholds under different environmental conditions. The correlation model can be constructed using rule-based models, regression models, lookup table models, or prediction models trained on historical samples. In one example, based on historical environmental parameters, historical pollution characteristics, and historical power generation efficiency loss data in a regional pollution characteristic database, the variation patterns of photovoltaic module power generation efficiency loss under different pollution types can be statistically analyzed, and the correlation model can be established accordingly. Through the correlation model, the impact of environmental change characteristics in the target photovoltaic power station area on power generation efficiency can be correlated with the surface pollution characteristics of photovoltaic modules, providing a basis for determining the basic early warning threshold and adjusting the dynamic early warning threshold. The regional contamination feature database is used to store at least environmental parameter data, historical contamination data of photovoltaic modules, power generation data, basic early warning thresholds, dynamic early warning threshold adjustment results, early warning results, and cleaning effect data to support the establishment, invocation, and continuous optimization of the correlation model. This invention integrates photovoltaic module surface image data, surface temperature distribution data, environmental parameter data, and power generation data to perform real-time monitoring, quantitative analysis, and graded early warning of the contamination status of photovoltaic module surfaces.

[0022] Specifically, temperature anomalies refer to temperature sampling points or temperature pixels in the photovoltaic module surface temperature distribution data that deviate abnormally from the average surface temperature, neighborhood temperature, or preset reference temperature of the photovoltaic module. When the temperature value of a certain temperature sampling point or temperature pixel is higher than a preset difference value of the average surface temperature of the photovoltaic module, the temperature sampling point or temperature pixel can be identified as a temperature anomaly. The preset difference value can be set according to the environmental conditions of the target photovoltaic power station area, the module model, historical operating status, or calibration results. Abnormal temperature fluctuations caused by instantaneous reflection, short-term cloud shadows, or environmental noise can be eliminated through time continuity filtering or spatial neighborhood filtering. The number of temperature anomalies is used to characterize the degree of local thermal anomalies caused by dirt adhesion on the surface of the photovoltaic module. By incorporating the number of temperature anomalies into the quantitative index of photovoltaic module contamination, the ability to identify local shading, uneven light exposure, and hot spot risks can be improved. Step S1 is used to establish regional basic data and early warning basis; step S2 is used to complete multi-source data acquisition and preprocessing; step S3 is used to realize contamination identification and degree quantification; step S4 is used to dynamically adjust the early warning threshold in combination with regional environmental characteristics; step S5 is used to output graded early warning results and operation and maintenance suggestions; and step S6 is used to verify the effect and optimize the model after cleaning.

[0023] Furthermore, multi-source data fusion can not only identify dirty areas on the surface of photovoltaic modules, but also determine the degree of impact of dirt on the power generation efficiency of photovoltaic modules, thereby improving the consistency between early warning results and actual operation and maintenance needs.

[0024] Furthermore, by introducing a regional dirt and grime feature database and a dynamic early warning threshold adjustment mechanism, the adaptability of this invention under different regions, different weather conditions, and different types of dirt and grime can be enhanced, reducing the probability of false alarms and missed alarms.

[0025] Based on the above basic implementation method, in step S1, the environmental parameter data includes at least wind speed, wind direction, rainfall, light intensity, and temperature and humidity; the historical dirt data of the photovoltaic module includes at least the type of dirt on the surface of the photovoltaic module and the percentage of dirt coverage area; and the power generation data includes at least the power generation capacity of the photovoltaic module.

[0026] Furthermore, the aforementioned environmental parameter data, historical contamination data of photovoltaic modules, and power generation data can be organized, correlated, and stored according to a unified time dimension. This data can be used to characterize the relationship between environmental changes in the area where the target photovoltaic power station is located, the surface contamination characteristics of photovoltaic modules, and the power generation efficiency loss of photovoltaic modules. It can also provide data support for determining the basic early warning threshold and adjusting the dynamic early warning threshold.

[0027] In one specific embodiment, the infrared thermal imaging module is deployed on the photovoltaic module array of the target photovoltaic power station at a preset interval of 5 to 8 meters; in step S2, the preset sampling period is 10 to 30 minutes.

[0028] In this embodiment, the infrared thermal imaging module, environmental sensor, drone, and monitoring terminal together constitute a multi-source data acquisition system. The infrared thermal imaging module is used to acquire abnormal temperature characteristics on the surface of the photovoltaic module, the environmental sensor is used to acquire environmental change information, the drone is used to acquire image information of the photovoltaic module surface, and the monitoring terminal is used to receive, summarize, and preprocess various types of data.

[0029] Furthermore, setting the deployment spacing of the infrared thermal imaging modules to 5 to 8 meters and the preset sampling period to 10 to 30 minutes helps to balance data acquisition coverage, monitoring real-time performance, and system deployment costs.

[0030] In another specific embodiment, in step S2, the collected component surface image data, surface temperature distribution data, and environmental parameter data are preprocessed, including: performing grayscale and noise reduction processing on the collected component surface image data, and removing interference caused by environmental reflection and shadow; performing normalization processing on the surface temperature distribution data, and extracting temperature anomalies on the surface of the photovoltaic modules in the target photovoltaic power station; and screening the environmental parameter data, retaining key environmental parameters that characterize the conditions for dirt formation on the surface of the photovoltaic modules in the target photovoltaic power station.

[0031] In this embodiment, the purpose of preprocessing multi-source data is to reduce the impact of environmental noise, abnormal data, and invalid data on the results of dirt identification and early warning.

[0032] Specifically, after performing grayscale conversion and noise reduction on the component surface image data, the impact of environmental reflection, shadows, and image noise on the identification results of dirty areas can be reduced; after normalizing the surface temperature distribution data, the comparability of temperature anomaly characteristics of different photovoltaic modules under different operating conditions can be improved.

[0033] Furthermore, by filtering and retaining key environmental parameters from the environmental parameter data, subsequent adjustments to the dynamic early warning threshold can better align with the actual environmental characteristics of the target photovoltaic power station's location.

[0034] In another specific embodiment, in step S3, the degree of dirtiness is characterized by the following dirtiness quantification value: Dirt quantification value = α × percentage of dirty coverage area + β × (1 - light transmittance) + γ × number of temperature anomalies; Wherein, α, β, and γ are weighting coefficients, with α ranging from 0.4 to 0.5, β ranging from 0.3 to 0.4, and γ ranging from 0.1 to 0.2; the weighting coefficients are adjusted according to the type of contamination in the area where the target photovoltaic power station is located; the transmittance is obtained from the results of the analysis of the module surface image, the results of on-site testing, or the preset calibration data; the transmittance is used to characterize the degree of influence of the contamination on the surface of the photovoltaic module on the transmission capacity of incident light. The transmittance can be represented by a normalized value, ranging from 0 to 1. A transmittance closer to 1 indicates a smaller impact of surface contamination on incident light, while a lower transmittance indicates a more significant blocking effect of surface contamination on incident light. Transmittance can be obtained from image analysis of the module surface, by comparing the brightness, grayscale, or reflection characteristics of the current photovoltaic module surface image with the characteristics of a baseline image under a preset clean state to obtain the normalized transmittance. Alternatively, the transmittance can be directly obtained from on-site transmittance detection results. It can also be calculated based on historical calibration data and the current percentage of contamination coverage. To reduce the impact of different weather conditions, light intensities, or shooting angles on the transmittance calculation results, illumination compensation, brightness normalization, or baseline calibration can be performed on the image before transmittance calculation to improve the stability of the transmittance representation results.

[0035] Furthermore, the percentage of dirty coverage area in the dirt quantification value is used to characterize the area characteristics of the dirty area on the surface of the photovoltaic module, the light transmittance is used to characterize the degree of impact of dirt on the light-receiving ability of the photovoltaic module, and the number of temperature anomalies is used to characterize the thermal anomalies caused by dirt.

[0036] Specifically, by weighting and fusing the above three parameters, a comprehensive quantitative index that takes into account image features, light-receiving features, and thermal features can be formed, thereby avoiding the bias caused by a single parameter and improving the accuracy of the quantitative results of the degree of dirtiness of photovoltaic modules.

[0037] In another specific embodiment, in step S3, the degree of contamination of the photovoltaic module is classified according to the contamination quantification value, wherein: when the contamination quantification value is 0.1 to 0.3, the surface of the photovoltaic module is slightly contaminated; when the contamination quantification value is 0.3 to 0.6, the surface of the photovoltaic module is moderately contaminated; and when the contamination quantification value is greater than 0.6, the surface of the photovoltaic module is heavily contaminated.

[0038] Furthermore, based on the quantification value of dirtiness, the degree of dirtiness on the surface of photovoltaic modules is divided into three levels: light dirtiness, moderate dirtiness, and heavy dirtiness, which is conducive to adopting differentiated operation and maintenance strategies for different degrees of dirtiness in the future.

[0039] Specifically, when the dirt level is in a low range, it indicates that the dirt on the surface of the photovoltaic module has little impact on the power generation efficiency; when the dirt level is in a high range, it indicates that the dirt on the surface of the photovoltaic module has already had a significant adverse impact on the power generation performance of the photovoltaic module, and cleaning measures need to be taken in a timely manner.

[0040] In another specific embodiment, in step S4, the dynamic warning threshold adjustment includes: lowering the warning threshold for mild dirt on photovoltaic modules during sandstorm weather; raising the warning threshold for mild dirt during rainy weather; and adjusting the warning thresholds for moderate and severe dirt during hot and humid weather.

[0041] Furthermore, the dynamic early warning threshold adjustment is not based on a fixed threshold, but rather on targeted adjustments made in conjunction with changes in environmental conditions and regional pollution characteristics in the area where the target photovoltaic power station is located.

[0042] Furthermore, by adjusting the warning thresholds under different weather conditions such as sandstorms, rainfall, and high temperature and humidity, the warning results can better reflect the actual changes in dirt on the surface of photovoltaic modules, thereby improving the regional adaptability and operational condition adaptability of the warnings.

[0043] In one specific implementation, in step S5, the output of cleaning suggestions or cleaning tasks corresponding to the warning level includes: when the photovoltaic modules in the target photovoltaic power station are determined to be slightly dirty, issuing a prompt warning and outputting a suggestion for natural cleaning based on subsequent rainfall; when the photovoltaic modules in the target photovoltaic power station are determined to be moderately dirty, issuing a warning and outputting a cleaning plan and cleaning priority; and when the photovoltaic modules in the target photovoltaic power station are determined to be heavily dirty, issuing an emergency warning, outputting a cleaning task, and simultaneously linking to the cleaning equipment scheduling system to remind maintenance personnel to handle the situation promptly.

[0044] Specifically, when the photovoltaic modules in the target photovoltaic power station are determined to be slightly dirty, it means that the dirt has a relatively small impact on the power generation efficiency of the photovoltaic modules, and natural cleaning can be prioritized in conjunction with subsequent rainfall. When they are determined to be moderately dirty, it means that the dirt has a significant impact on the power generation efficiency of the photovoltaic modules, and manual or equipment cleaning should be arranged according to the cleaning plan and cleaning priority. When they are determined to be heavily dirty, it means that the dirt has significantly affected the operating status of the photovoltaic modules, and cleaning tasks should be carried out in a timely manner and the cleaning equipment scheduling system should be activated for processing.

[0045] Furthermore, the aforementioned tiered operation and maintenance guidance method enables cleaning resources to be prioritized for photovoltaic modules that are more affected by dirt, thereby reducing ineffective and delayed cleaning, improving operation and maintenance efficiency, and reducing operation and maintenance costs.

[0046] In one specific implementation, in step S5, based on the quantification result of the degree of dirtiness of the photovoltaic module obtained in step S3 and the dynamic warning threshold obtained in step S4, different levels of dirtiness warnings are triggered, and cleaning suggestions or cleaning tasks are output according to the triggered warning level. The dirtiness warning information, the quantification result of the degree of dirtiness of the photovoltaic module and the cleaning suggestions or cleaning tasks are pushed to the operation and maintenance terminal through the cloud platform. In step S6, by comparing the power generation data of the photovoltaic module before and after cleaning, the power generation efficiency recovery rate of the photovoltaic module after cleaning is calculated. Based on the module surface image data, surface temperature distribution data, power generation data, environmental parameter data, early warning results and cleaning effect data before and after cleaning, the association model and the dirt identification algorithm used to perform dirt identification in step S3 are optimized.

[0047] In this embodiment, by pushing dirt warning information, quantitative results of the degree of dirtiness of photovoltaic modules, and cleaning suggestions or cleaning tasks to the operation and maintenance terminal, operation and maintenance personnel can promptly grasp the dirt status of photovoltaic modules in the target photovoltaic power station and the need for disposal.

[0048] Specifically, after cleaning is completed, by comparing the power generation data of the photovoltaic modules before and after cleaning, and combining the photovoltaic module surface image data, surface temperature distribution data and environmental parameter data, the actual effect of the cleaning measures can be verified. The verification results are written back to the regional dirt feature database for continuous optimization of the correlation model and the dirt identification algorithm used to perform dirt identification in step S3, thereby improving the accuracy and adaptability of subsequent early warnings.

[0049] In one specific embodiment, the present invention is applied to a centralized photovoltaic power station for new energy in areas prone to sandstorms. When sandstorms occur in the area where the target photovoltaic power station is located, environmental parameter data collected by environmental sensors indicate increased wind speed, continuous changes in wind direction, and low or no effective rainfall. Simultaneously, the monitoring terminal receives surface temperature distribution data and environmental parameter data of the photovoltaic modules according to a preset sampling period, and acquires surface image data of the photovoltaic modules collected by a drone inspection. The aforementioned data is then preprocessed. Since wind speed, wind direction, rainfall, light intensity, and temperature and humidity are used as environmental parameters, and an infrared thermal imaging module is deployed on the photovoltaic module array, drones collect module inspection images, and the monitoring terminal collects or receives data according to a preset sampling period, this embodiment can utilize the aforementioned multi-source data to continuously monitor the surface condition of the modules during sandstorms. The preprocessed photovoltaic module surface image data is then used for dirt identification. When a large area of ​​dust or sand adheres to the module surface is identified, the percentage of dirt coverage and the average gray value of the dirt are calculated. Combined with the number of temperature anomalies extracted by the infrared thermal imaging module, a quantitative index of the degree of dirt on the photovoltaic modules is constructed. The degree of contamination is quantified by weighting the percentage of contamination coverage area, light transmittance, and the number of temperature anomalies. The degree of contamination is divided into three levels: light contamination, moderate contamination, and heavy contamination. Therefore, in dusty weather, the degree of impact of dust deposition on the light-receiving capacity and heat distribution of photovoltaic module surfaces can be determined in real time through the degree of contamination quantification.

[0050] In this embodiment, under sandstorm conditions, based on the regional pollution characteristic database constructed in step S1 and combined with the real-time environmental parameter data collected in step S2, the basic early warning threshold is dynamically adjusted through an association model. In this embodiment, the system prioritizes lowering the early warning threshold for mild pollution, enabling photovoltaic modules to enter the early warning state earlier in the initial stage of sandstorm deposition, thereby improving the foresight of the early warning and avoiding a significant decrease in module power generation efficiency in a short period of time due to rapid sandstorm accumulation.

[0051] Specifically, when the quantified result of the contamination level of the photovoltaic module reaches the mild contamination warning threshold after correction for sandstorm weather, the system triggers a prompt warning and pushes the contamination warning information, the quantified result of the photovoltaic module's contamination level, and warning handling suggestions to the operation and maintenance terminal through the cloud platform. When the contamination level further increases and reaches the moderate contamination level, the system issues a warning, pushes a cleaning plan, and clarifies the cleaning priority. When the contamination level continues to rise and reaches the severe contamination level, the system issues an emergency warning, outputs a cleaning task, and simultaneously links to the cleaning equipment scheduling system to remind operation and maintenance personnel to handle the situation promptly. After cleaning is completed, the monitoring terminal collects surface image data, surface temperature distribution data, power generation data, and environmental parameter data for the same period after cleaning of the photovoltaic module. By comparing the power generation data before and after cleaning, the system calculates the power generation efficiency recovery rate of the photovoltaic module after cleaning and records the data before and after cleaning, the warning results, and the cleaning effect data into the regional contamination feature database to continuously optimize the correlation model and contamination identification algorithm. Through the above methods, the system's accuracy in judging the rate of contamination deposition, changes in contamination type, and the timing of warnings under sandstorm weather conditions can be gradually improved.

[0052] In one specific embodiment, the present invention is applied to a centralized photovoltaic power station in a new energy region experiencing significant rainfall. When rainfall occurs in the area where the target photovoltaic power station is located, environmental parameter data collected by environmental sensors indicate increased rainfall, higher air humidity, and fluctuations in light intensity. The monitoring terminal receives surface temperature distribution data and environmental parameter data of the photovoltaic modules according to a preset sampling period, and acquires surface image data of the photovoltaic modules collected by drone inspections. The above data is preprocessed to identify the dirt status of the photovoltaic module surface under rainfall conditions. Since rainfall, temperature, humidity, and light intensity have been included in the environmental parameter range, and dirt identification and graded early warning through multi-parameter fusion have been publicly disclosed, in this embodiment, real-time environmental information under rainfall conditions can be used to specifically modify the early warning logic. Dirt identification is performed on the preprocessed photovoltaic module surface image data, and temperature anomalies are extracted by combining surface temperature distribution data to calculate the proportion of dirt coverage area, the average gray value of dirt, and the quantitative value of dirt degree. When the system determines that the surface dirt of the current component is mainly manifested as light dust adhesion, and the regional meteorological conditions indicate that there will be continuous rainfall or natural scouring conditions in the short term, the basic warning threshold is dynamically adjusted based on the regional dirt feature database and real-time environmental parameter data.

[0053] In this embodiment, when the quantification result of the degree of contamination of the photovoltaic module is in the original slightly contaminated range, but has not yet reached the slightly contamination warning threshold after rainfall, the system does not immediately trigger a forced cleaning task. Instead, it outputs a suggestion for natural cleaning based on subsequent rainfall, and pushes this suggestion, along with the contamination warning information and the contamination quantification result, to the operation and maintenance terminal through the cloud platform. If the data collected again after the rainfall shows that the contamination on the module surface has been significantly reduced, and the power generation efficiency recovery rate after cleaning reaches the expected level, then the self-cleaning effect of this rainfall is written as a sample into the regional contamination feature database for subsequent optimization of the correlation model.

[0054] Specifically, during rainy weather, if re-identification reveals that the surface of the photovoltaic modules still retains dirt that is difficult to remove by natural rainfall, such as highly adhesive mud spots, bird droppings, or mixed pollutants, causing the quantification result of the photovoltaic modules' dirt level to reach the moderate or heavy dirt level, the system will issue a warning or emergency warning according to the graded early warning logic, and output a cleaning plan and cleaning priority, or directly output a cleaning task and link it with the cleaning equipment scheduling system to remind maintenance personnel to handle it in a timely manner. This can make full use of the natural cleaning effect of rainy weather and avoid the problem of untimely treatment of stubborn dirt. In the rainy weather embodiment, the monitoring terminal also collects photovoltaic module surface image data, surface temperature distribution data, power generation data, and environmental parameter data for the same period again after cleaning is completed or after the rain ends. By comparing the power generation data of the photovoltaic modules before and after cleaning, the power generation efficiency recovery rate of the photovoltaic modules after cleaning is calculated, and the data before and after cleaning, the warning results, and the cleaning effect data are entered into the regional dirt feature database to continuously optimize the correlation model and dirt identification algorithm. Through the above methods, the system can gradually form a warning threshold correction rule and a natural cleaning effect evaluation mechanism applicable to rainy weather conditions.

[0055] To aid in a better understanding of the present invention, a more comprehensive and specific embodiment is described. In this embodiment, the present invention provides a photovoltaic module contamination classification and early warning method based on multi-parameter fusion, comprising the following steps: S1: Collect environmental parameter data representing the conditions for the formation of surface contamination of photovoltaic modules in the target photovoltaic power station, historical contamination data of photovoltaic modules formed during the historical operation of the target photovoltaic power station, and power generation data of photovoltaic modules in the target photovoltaic power station within the area where the target photovoltaic power station is located. Preprocess the collected environmental parameter data, historical contamination data of photovoltaic modules, and power generation data and remove outliers. Construct a regional contamination feature database and establish a correlation model between the contamination features of photovoltaic modules in the area where the target photovoltaic power station is located and the power generation efficiency loss of photovoltaic modules in the target photovoltaic power station. Determine basic early warning thresholds for different types of contamination. S2: Deploy an infrared thermal imaging module on the photovoltaic module array of the target photovoltaic power station to collect surface temperature distribution data of the photovoltaic modules; collect environmental parameter data using environmental sensors; collect surface image data of the photovoltaic modules using drone inspection; receive surface temperature distribution data and environmental parameter data according to a preset sampling period by the monitoring terminal, and acquire surface image data of the photovoltaic modules, and preprocess the surface temperature distribution data, environmental parameter data and surface image data of the photovoltaic modules; S3: Perform dirt identification on the pre-processed photovoltaic module surface image data to identify dirty areas on the surface of photovoltaic modules in the target photovoltaic power station; after identifying the dirty areas, calculate the proportion of dirty coverage area and the average gray value of dirt corresponding to the dirty areas, and combine the number of temperature anomalies on the surface of photovoltaic modules in the target photovoltaic power station to construct a quantitative index of the degree of dirtiness of photovoltaic modules, so as to classify the degree of dirtiness of photovoltaic modules in the target photovoltaic power station into light dirtiness, moderate dirtiness and heavy dirtiness according to the quantitative index of the degree of dirtiness of photovoltaic modules; S4: Based on the association model established in step S1 and combined with the environmental parameter data collected in step S2, the basic early warning thresholds corresponding to different types of dirt are dynamically adjusted to obtain dynamic early warning thresholds that are compatible with the environmental conditions of the target photovoltaic power station area and the dirt deposition characteristics of photovoltaic modules. S5: Compare the quantitative result of the degree of dirtiness of the photovoltaic module obtained in step S3 with the dynamic early warning threshold obtained in step S4, trigger the corresponding level of dirtiness warning, and output the cleaning suggestion or cleaning task corresponding to the warning level. S6: After cleaning the photovoltaic modules in the target photovoltaic power station, collect surface image data, surface temperature distribution data, power generation data, and environmental parameter data for the same period after cleaning of the photovoltaic modules to verify the dirt removal effect of the photovoltaic modules and calculate the power generation efficiency recovery rate of the photovoltaic modules after cleaning; enter the surface image data, surface temperature distribution data, power generation data, environmental parameter data, early warning results, and cleaning effect data of the photovoltaic modules before and after cleaning into the regional dirt feature database to optimize the correlation model.

[0056] In this embodiment, in step S1, the environmental parameter data includes at least wind speed, wind direction, rainfall, light intensity, and temperature and humidity; the historical dirt data of the photovoltaic modules includes at least the type of dirt on the surface of the photovoltaic modules and the percentage of dirt-covered area; the power generation data includes at least the power generation of the photovoltaic modules; the infrared thermal imaging module is deployed on the photovoltaic module array of the target photovoltaic power station at a preset spacing of 5 to 8 meters; in step S2, the preset sampling period is 10 to 30 minutes; in step S2, the collected module surface image data, surface temperature distribution data, and environmental parameter data are preprocessed, including: grayscale conversion and noise reduction of the collected module surface image data, and removal of interference caused by environmental reflection and shadows; normalization of the surface temperature distribution data, and extraction of temperature anomalies on the surface of the photovoltaic modules in the target photovoltaic power station; and screening of the environmental parameter data, retaining those that characterize the surface of the photovoltaic modules in the target photovoltaic power station. Key environmental parameters for surface contamination formation; in step S3, the degree of contamination is quantified using the following contamination quantification value: Contamination quantification value = α × percentage of contamination-covered area + β × (1 - transmittance) + γ × number of temperature anomalies; where α, β, and γ are weighting coefficients, and the value of α ranges from 0.4 to 0.5, the value of β ranges from 0.3 to 0.4, and the value of γ ranges from 0.1 to 0.2; the weighting coefficients are adjusted according to the type of contamination in the area where the target photovoltaic power station is located; the transmittance is obtained from the results of component surface image analysis, on-site detection results, or preset calibration data; in step S3, the degree of contamination of the photovoltaic components is classified according to the contamination quantification value, where: when the contamination quantification value is 0.1 to 0.3, the surface of the photovoltaic component is lightly contaminated; when the contamination quantification value is 0.3 to 0.6, the surface of the photovoltaic component is moderately contaminated; when the contamination quantification value is greater than 0.6, the surface of the photovoltaic component is heavily contaminated.

[0057] Specifically, in step S4, the dynamic warning threshold adjustment includes: lowering the warning threshold for light dirt on photovoltaic modules during sandstorms; raising the warning threshold for light dirt during rainy weather; and adjusting the warning thresholds for moderate and heavy dirt during hot and humid weather. In step S5, the output of cleaning suggestions or tasks corresponding to the warning level includes: issuing a warning when photovoltaic modules in the target photovoltaic power station are determined to be lightly dirty, and outputting suggestions for natural cleaning based on subsequent rainfall; issuing a warning when photovoltaic modules in the target photovoltaic power station are determined to be moderately dirty, and outputting a cleaning plan and cleaning priority; and issuing an emergency warning when photovoltaic modules in the target photovoltaic power station are determined to be heavily dirty, outputting a cleaning task, and simultaneously coordinating with the scheduling of cleaning equipment. The system reminds maintenance personnel to handle issues promptly. In step S5, based on the quantification result of the degree of contamination of the photovoltaic module obtained in step S3 and the dynamic warning threshold obtained in step S4, different levels of contamination warnings are triggered. Cleaning suggestions or cleaning tasks are output according to the triggered warning level, and the contamination warning information, the quantification result of the degree of contamination of the photovoltaic module, and the cleaning suggestions or cleaning tasks are pushed to the maintenance terminal through the cloud platform. In step S6, by comparing the power generation data before and after cleaning the photovoltaic module, the power generation efficiency recovery rate of the photovoltaic module after cleaning is calculated. Based on the module surface image data, surface temperature distribution data, power generation data, environmental parameter data, warning results, and cleaning effect data before and after cleaning, the association model and the contamination recognition algorithm used to perform contamination recognition in step S3 are optimized.

[0058] In summary, the embodiments disclosed herein have at least the following technical effects: This invention integrates photovoltaic module surface image data, surface temperature distribution data, environmental parameter data, and power generation data to construct a regional contamination feature database and establishes a correlation model between photovoltaic module contamination features and power generation efficiency loss. This enables accurate identification and quantitative analysis of the contamination status of photovoltaic module surfaces, improves the accuracy and real-time performance of contamination early warning, and overcomes the problems of traditional manual inspection being subjective, inefficient, and susceptible to interference from external factors and having low early warning accuracy due to single-parameter monitoring. This invention determines the basic warning threshold based on different types of contamination and dynamically adjusts the warning threshold in combination with real-time environmental parameters. In particular, it can adaptively correct the warning threshold for sandstorms, rainy weather, and high temperature and humidity weather, so that the warning result is adapted to the environmental conditions and contamination deposition characteristics of the target photovoltaic power station area. This solves the problems of fixed warning thresholds, poor regional adaptability, and difficulty in meeting the operation and maintenance needs under different environments in the existing technology. This invention constructs a quantitative index of dirt levels to classify the dirt levels of photovoltaic modules into light, moderate, and heavy dirt levels, and outputs natural cleaning suggestions, cleaning plans, cleaning priorities, or emergency cleaning tasks for different levels. This achieves graded early warning and differentiated operation and maintenance guidance for photovoltaic module dirt, which can avoid the waste of operation and maintenance resources caused by over-cleaning or untimely cleaning, improve the operation and maintenance efficiency of photovoltaic power plants, and reduce operation and maintenance costs. This invention also pushes early warning information, quantified dirt data, and cleaning suggestions to the operation and maintenance terminal, and collects relevant data again after cleaning to verify the dirt removal effect, calculate the power generation efficiency recovery rate after cleaning, and enters the data before and after cleaning into the regional dirt feature database. It continuously optimizes the correlation model and dirt identification algorithm to form a closed-loop mechanism of "monitoring-early warning-cleaning-verification-optimization", thereby further improving the system's early warning accuracy, environmental adaptability, and long-term operational stability.

[0059] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for graded early warning of contamination in photovoltaic modules based on multi-parameter fusion, characterized in that, Includes the following steps: S1: Collect environmental parameter data characterizing the conditions for surface contamination of photovoltaic modules within the target photovoltaic power station area, historical contamination data of photovoltaic modules formed during the historical operation of the target photovoltaic power station, and power generation data of photovoltaic modules in the target photovoltaic power station. Preprocess the collected environmental parameter data, historical contamination data of photovoltaic modules, and power generation data to construct a regional contamination feature database. Establish a correlation model between the contamination features of photovoltaic modules within the target photovoltaic power station area and the power generation efficiency loss of photovoltaic modules in the target photovoltaic power station. Determine basic early warning thresholds for different types of contamination. S2: Obtain surface temperature distribution data and surface image data of photovoltaic modules in the target photovoltaic power station, and preprocess the surface temperature distribution data, surface image data of photovoltaic modules, and environmental parameter data; S3: Perform dirt identification on the preprocessed photovoltaic module surface image data to identify the dirty areas on the surface of the photovoltaic modules in the target photovoltaic power station, and construct a quantitative index of the degree of dirtiness of the photovoltaic modules based on the dirty areas and the surface temperature distribution data, so as to classify the degree of dirtiness of the photovoltaic modules in the target photovoltaic power station according to the quantitative index of the degree of dirtiness of the photovoltaic modules. S4: Based on the correlation model and combined with the environmental parameter data, the basic early warning thresholds for different types of dirt are dynamically adjusted to obtain dynamic early warning thresholds that conform to the environmental conditions of the area where the target photovoltaic power station is located and the dirt deposition characteristics of the photovoltaic modules. S5: Compare the obtained quantitative result of the degree of dirtiness of the photovoltaic module with the dynamic early warning threshold, trigger the corresponding level of dirtiness warning, and output cleaning suggestions or cleaning tasks according to the warning level.

2. The photovoltaic module pollution classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S1, the environmental parameter data includes at least wind speed, wind direction, rainfall, light intensity, and temperature and humidity; the historical dirt data of the photovoltaic module includes at least the type of dirt on the surface of the photovoltaic module and the percentage of dirt coverage area; and the power generation data includes at least the power generation capacity of the photovoltaic module.

3. The photovoltaic module pollution classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S2, infrared thermal imaging modules are deployed on the photovoltaic module array of the target photovoltaic power station at a preset spacing of 5 to 8 meters to collect surface temperature distribution data of the photovoltaic modules. Environmental parameter data is collected using environmental sensors; surface image data of photovoltaic modules is collected using drone inspection; and the surface temperature distribution data and environmental parameter data are received by the monitoring terminal according to a preset sampling period, and the surface image data of the photovoltaic modules is obtained.

4. The photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion according to claim 3, characterized in that, In step S2, the preset sampling period is 10 minutes to 30 minutes.

5. The photovoltaic module pollution classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S2, the acquired component surface image data, surface temperature distribution data, and environmental parameter data are preprocessed, including: grayscale conversion and noise reduction of the component surface image data, and removal of interference caused by environmental reflection and shadow; normalization of the surface temperature distribution data, and extraction of temperature anomalies on the surface of the photovoltaic modules in the target photovoltaic power station; and screening of the environmental parameter data, retaining key environmental parameters that characterize the conditions for dirt formation on the surface of the photovoltaic modules in the target photovoltaic power station.

6. The photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S3, the degree of dirtiness is characterized by the following dirtiness quantification values: Dirt quantification value = α × percentage of dirty coverage area + β × (1 - light transmittance) + γ × number of temperature anomalies; Wherein, α, β, and γ are weighting coefficients, and the value of α ranges from 0.4 to 0.5, the value of β ranges from 0.3 to 0.4, and the value of γ ranges from 0.1 to 0.2; the weighting coefficients are adjusted according to the type of pollution in the area where the target photovoltaic power station is located.

7. The photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion according to claim 6, characterized in that, In step S3, the degree of dirtiness of the photovoltaic module is classified according to the dirt quantification value, wherein: when the dirt quantification value is 0.1 to 0.3, the surface of the photovoltaic module is lightly dirty; when the dirt quantification value is 0.3 to 0.6, the surface of the photovoltaic module is moderately dirty; and when the dirt quantification value is greater than 0.6, the surface of the photovoltaic module is heavily dirty.

8. The photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S4, the dynamic warning threshold adjustment includes: lowering the light pollution warning threshold of the photovoltaic module during sandstorm weather; raising the light pollution warning threshold during rainy weather; and adjusting the moderate and heavy pollution warning thresholds during hot and humid weather.

9. The photovoltaic module pollution classification and early warning method based on multi-parameter fusion according to claim 1, characterized in that, In step S5, the output of cleaning recommendations or cleaning tasks corresponding to the warning level includes: When the photovoltaic modules in the target photovoltaic power station are determined to be slightly dirty, a warning is issued and suggestions for natural cleaning based on subsequent rainfall are provided. When the photovoltaic modules in the target photovoltaic power station are determined to be moderately dirty, a warning alert is issued, and a cleaning plan and cleaning priority are output; and When the photovoltaic modules in the target photovoltaic power station are determined to be heavily soiled, an emergency warning is issued, a cleaning task is output, and the cleaning equipment scheduling system is simultaneously linked to remind the operation and maintenance personnel to handle the situation.

10. The photovoltaic module dirt and grime classification and early warning method based on multi-parameter fusion according to any one of claims 1 to 9, characterized in that: After cleaning the photovoltaic modules in the target photovoltaic power station, surface image data, surface temperature distribution data, power generation data, and environmental parameter data for the same period after cleaning are collected to verify the dirt removal effect of the photovoltaic modules and calculate the power generation efficiency recovery rate after cleaning. The surface image data, surface temperature distribution data, power generation data, environmental parameter data, early warning results, and cleaning effect data of the photovoltaic modules before and after cleaning are entered into the regional dirt feature database to optimize the association model and the dirt identification algorithm used to perform dirt identification in step S3.

Images