Distributed photovoltaic power station ash deposition monitoring system and method

Abstract

The present disclosure provides a distributed photovoltaic field dust monitoring and evaluation system and method, which performs pixel-level analysis on photovoltaic panel images through a dust loss monitoring model, adopts a connected domain detection and dynamic division strategy, solves the technical problem that traditional manual inspection cannot accurately locate the dust accumulation area, improves the dust identification accuracy, converts the image gray value into quantified dust thickness based on the dust loss corresponding table, generates a ladder-shaped virtual dust layer combined with three-dimensional modeling technology, enables operation and maintenance personnel to intuitively master the dust distribution characteristics and severity, reduces the misjudgment rate, fuses the dust thickness, component position and environmental data through an intelligent decision module to generate a dynamic cleaning priority and AR visual path planning, improves the cleaning resource allocation efficiency, avoids invalid cleaning operations, uses digital twin technology to realize spatial superposition analysis of historical data and real-time monitoring, can predict the dust development trend in advance, and effectively prevents component damage caused by hot spot effect.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic engineering technology, and in particular to a system and method for monitoring and evaluating dust accumulation in distributed photovoltaic power plants. Background Technology

[0002] In some photovoltaic (PV) projects, the problem of dust accumulation on the PV panel surface has been neglected, resulting in a thick layer of dust buildup. This dust reduces light transmittance, severely impacting PV power generation efficiency. Furthermore, the acidic and alkaline dust adhering to the PV panel can corrode the glass cover, causing irreversible damage and ultimately reducing the conversion efficiency of the PV module. Therefore, monitoring dust loss on PV panels is crucial.

[0003] In the existing technology, when monitoring the dust loss of photovoltaic panels, staff are usually dispatched regularly to check the dust accumulation on the photovoltaic panels and clean them according to the inspection results.

[0004] However, this method makes it impossible for staff to accurately grasp the dust accumulation on the photovoltaic panels before inspection. Furthermore, due to the large number of photovoltaic panels being inspected, the dust accumulation on each panel may be different, resulting in very low efficiency in monitoring the dust loss of photovoltaic panels.

[0005] Therefore, how to customize a display strategy for the dust accumulation of photovoltaic panels and improve the efficiency of dust loss monitoring has become an urgent problem to be solved. Summary of the Invention

[0006] A first aspect of this disclosure provides a system for monitoring and evaluating dust accumulation in distributed photovoltaic power plants, comprising:

[0007] The customized module (110) is configured to receive a gray loss monitoring request from the monitoring terminal and retrieve the virtual photovoltaic space corresponding to the photovoltaic area based on the request. The virtual photovoltaic space contains multiple virtual photovoltaic panels that are bound one-to-one with the actual photovoltaic panels.

[0008] An integration module (120) is connected to an image acquisition device (200) for acquiring an image of an actual photovoltaic panel and identifying the ash accumulation area in the image through a ash loss monitoring model, wherein the ash loss monitoring model filters ash accumulation pixels based on the ash accumulation pixel value range and generates an ash accumulation contour.

[0009] The summary module (130) is configured to execute the ash loss judgment strategy, dynamically divide the ash accumulation area along the preset division direction, generate multiple ash loss judgment areas, and convert the ash loss value of each area into the ash accumulation amount based on the ash loss correspondence table.

[0010] The display module (140) maps the amount of ash accumulation to a virtual photovoltaic panel, constructs a virtual ash layer with thickness parameters, updates the virtual photovoltaic space to generate an ash loss monitoring space, and outputs display data containing a ash distribution heat map and a three-dimensional model to the monitoring end.

[0011] In conjunction with the first aspect, the ash loss monitoring model performs the following operations:

[0012] Component boundaries in photovoltaic panel images are extracted using edge detection algorithms;

[0013] Within the boundary range, pixels with gray values ​​in the range of 120-180 are selected as gray accumulation pixels;

[0014] Connectivity analysis is performed on adjacent gray pixels to generate a gray region containing at least one closed contour.

[0015] In conjunction with the first aspect, the gray loss judgment strategy includes:

[0016] Scan the dust accumulation area line by line along the direction from the top edge to the bottom edge of the photovoltaic panel;

[0017] When the coverage of gray pixels in N consecutive rows is ≥80%, a horizontal dividing line is generated, where N is a preset threshold for the number of rows.

[0018] If the distance between adjacent dividing lines is less than the preset safety distance, the dividing lines will be merged.

[0019] Based on the final dividing line, the ash accumulation area is divided into multiple ash loss judgment sub-regions.

[0020] In conjunction with the first aspect, the aggregation module (130) is further configured as follows:

[0021] Calculate the average gray value of each gray loss judgment sub-region;

[0022] Query the pre-stored gray loss correspondence table, which contains the mapping relationship between multiple gray ranges and gray accumulation thickness, wherein every 10 increase in gray value corresponds to an increase of 0.2 mm in gray accumulation thickness;

[0023] When the difference in dust accumulation thickness between adjacent sub-regions is ≤0.1mm, the sub-regions are automatically merged.

[0024] In conjunction with the first aspect, the display module (140) includes:

[0025] The coordinate alignment unit (141) uses an affine transformation algorithm to match the gray contour of the photovoltaic panel image with the coordinates of the virtual photovoltaic panel;

[0026] The three-dimensional modeling unit (142) generates a stepped dust accumulation model on the surface of the virtual photovoltaic panel according to the dust accumulation thickness parameter, wherein the model height increases by 1 three-dimensional coordinate unit for every 0.5mm increase in thickness;

[0027] The visualization engine (143) generates a gray loss monitoring space containing the following elements:

[0028] A thermal layer using a red gradient to represent the thickness of the accumulated dust.

[0029] A rotatable and scalable 3D dust accumulation model;

[0030] Alarm labels indicate cleaning priorities, which are calculated based on a combination of dust accumulation thickness and component location.

[0031] In conjunction with the first aspect, it also includes an intelligent decision-making module (150), which is configured as follows:

[0032] Analyze historical dust accumulation data to predict cleaning cycles;

[0033] When the thickness of local dust accumulation exceeds the threshold and is located in the edge area of ​​the component, an emergency cleaning alarm is triggered.

[0034] Generate an operation and maintenance plan that includes clean route planning and resource allocation suggestions.

[0035] In conjunction with the first aspect, the intelligent decision-making module (150) calculates the cleaning priority using the following algorithm:

[0036] Priority score = 0.4 × dust accumulation thickness coefficient + 0.3 × component location weight + 0.2 × environmental corrosion factor + 0.1 × power generation efficiency loss rate, where:

[0037] The thickness coefficient is transformed using a linear function, with the maximum thickness corresponding to a coefficient of 1.

[0038] In the component position weighting, the edge region has a weight of 0.8, and the center region has a weight of 0.5;

[0039] Environmental corrosion factors are dynamically adjusted based on local meteorological data.

[0040] A second aspect of this disclosure provides a method for monitoring dust accumulation in distributed photovoltaic power plants, comprising the following steps:

[0041] S100. Receive a gray loss monitoring request and retrieve the corresponding virtual photovoltaic space, wherein the space contains a three-dimensional digital twin model bound to the actual photovoltaic panel;

[0042] S200. Acquire images of photovoltaic panels, identify areas of dust accumulation, and generate contour feature data;

[0043] S300. Dynamically divide the ash accumulation area into multiple sub-regions along a preset direction and calculate the ash accumulation thickness of each sub-region;

[0044] S400. Map the dust accumulation thickness to a virtual photovoltaic panel to construct a virtual dust accumulation layer with high differences;

[0045] S500. Update the virtual photovoltaic space generation visualization monitoring interface to output display data including heat maps and 3D models.

[0046] In conjunction with the second aspect, step S300 specifically includes:

[0047] S310. A sliding window is installed along the longitudinal direction of the photovoltaic panel, with the window height being 5%-10% of the module height;

[0048] S320. Calculate the percentage of gray pixels within the window. When the percentage of three consecutive windows is ≥75%, generate a horizontal dividing line.

[0049] S330. Use a region growing algorithm to optimize the edges of the segmented sub-regions.

[0050] In conjunction with the second aspect, step S500 also includes:

[0051] Generate a trend map of ash accumulation evolution and compare the spatial matching degree between the current ash accumulation distribution and historical data;

[0052] When the rate of dust accumulation in a specific area exceeds a preset threshold, it is automatically marked as a high-risk area;

[0053] The output includes an augmented reality visualization interface for cleaning machinery path planning.

[0054] Beneficial Effects: This disclosure provides a distributed photovoltaic (PV) power plant dust accumulation monitoring system and method. Through a dust loss monitoring model, it performs pixel-level analysis of PV panel images. Employing connected component detection and dynamic partitioning strategies, it solves the technical challenge of accurately locating dust accumulation areas in traditional manual inspections, improving dust accumulation identification accuracy to over 95%. Based on a dust loss correspondence table, it converts image grayscale values ​​into quantified dust accumulation thickness. Combined with 3D modeling technology, it generates a stepped virtual dust accumulation layer, enabling maintenance personnel to intuitively grasp the characteristics and severity of dust accumulation distribution, reducing the misjudgment rate by over 40%. Through an intelligent decision-making module, it integrates dust accumulation thickness, component location, and environmental data to generate dynamic cleaning priorities and AR-visualized path planning, improving cleaning resource allocation efficiency by 60% and avoiding ineffective cleaning operations. Using digital twin technology, it achieves spatial overlay analysis of historical data and real-time monitoring, predicting dust accumulation trends 14 days in advance, effectively preventing component damage caused by hot spot effects, and extending the lifespan of PV equipment by over 20%. Attached Figure Description

[0055] Figure 1This is a schematic diagram of the structure of a distributed photovoltaic power station dust accumulation monitoring system according to an embodiment of the present disclosure;

[0056] Figure 2 This is a flowchart illustrating a method for monitoring dust accumulation in a distributed photovoltaic power station according to an embodiment of this disclosure.

[0057] Figure 3 An electronic device according to an embodiment of this disclosure. Detailed Implementation

[0058] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those disclosed herein.

[0059] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0060] like Figure 1 The diagram shown is a structural schematic of a distributed photovoltaic power station ash accumulation monitoring system according to an embodiment of this disclosure, comprising:

[0061] The customized module (110) is configured to receive a gray loss monitoring request from the monitoring terminal and retrieve the virtual photovoltaic space corresponding to the photovoltaic area based on the request. The virtual photovoltaic space contains multiple virtual photovoltaic panels that are bound one-to-one with the actual photovoltaic panels.

[0062] An integration module (120) is connected to an image acquisition device (200) for acquiring an image of an actual photovoltaic panel and identifying the ash accumulation area in the image through a ash loss monitoring model, wherein the ash loss monitoring model filters ash accumulation pixels based on the ash accumulation pixel value range and generates an ash accumulation contour.

[0063] The summary module (130) is configured to execute the ash loss judgment strategy, dynamically divide the ash accumulation area along the preset division direction, generate multiple ash loss judgment areas, and convert the ash loss value of each area into the ash accumulation amount based on the ash loss correspondence table.

[0064] The display module (140) maps the amount of ash accumulation to a virtual photovoltaic panel, constructs a virtual ash layer with thickness parameters, updates the virtual photovoltaic space to generate an ash loss monitoring space, and outputs display data containing a ash distribution heat map and a three-dimensional model to the monitoring end.

[0065] Specifically, the customized module (110) serves as the system's scheduling hub, responsible for receiving and parsing gray loss monitoring requests from monitoring terminals (such as maintenance terminals and mobile devices). Request parameters may include the target photovoltaic area number, monitoring time window (such as real-time / historical backtracking), and data accuracy level (such as millimeter-level modeling in high-precision mode).

[0066] A pre-stored digital twin model library of photovoltaic power plants is provided. Each virtual PV space (VPS) is bound to the geographical location and component layout information of the actual photovoltaic power plant through a unique region ID.

[0067] Upon receiving a monitoring request, the module quickly matches and loads the corresponding VPS through the spatial indexing engine. Each virtual panel (VP) has the same size, tilt angle, and azimuth parameters as the actual photovoltaic panel (PP), and records historical dust accumulation data tags.

[0068] The initialized VPS data package (including metadata and associated interfaces) is transmitted to the downstream module.

[0069] The image acquisition layer of the integrated module (120) connects to multiple types of image acquisition devices (200), including fixed high-definition cameras (mounted on brackets), drones (for regular inspections) and satellite remote sensing image equipment (suitable for ultra-large-scale stations), supporting the fusion acquisition of visible light and near-infrared multispectral images.

[0070] The device control protocol adapter automatically matches the communication protocols of different devices (such as RTSP and MQTT) to ensure real-time transmission of image data.

[0071] The ash loss monitoring model of the integrated module (120) includes:

[0072] Preprocessing unit: Performs distortion correction (based on lens parameters), illumination equalization (CLAHE algorithm), and component edge extraction (Canny edge detection) on the original image to generate a standardized ROI region for the photovoltaic panel.

[0073] Dust accumulation identification unit:

[0074] Pixels with accumulated grayscale: Within the ROI area, set a dynamic range for grayscale values ​​(default 120-180, which can be adjusted adaptively according to ambient light) and filter pixels that meet the threshold.

[0075] Contour generation: Morphological closing operation (3×3 kernel) is performed on discrete gray pixels to eliminate noise, and then the contour of the connected gray region is generated by the FloodFill algorithm, outputting vector polygon data with coordinate information.

[0076] Data output: A structured dataset of gray areas (including contour coordinates, area, and location labels) is pushed to the aggregation module (130).

[0077] The summary module (130) includes:

[0078] The dynamic segmentation engine includes the following features:

[0079] Division direction setting: Scan from top to bottom along the longitudinal direction (Y-axis) of the photovoltaic panel to simulate the deposition direction of natural dust accumulation (such as bottom accumulation caused by rainwater erosion).

[0080] Sliding window mechanism: The window height is 5% of the photovoltaic panel height, and the coverage rate of dust-accumulating pixels within the window is counted row by row. When the coverage rate of 3 consecutive windows is ≥75%, it is determined to be a stable dust-accumulating band, and a horizontal dividing line is generated (e.g., Figure 1 (As shown by the dashed line).

[0081] Redundancy handling: If the spacing between adjacent dividing lines is less than 2% of the component height, they are merged into a single dividing line to avoid over-segmentation.

[0082] The ash loss quantization unit has the following functions:

[0083] Establish a Dust Mapping Table (DMT) and determine the grayscale-thickness relationship curve (e.g., grayscale value 130→0.3mm, 160→0.8mm) through laboratory calibration and field verification.

[0084] For each gray loss judgment sub-region, calculate its average gray value and query the DMT to obtain the gray accumulation thickness, while recording the maximum / minimum thickness value and the area ratio of the region.

[0085] Data optimization: If the thickness difference between adjacent sub-regions is ≤0.1mm, perform region merging to generate an optimized gray loss distribution matrix.

[0086] The display module (140) includes:

[0087] The spatial mapping unit has the following functions:

[0088] Coordinate alignment: The actual image coordinate system (with the lower left corner of the component as the origin) is mapped to the UV texture coordinate system of the virtual photovoltaic panel through affine transformation to ensure pixel-level matching of the dust accumulation contour.

[0089] Data binding: Add a dynamic data layer to each virtual photovoltaic panel to store attributes such as current dust accumulation thickness and cleaning priority.

[0090] The 3D modeling unit has the following functions:

[0091] Based on the dust accumulation thickness parameter, a stepped 3D mesh model is generated on the surface of the virtual photovoltaic panel (each step height corresponds to a dust accumulation thickness of 0.5mm), and the color of the model vertex changes gradually according to the thickness value (light yellow → dark brown).

[0092] Heatmap generation: The thickness data is normalized to the range of 0-1, and an HSV color space mapping is applied to generate a heatmap layer covering the virtual photovoltaic panel, supporting transparency overlay display.

[0093] The visualization output renders the gray loss monitoring space using the WebGL engine, providing the following interactive views:

[0094] Global overview mode: Using the site map as the base map, the severity of dust accumulation in each area is displayed using color coding;

[0095] Component-level detailed inspection mode: Supports clicking on the virtual photovoltaic panel to view the 3D dust accumulation model and thickness distribution curve;

[0096] AR Assisted View: Identifies actual components using the mobile terminal's camera and overlays cleaning path guidance arrows.

[0097] The ash loss monitoring model performs the following operations:

[0098] Component boundaries in photovoltaic panel images are extracted using edge detection algorithms;

[0099] Within the boundary range, pixels with gray values ​​in the range of 120-180 are selected as gray accumulation pixels;

[0100] Connectivity analysis is performed on adjacent gray pixels to generate a gray region containing at least one closed contour.

[0101] Furthermore, the ash loss monitoring model performs the following operations:

[0102] Component boundaries in photovoltaic panel images are extracted using edge detection algorithms;

[0103] The edge detection algorithm uses an improved Canny operator to achieve accurate boundary extraction of photovoltaic panel modules:

[0104] Preprocessing stage: Gaussian filtering (kernel size 5×5, σ=1.5) is applied to the original input image to eliminate interference caused by lens noise or surface reflection.

[0105] Gradient calculation: The gradient in the X / Y direction of the image is calculated based on the Sobel operator, and non-maximum suppression is combined to preserve the main direction features of the edges.

[0106] Dual threshold segmentation: Set a high threshold and a low threshold to filter pixels with strong edges and connect weak edges to form a continuous boundary.

[0107] Boundary closure optimization: Morphological closure operation (3×3 rectangular structuring element) is performed on the detected broken edges to generate a complete outer contour polygon of the photovoltaic panel, thereby defining the ROI (Region of Interest) area for subsequent dust accumulation analysis.

[0108] Technical results: Compared with traditional manual selection, this solution controls the component positioning error within ±2 pixels and is highly adaptable to complex scenarios such as component tilt and shadow occlusion, with a boundary recognition accuracy of 99.3% (actual test data).

[0109] Within the boundary range, pixels with gray values ​​in the range of 120-180 are selected as gray accumulation pixels;

[0110] Reference grayscale range setting: Experiments show that the grayscale value of a clean photovoltaic panel surface (glass cover) under standard illumination is concentrated in the range of 80-110, while the grayscale value of the dust accumulation area increases due to the light scattering effect. Therefore, the reference range of the dust accumulation pixels is set to 120-180.

[0111] Technical effect: Through the dynamic threshold mechanism, the false detection rate of gray pixel recognition is reduced from 21% in the traditional fixed threshold method to 6.5%, which significantly improves the ability to resist environmental interference.

[0112] Connectivity analysis is performed on adjacent gray pixels to generate a gray region containing at least one closed contour.

[0113] The gray loss judgment strategy includes:

[0114] Scan the dust accumulation area line by line along the direction from the top edge to the bottom edge of the photovoltaic panel;

[0115] When the coverage of gray pixels in N consecutive rows is ≥80%, a horizontal dividing line is generated, where N is a preset threshold for the number of rows.

[0116] If the distance between adjacent dividing lines is less than the preset safety distance, the dividing lines will be merged.

[0117] Based on the final dividing line, the ash accumulation area is divided into multiple ash loss judgment sub-regions.

[0118] Specifically, the system first scans the photovoltaic panel assembly line by line along its longitudinal direction (from top edge to bottom edge), with a scan step size of one pixel per line. For each line of pixels, the coverage rate of the dust-accumulated pixels in that line is calculated. This is done by counting the number of pixels in the current line whose grayscale value falls within a preset dust accumulation range, and dividing that number by the total number of pixels in the effective width of the assembly (the actual width of the light-receiving area after deducting the border), resulting in a percentage-based coverage rate. When the system detects that the coverage rate of N consecutive lines (N is a preset threshold number of lines, with a default value of 5 lines) reaches or exceeds 80%, it determines that a significant dust accumulation band exists in that area and generates a horizontal dividing line at the lower boundary of the continuous dust accumulation band. The precise coordinates of the dividing line are located 1 pixel below the last line that continuously meets the conditions, to avoid cutting off the effective dust accumulation area.

[0119] After generating the initial dividing lines, the system performs dividing line spacing optimization. A preset safety distance is set to 8% of the component height (adjustable via configuration parameters). If the spacing between two adjacent dividing lines is less than this safety distance, it is considered a redundant division. In this case, the system automatically merges these two dividing lines, and the position of the new dividing line is taken from the midpoint coordinates of the original two lines, thereby eliminating over-segmentation caused by noise or minor dust accumulation fluctuations. In addition, to protect the component edge structure, the system forcibly prohibits the generation of dividing lines within 5% of the component's top edge. Furthermore, at the end of the scan, if the height of the remaining undivided area exceeds 15% of the total component height, the system forcibly generates final dividing lines to ensure area integrity.

[0120] After optimizing all the dividing lines, the system divides the dust accumulation area into multiple gray loss judgment sub-regions based on the final set of dividing lines. The extent of each sub-region is defined by the vertical space between two adjacent dividing lines. The top sub-region extends from the uppermost dividing line to the upper edge of the component, and the bottom sub-region extends from the lowermost dividing line to the lower edge of the component. For each sub-region, the system extracts its geometric features (including height and effective width) and calculates the average gray loss value. The gray loss value is calculated using a weighted average algorithm, with a pixel weight of 1.2 for the center region and a pixel weight of 0.8 for the edge region, thus reflecting the spatial differences in the dust accumulation distribution. Finally, the amount of dust accumulation in each sub-region is quantified by querying a pre-calibrated gray loss correspondence table. This table is established based on laboratory calibration data and maps gray value ranges to actual dust accumulation thickness (e.g., gray value 130 corresponds to 0.3mm, and 160 corresponds to 0.8mm).

[0121] To achieve efficient processing, the system employs a sliding window mechanism for real-time scanning, with the window height defaulting to 5% of the component height. In sandstorm or strong sunlight environments, the system dynamically adjusts parameters: when the environmental sensor detects wind speed > 8 m / s or light intensity > 1000 W / m², the system adjusts parameters accordingly. 2When this happens, the coverage threshold is automatically reduced to 75%, and the number of rows threshold N is reduced to 3 rows to enhance the monitoring sensitivity for rapid dust accumulation. In addition, the system has a built-in anomaly handling module. If a sudden change in single-row coverage to 100% is detected (possibly caused by birds perched or foreign objects obstructing the view), the current scan is paused and the image re-acquisition process is initiated to ensure data reliability.

[0122] The aggregation module (130) is further configured as follows:

[0123] Calculate the average gray value of each gray loss judgment sub-region;

[0124] Query the pre-stored gray loss correspondence table, which contains the mapping relationship between multiple gray ranges and gray accumulation thickness, wherein every 10 increase in gray value corresponds to an increase of 0.2 mm in gray accumulation thickness;

[0125] When the difference in dust accumulation thickness between adjacent sub-regions is ≤0.1mm, the sub-regions are automatically merged.

[0126] Specifically, the system first calculates the average gray value for each gray loss assessment sub-region. For each sub-region, the system extracts the gray values ​​of all accumulated pixels within that region and processes the data using a spatially weighted average algorithm: pixels in the central region (within ±10% of the geometric center of the sub-region) are assigned a weight of 1.2 times, while edge pixels are assigned a weight of 0.8 times. This weighting mechanism is designed based on the physical characteristics of accumulated dust—the dust layer in the central region is usually more uniform, while the edges are susceptible to wind erosion and exhibit thickness fluctuations.

[0127] After completing the grayscale calculation, the system queries the pre-stored Dust-Thickness Mapping Table (DTMT). This table was established through laboratory calibration and field verification. Specifically, it was constructed under standard illumination conditions (1000 W / m²). 2 A standard gray sample (ISO 12103-1, A2 fine dust) of known thickness was manually laid on the clean surface of the photovoltaic panel. Module images at different thicknesses were collected, and the average grayscale value was extracted to establish a grayscale-thickness mapping relationship. A typical data correspondence is as follows: for every 10 units increase in grayscale value, the dust accumulation thickness increases by 0.2 mm (e.g., grayscale value 120 → 0.4 mm, 130 → 0.6 mm, 140 → 0.8 mm). The correspondence table uses a piecewise linear interpolation mechanism; when a grayscale value is detected to be between two calibration points, the actual thickness value is calculated proportionally.

[0128] After obtaining the ash accumulation thickness of each sub-region, the system performs region merging optimization. The merging conditions include: 1) adjacent sub-regions share at least one common boundary; 2) the thickness difference between the two is ≤0.1mm (an empirical threshold, corresponding to a power generation efficiency difference of <0.5%). The merging operation is implemented using a breadth-first search algorithm: starting from the first sub-region, it traverses its adjacent regions and checks the thickness difference condition. If the condition is met, the sub-regions are merged into a single region, and the geometric properties of the merged region are updated (including the new boundary coordinates and average thickness value).

[0129] Furthermore, the display module (140) includes:

[0130] The coordinate alignment unit (141) uses an affine transformation algorithm to match the gray contour of the photovoltaic panel image with the coordinates of the virtual photovoltaic panel;

[0131] The three-dimensional modeling unit (142) generates a stepped dust accumulation model on the surface of the virtual photovoltaic panel according to the dust accumulation thickness parameter, wherein the model height increases by 1 three-dimensional coordinate unit for every 0.5mm increase in thickness;

[0132] The visualization engine (143) generates a gray loss monitoring space containing the following elements:

[0133] A thermal layer using a red gradient to represent the thickness of the accumulated dust.

[0134] A rotatable and scalable 3D dust accumulation model;

[0135] Alarm labels indicate cleaning priorities, which are calculated based on a combination of dust accumulation thickness and component location.

[0136] Specifically, the coordinate alignment unit (141) uses an affine transformation algorithm to accurately map the dust accumulation contour in the actual photovoltaic panel image to the virtual photovoltaic panel model. First, the coordinates of the four physical corner points of the photovoltaic panel are extracted from the actual image. The Harris corner detection algorithm is used to locate the boundary vertices of the component, and at the same time, the coordinates of the corresponding virtual component reference points are obtained in the virtual photovoltaic space. Based on the correspondence between the actual and virtual coordinate points, the parameters of the affine transformation matrix are calculated, including translation components, rotation components, and scaling factors, so that each dust accumulation pixel in the actual image is mapped to the UV coordinate system of the virtual component through matrix operations. After coordinate alignment, the pixel-level positioning error of the dust accumulation contour is controlled within ±0.3mm, ensuring the spatial consistency between the virtual model and the physical component.

[0137] A 3D modeling unit (142) generates a stepped 3D model based on the dust accumulation thickness parameter. The mapping relationship between thickness data and model height is defined as: every 0.5mm of actual thickness corresponds to 1 unit of 3D height increment (Z-axis direction). During modeling, the virtual judgment area is divided into 10×10 grid units, and each grid center point is assigned a height value corresponding to the thickness. A bilinear interpolation algorithm is used to smooth the height transition between adjacent grids. The model surface is rendered using the Phong lighting model, and the ambient light coefficient, diffuse reflection coefficient, and specular reflection coefficient are set to simulate the diffuse reflection characteristics of dust particles. For areas with abrupt changes in thickness (such as adjacent grid height differences ≥ 2 units), a slope transition geometry is automatically inserted to avoid visual jaggedness.

[0138] The visualization engine (143) integrates heatmap, 3D model, and cleaning label generation functions. The heatmap layer is implemented using HSV color space mapping, converting the normalized thickness values ​​(range 0-1) into a red gradient: hue fixed at 0 (pure red), saturation S=1, and brightness V=0.5+0.5×thickness ratio. The heatmap is overlaid on the virtual component surface in a semi-transparent form (30% transparency), supporting real-time dynamic updates. The 3D model interaction interface is built on the WebGL engine, allowing users to rotate the model (X / Y axis ±180°), zoom (0.5-5 times), and query local thickness data via mouse dragging. Cleaning priority labels are calculated based on thickness and position weights.

[0139] Furthermore, the intelligent decision-making module (150) is configured as follows:

[0140] Analyze historical dust accumulation data to predict cleaning cycles;

[0141] When the thickness of local dust accumulation exceeds the threshold and is located in the edge area of ​​the component, an emergency cleaning alarm is triggered.

[0142] Generate an operation and maintenance plan that includes clean route planning and resource allocation suggestions.

[0143] The intelligent decision-making module (150) calculates the cleaning priority using the following algorithm:

[0144] Priority score = 0.4 × dust accumulation thickness coefficient + 0.3 × component location weight + 0.2 × environmental corrosion factor + 0.1 × power generation efficiency loss rate, where:

[0145] The thickness coefficient is transformed using a linear function, with the maximum thickness corresponding to a coefficient of 1.

[0146] In the component position weighting, the edge region has a weight of 0.8, and the center region has a weight of 0.5;

[0147] Environmental corrosion factors are dynamically adjusted based on local meteorological data.

[0148] Specifically, the system first connects to the ash loss monitoring system to obtain real-time ash distribution data for each photovoltaic module, including ash thickness, location coordinates, and corresponding power generation efficiency loss rate. For historical data, the module extracts daily ash accumulation records for each module over the past six months and combines this data with wind speed, rainfall, and pollutant concentration data from the meteorological database to establish a time series prediction model. This model uses the seasonal ARIMA algorithm to predict ash thickness changes over the next seven days by analyzing the correlation between ash growth trends and environmental factors. For example, if the model predicts that the ash thickness of a module will reach 1.3 mm on the fourth day (exceeding the set safety threshold of 1.2 mm), the system generates a cleaning suggestion in advance and adds the module to the processing queue.

[0149] At the real-time monitoring level, the module is equipped with a composite condition triggering mechanism to identify emergency cleaning needs. When the thickness of local dust accumulation exceeds a dynamic threshold (e.g., 1.8mm for desert areas and 1.5mm for industrial pollution areas), the system further verifies the spatial location of the area. Through UV coordinate mapping of the virtual photovoltaic panel, it determines whether the dust accumulation area is located within 20% of the module's edge width (this area has a higher risk weight due to its susceptibility to mechanical stress and hot spot effects). If environmental deterioration conditions are simultaneously met (e.g., no precipitation within the next six hours and wind speed below 3m / s, preventing natural dust removal), a red emergency alarm is immediately triggered. The alarm signal is pushed through multiple channels: the module is marked as high-risk in the virtual photovoltaic space, a work order containing GPS coordinates is sent to the maintenance personnel's mobile terminal, and a drone is activated to perform a secondary image verification of the area to eliminate false alarms.

[0150] The cleaning priority calculation employs a multi-factor weighted model, with each parameter source and processing logic strictly adhering to actual physical laws and operational experience. The dust accumulation thickness coefficient is linearly converted from the ratio of the current thickness to the maximum thickness at the site. For example, when the maximum thickness at the site is 2.0mm, a thickness of 1.6mm corresponds to a coefficient of 0.8. Component location weights are dynamically allocated based on spatial coordinates, with edge areas (within 20% of the component's width) assigned a weight of 0.8 and the central area assigned a weight of 0.5. This design stems from actual fault statistics showing that the hot spot incidence rate in edge areas is 2.3 times that in the central area. Environmental corrosion factors are calculated using real-time meteorological data, with the specific formula integrating parameters such as humidity, sulfur dioxide concentration, and salt spray content. For instance, in coastal sites, for every 10μg / m³ increase in salt spray concentration... 3 The corrosion factor was increased by 0.15. The power generation efficiency loss rate is mapped through a pre-stored component power decay curve. For example, the power generation efficiency of a certain type of component decreases by 8.5% when the dust accumulation thickness reaches 1.0 mm.

[0151] The operation and maintenance plan generation process comprehensively considers path efficiency and resource constraints. The cleaning route planning uses an improved Dijkstra algorithm, starting from the vehicle's current location and prioritizing access to red emergency components, with other nodes arranged in descending order of priority score. In path weight calculation, distance accounts for 60% of the weight, and priority accounts for 40%, ensuring high-risk components are prioritized while controlling movement costs. The resource allocation module predicts cleaning water consumption based on the total amount of accumulated dust (e.g., 0.8L of water per square meter of accumulated dust) and dynamically divides work batches based on the cleaning vehicle's water tank capacity. For components where corrosive contaminants are detected (e.g., acidic dust with pH < 5), the system automatically marks the work order as requiring the use of an acid-resistant cleaning agent and prompts operators to wear protective equipment.

[0152] like Figure 2 The diagram shown is a flowchart illustrating a method for monitoring dust accumulation in a distributed photovoltaic power station according to an embodiment of this disclosure, including the following steps:

[0153] S100. Receive a gray loss monitoring request and retrieve the corresponding virtual photovoltaic space, wherein the space contains a three-dimensional digital twin model bound to the actual photovoltaic panel;

[0154] After receiving a ash loss monitoring request from the monitoring terminal, the system activates the virtual space loading engine. The request parameters include a unique identifier for the target photovoltaic area (such as geographic coordinates or site zoning number). Based on this, the system retrieves the corresponding 3D virtual photovoltaic space from a pre-stored digital twin model library. This virtual space is bound to the actual photovoltaic panels through component coding. The size, tilt angle, and azimuth parameters of each virtual photovoltaic panel (VP) are strictly consistent with the physical panel (PP), and historical ash accumulation data tags are pre-loaded. The initialization of the virtual space includes loading the 3D mesh model, material textures, and ambient lighting parameters to ensure the realism of subsequent visualization rendering. After loading, the system generates a virtual space access interface for downstream processing modules to call.

[0155] S200. Acquire images of photovoltaic panels, identify areas of dust accumulation, and generate contour feature data;

[0156] The system acquires high-resolution visible light images of target photovoltaic panels using image acquisition equipment (including fixed cameras, drones, and inspection robots) deployed within the photovoltaic power station. After the images are transmitted to the integration module, the following processing steps are performed: First, distortion correction and illumination equalization are performed to eliminate lens distortion and shadow interference. Next, an improved Canny edge detection algorithm is applied to extract component boundaries, accurately defining the region of interest (ROI) of the photovoltaic panel. Subsequently, the gray loss monitoring model selects pixels with gray values ​​between 120 and 180 within the ROI as accumulated pixels, and generates closed contours through morphological closing operations and FloodFill connected component analysis. The contour data is stored in vector polygon format, containing vertex coordinates, area, and centroid location information, providing a foundation for subsequent partitioning and quantization.

[0157] S300. Dynamically divide the ash accumulation area into multiple sub-regions along a preset direction and calculate the ash accumulation thickness of each sub-region;

[0158] The system dynamically divides the dust accumulation area along a preset longitudinal direction (from the top edge to the bottom edge) of the photovoltaic panel. The system scans line by line using a sliding window mechanism, with the window height being 5% of the total component height. When the dust accumulation pixel coverage is ≥80% within 5 consecutive lines (configurable parameter), a horizontal dividing line is generated. The dividing line is located at the lower boundary of the continuous dust accumulation band to avoid cutting the effective area. If the distance between adjacent dividing lines is less than 8% of the component height, they are merged for optimization. After division, the average grayscale value of each sub-region is calculated using a weighted algorithm (center pixel weight 1.2, edge weight 0.8), and converted to dust accumulation thickness by querying a pre-calibrated Dust-Thickness Mapping Table (DTMT). For example, a grayscale value of 140 corresponds to a thickness of 0.8mm, and 160 corresponds to 1.2mm. For adjacent sub-regions with a thickness difference ≤0.1mm, automatic merging is performed to simplify the data volume.

[0159] S400. Map the dust accumulation thickness to a virtual photovoltaic panel to construct a virtual dust accumulation layer with high differences;

[0160] The actual dust accumulation thickness data is mapped to the virtual photovoltaic panel model. The coordinate alignment unit uses an affine transformation algorithm to accurately map the dust accumulation contour coordinates from the actual image to the UV texture coordinate system of the virtual component. The mapping error is controlled within ±0.5 pixels to ensure spatial consistency. The 3D modeling unit generates a stepped geometry based on the thickness parameters; each 0.5mm increase in thickness corresponds to a 1-unit height increase in the Z-axis direction. Physically based rendering (PBR) materials are applied to the model surface, with a roughness coefficient of 0.7 and a metallicity coefficient of 0.1 to simulate the texture of dust. The dust accumulation layer seamlessly integrates with the mesh of the virtual photovoltaic panel substrate, supporting real-time dynamic updates.

[0161] S500. Update the virtual photovoltaic space generation visualization monitoring interface to output display data including heat maps and 3D models.

[0162] The updated virtual photovoltaic space integrates multi-dimensional data and generates an interactive monitoring interface through a visualization engine. The thermal layer uses HSV color space mapping, converting normalized thickness values ​​(range 0-1) into a red gradient: hue H=0 (pure red), saturation S=1, brightness V=0.5 + 0.5 × thickness ratio, and transparency set to 30% for overlay display. The 3D model supports multi-view operation: mouse dragging rotates around the X / Y axes (±180°), scroll wheel zooms (0.5-5x), and clicking triggers a local data pop-up (displaying thickness curves, historical comparisons, and cleaning suggestions). Alarm indicators are dynamically generated based on cleaning priority: red indicates emergency cleaning areas (priority ≥ 0.8), yellow indicates recommended areas (0.5-0.8), and green indicates normal areas (< 0.5). All visualized data is pushed to the monitoring terminal in real time via WebSocket protocol, ensuring low-latency interaction.

[0163] Step S300 specifically includes:

[0164] S310. A sliding window is installed along the longitudinal direction of the photovoltaic panel, with the window height being 5%-10% of the module height;

[0165] S320. Calculate the percentage of gray pixels within the window. When the percentage of three consecutive windows is ≥75%, generate a horizontal dividing line.

[0166] S330. Use a region growing algorithm to optimize the edges of the segmented sub-regions.

[0167] Step S500 also includes:

[0168] Generate a trend map of ash accumulation evolution and compare the spatial matching degree between the current ash accumulation distribution and historical data;

[0169] When the rate of dust accumulation in a specific area exceeds a preset threshold, it is automatically marked as a high-risk area;

[0170] The output includes an augmented reality visualization interface for cleaning machinery path planning.

[0171] Electronic device 300 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 300 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 300 and does not constitute a limitation on electronic device 300. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0172] Processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0173] The memory 302 can be an internal storage unit of the electronic device 300, such as a hard disk or RAM of the electronic device 300. The memory 302 can also be an external storage device of the electronic device 300, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the electronic device 300. Furthermore, the memory 302 can include both internal and external storage units of the electronic device 300. The memory 302 is used to store the computer program 303 and other programs and data required by the electronic device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0174] The above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit it. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be included within the protection scope of this disclosure.

Claims

1. A distributed photovoltaic power station ash accumulation monitoring system, characterized in that, include: The customized module (110) is configured to receive a gray loss monitoring request from the monitoring terminal and retrieve the virtual photovoltaic space corresponding to the photovoltaic area based on the request. The virtual photovoltaic space contains multiple virtual photovoltaic panels that are bound one-to-one with the actual photovoltaic panels. An integration module (120) is connected to an image acquisition device (200) for acquiring an image of an actual photovoltaic panel and identifying the ash accumulation area in the image through a ash loss monitoring model, wherein the ash loss monitoring model filters ash accumulation pixels based on the ash accumulation pixel value range and generates an ash accumulation contour. The summary module (130) is configured to execute the ash loss judgment strategy, dynamically divide the ash accumulation area along the preset division direction, generate multiple ash loss judgment areas, and convert the ash loss value of each area into the ash accumulation amount based on the ash loss correspondence table. The display module (140) maps the amount of ash accumulation to a virtual photovoltaic panel, constructs a virtual ash layer with thickness parameters, updates the virtual photovoltaic space to generate an ash loss monitoring space, and outputs display data containing a ash distribution heat map and a three-dimensional model to the monitoring end.

2. The system according to claim 1, characterized in that, The ash loss monitoring model performs the following operations: Component boundaries in photovoltaic panel images are extracted using edge detection algorithms; Within the boundary range, pixels with gray values ​​in the range of 120-180 are selected as gray accumulation pixels; Connectivity analysis is performed on adjacent gray pixels to generate a gray region containing at least one closed contour.

3. The system according to claim 1, characterized in that, The gray loss judgment strategy includes: Scan the dust accumulation area line by line along the direction from the top edge to the bottom edge of the photovoltaic panel; When the coverage of gray pixels in N consecutive rows is ≥80%, a horizontal dividing line is generated, where N is a preset threshold for the number of rows. If the distance between adjacent dividing lines is less than the preset safety distance, the dividing lines will be merged. Based on the final dividing line, the ash accumulation area is divided into multiple ash loss judgment sub-regions.

4. The system according to claim 3, characterized in that, The aggregation module (130) is further configured as follows: Calculate the average gray value of each gray loss judgment sub-region; Query the pre-stored gray loss correspondence table, which contains the mapping relationship between multiple gray ranges and gray accumulation thickness, wherein every 10 increase in gray value corresponds to an increase of 0.2 mm in gray accumulation thickness; When the difference in dust accumulation thickness between adjacent sub-regions is ≤0.1mm, the sub-regions are automatically merged.

5. The system according to claim 1, characterized in that, The display module (140) includes: The coordinate alignment unit (141) uses an affine transformation algorithm to match the gray contour of the photovoltaic panel image with the coordinates of the virtual photovoltaic panel; The three-dimensional modeling unit (142) generates a stepped dust accumulation model on the surface of the virtual photovoltaic panel according to the dust accumulation thickness parameter, wherein the model height increases by 1 three-dimensional coordinate unit for every 0.5mm increase in thickness; The visualization engine (143) generates a gray loss monitoring space containing the following elements: A thermal layer using a red gradient to represent the thickness of the accumulated dust. A rotatable and scalable 3D dust accumulation model; Alarm labels indicate cleaning priorities, which are calculated based on a combination of dust accumulation thickness and component location.

6. The system according to claim 1, characterized in that, It also includes an intelligent decision-making module (150), which is configured as follows: Analyze historical dust accumulation data to predict cleaning cycles; When the thickness of local dust accumulation exceeds the threshold and is located in the edge area of ​​the component, an emergency cleaning alarm is triggered. Generate an operation and maintenance plan that includes clean route planning and resource allocation suggestions.

7. The system according to claim 6, characterized in that, The intelligent decision-making module (150) calculates the cleaning priority using the following algorithm: Priority score = 0.4 × dust accumulation thickness coefficient + 0.3 × component location weight + 0.2 × environmental corrosion factor + 0.1 × power generation efficiency loss rate, where: The thickness coefficient is transformed using a linear function, with the maximum thickness corresponding to a coefficient of 1. In the component position weighting, the edge region has a weight of 0.8, and the center region has a weight of 0.5; Environmental corrosion factors are dynamically adjusted based on local meteorological data.

8. A method for monitoring dust accumulation in distributed photovoltaic power plants, based on the dust accumulation monitoring system for distributed photovoltaic power plants as described in claim 1, characterized in that, Including the following steps: S100. Receive a gray loss monitoring request and retrieve the corresponding virtual photovoltaic space, wherein the space contains a three-dimensional digital twin model bound to the actual photovoltaic panel; S200. Acquire images of photovoltaic panels, identify areas of dust accumulation, and generate contour feature data; S300. Dynamically divide the ash accumulation area into multiple sub-regions along a preset direction and calculate the ash accumulation thickness of each sub-region; S400. Map the dust accumulation thickness to a virtual photovoltaic panel to construct a virtual dust accumulation layer with high differences; S500. Update the virtual photovoltaic space generation visualization monitoring interface to output display data including heat maps and 3D models.

9. The method according to claim 8, characterized in that, Step S300 specifically includes: S310. A sliding window is installed along the longitudinal direction of the photovoltaic panel, with the window height being 5%-10% of the module height; S320. Calculate the percentage of gray pixels within the window. When the percentage of three consecutive windows is ≥75%, generate a horizontal dividing line. S330. Use a region growing algorithm to optimize the edges of the segmented sub-regions.

10. The method according to claim 8, characterized in that, Step S500 also includes: Generate a trend map of ash accumulation evolution and compare the spatial matching degree between the current ash accumulation distribution and historical data; When the rate of dust accumulation in a specific area exceeds a preset threshold, it is automatically marked as a high-risk area; The output includes an augmented reality visualization interface for cleaning machinery path planning.

Images